“Failures” to be celebrated: an analysis of major pivots of software startups

Abstract

In the context of software startups, project failure is embraced actively and considered crucial to obtain validated learning that can lead to pivots. A pivot is the strategic change of a business concept, product or the different elements of a business model. A better understanding is needed on different types of pivots and different factors that lead to failures and trigger pivots, for software entrepreneurial teams to make better decisions under chaotic and unpredictable environment. Due to the nascent nature of the topic, the existing research and knowledge on the pivots of software startups are very limited. In this study, we aimed at identifying the major types of pivots that software startups make during their startup processes, and highlighting the factors that fail software projects and trigger pivots. To achieve this, we conducted a case survey study based on the secondary data of the major pivots happened in 49 software startups. 10 pivot types and 14 triggering factors were identified. The findings show that customer need pivot is the most common among all pivot types. Together with customer segment pivot, they are common market related pivots. The major product related pivots are zoom-in and technology pivots. Several new pivot types were identified, including market zoom-in, complete and side project pivots. Our study also demonstrates that negative customer reaction and flawed business model are the most common factors that trigger pivots in software startups. Our study extends the research knowledge on software startup pivot types and pivot triggering factors. Meanwhile it provides practical knowledge to software startups, which they can utilize to guide their effective decisions on pivoting.

Full text

BFailures^to be celebrated: an analysis of major pivots
of software startups
Sohaib Shahid Bajwa
1
&Xiaofeng Wang
1
&
Anh Nguyen Duc
2
&Pekka Abrahamsson
2
#Springer Science+Business Media New York 2016
Abstract In the context of software startups, project failure is embraced actively and consid-
ered crucial to obtain validated learning that can lead to pivots. A pivot is the strategic change
of a business concept, product or the different elements of a business model. A better
understanding is needed on different types of pivots and different factors that lead to failures
and trigger pivots, for software entrepreneurial teams to make better decisions under chaotic
and unpredictable environment. Due to the nascent nature of the topic, the existing research
and knowledge on the pivots of software startups are very limited. In this study, we aimed at
identifying the major types of pivots that software startups make during their startup processes,
and highlighting the factors that fail software projects and trigger pivots. To achieve this, we
conducted a case survey study based on the secondary data of the major pivots happened in 49
software startups. 10 pivot types and 14 triggering factors were identified. The findings show
that customer need pivot is the most common among all pivot types. Together with customer
segment pivot, they are common market related pivots. The major product related pivots are
zoom-in and technology pivots. Several new pivot types were identified, including market
zoom-in, complete and side project pivots. Our study also demonstrates that negative customer
Empir Software Eng
DOI 10.1007/s10664-016-9458-0
Communicated by: Magne Jørgensen, Mika Mäntylä, Paul Ralph and Hakan Erdogmus
*Sohaib Shahid Bajwa
bajwa@inf.unibz.it
Xiaofeng Wang
xiaofeng.wang@unibz.it
Anh Nguyen Duc
anhn@idi.ntnu.no
Pekka Abrahamsson
pekkaa@ntnu.no
1
Faculty of Computer Science, Free University of Bozen –Bolzan, Bozen (Bolzano), Italy
2
Department of Computer and Information Science, Norwegian University of Science and
Technology, Trondheim, Norway

reaction and flawed business model are the most common factors that trigger pivots in
software startups. Our study extends the research knowledge on software startup pivot types
and pivot triggering factors. Meanwhile it provides practical knowledge to software startups,
which they can utilize to guide their effective decisions on pivoting.
Keywords Pivot .Software startups .Lean startup .Validated learning .Pivoting factor .Side
project
1 Introduction
Startups are human institutions that create innovative products or services and search for
sustainable business models under extreme uncertainty (Blank 2005;Ries2011). Software
startups are startups that build software-intensive products/services. Similar to established
software companies in which software development projects have a reputation for failure
(Savolainen et al. 2011), projects in software startups do fail as well. The consequence of project
failureforasoftwarestartupcanbeevenmoresevere than that for an established software
company. This is because a majority of software startups are focused on one single project at a
time. One project failure could put a software startup out of business (Giardino et al. 2016).
However, interestingly, failure is treated with positive attitude in software startups, to the
extent that a few companies have the practice of celebrating each failed project, such as
Supercell, according to some anecdotal evidence (Kelly 2013). Why do software startups
embrace and even celebrate failures? Since the environments of software startups are extreme-
ly unpredictable and even chaotic, failures are considered a crucial way (sometimes the only
way) for them to obtain important learning to validate key assumptions they make about their
software products and business (Eisenmann et al. 2012). In fact, the ultimate goal of these
intermediate failures along the way is to avoid the final fatal failures that put startups out of
business. These intermediate failures are what we focused on in our study.
The validated learning obtained through failing fast and failing often leads software startups
to making the strategic change of a business concept, product, or different elements of a
business model. This type of change is called pivot in the Lean Startup approach (Ries 2011). It
is claimed that pivot is the most frequently occurring commonality among different successful
startups (Ries 2011). Pivot is inevitable for almost all software startups to survive, grow and
eventually obtain sustainable business models. Only a few startups get their business models
right immediately. It is evidenced by the fact that many successful software startups did not
turn out to be what they had initially started with. For instance, Flickr used to be an online
multiplayer role playing game rather than a photo managing and sharing service (Nazar 2013),
while Twitter was initially developed as a podcast service, not a microblogging service
(Carlson 2011).
The importance of pivot for software startups deserves research attention. However, due to
the nascent nature of the research on software startups, there is a scarcity of studies on pivots in
software startups. Comprehensive and valid knowledge is yet to be built on what trigger
software startups to pivot, how and why they make certain pivot decisions, and how they
actually pivot. The study reported in this paper is one of the first attempts to fill this knowledge
gap. The objective of the study is to lay the foundations for future studies on software startup
pivots by providing the basic understanding of pivots in software startups. The basic under-
standing includes the factors that trigger software startups to pivot, and the major types of
Empir Software Eng

pivots that software startups make when failures happen. To this end, the research questions
that guided the study are phrased as following:
RQ1: What are the factors that trigger software startups to pivot?
RQ2: What are the types of pivots software startups undertake?
To answer the research questions, we employed a systematic research process. We collected
online materials as secondary data and analysed the major pivots in 49 software startups
reported in these materials, including the well-known companies such as YouTube, Flickr,
Pinterest and Twitter. The online materials allowed us to quickly obtain useful data on as many
significant pivots in software startups as possible. Based on the analysis of the pivots in these
49 software startups, we extracted a list of factors that triggered them to pivot, and identified a
set of major types of pivots they conducted. To better structure the triggering factors and pivot
types, we categorized them into different groups respectively.
The rest of this paper is organized as follows: in Section 2, the background literature and
related work are reviewed. Section 3describes the research approach employed in the study.
The research findings are presented in detail in Section 4, and further discussed in Section 5.
The paper is summarized in Section 6, which also outlines the future research.
2 Background and Related Work
2.1 Software Startups and Lean Startup
Software startups are challenging endeavours. A systematic mapping study (Paternoster et al.
2014) reveals the most frequently reported contextual features of a software startup: general
lack of resources, high reactiveness and flexibility, intense time pressure, uncertain conditions,
and tackling dynamic and fast growing markets. Software startups are dealing with various
difficulties constantly emerging from different directions. Some of the top challenges include
developing technologically innovative software products that require novel development tools
and techniques, defining minimum viable product to capture and evaluate the riskiest assump-
tions that might fail a business concept, and discovering an appropriate business strategy to
deliver value (Giardino et al. 2015).
Inspired by the lean principles from Toyota manufacturing and production system
(Womack et al. 1990), Ries (2011) presents a new approach of entrepreneurship and innova-
tion –referred to as Lean Startup. Lean Startup focuses on the efforts that create value to
customers and eliminate waste during the development phase. However, since customers are
often unknown, what they could perceive as value is also unknown. Therefore, entrepreneurs
should Bget out of the building^to discover customers from day one (Blank 2013). Instead of
emphasizing on a business plan, Lean Startup advocates to build the product iteratively and
deliver to the market for earlier feedback. The core activity of any lean startup is based on the
Build-Measure-Learn (BML) loops, through which a startup turns an idea into a product,
measures customer response, and then learns. This can be done through developing minimum
viable products (MVP). This learning is referred to as validated learning, where each hypoth-
esis on a business model is validated, and then a decision is made on whether to pivot or
persevere. Therefore, Lean Startup is also referenced as hypothesis-driven entrepreneurship
(Eisenmann et al. 2012).
Empir Software Eng

2.2 Failure in Software Startups
In Software Engineering literature, software project failures are defined in terms of cost and
schedule over-runs, project cancellations, and lost opportunities for the organizations that
embark on the difficult journey of software development (Linberg 1999). While software
project failures can lead to business failures in established companies, it is not so drastic as in
the context of startup where one failed project could put a startup company out of business
(Giardino et al. 2016), the eventual failure of a startup when it passes a point of no return,
leading to the termination of business. Software startup failure rate can be as high as 75 % to
90 % (Nobel 2011;Marmeretal.2011).
The essence of Lean Startup methodology is to help startups make early and cheap failures
as often as possible, and learn from these intermediate failures in order to avoid final
catastrophes (Ries 2011). These intermediate failures that occur during the courses of startup
processes are the ones that should be embraced actively and that can lead to pivoting, therefore
the focus of our study. The related work is also reviewed based on this perspective on failure in
software startups. Learning from this type of failures is crucial. However there is a paucity of
studies on failures in software startups. One exception is Giardino et al. (2014), in which two
software startup failures were documented and the reasons identified. One main reason is not
changing directions when they were needed, or in other words, necessary pivots were not taken
at due time. This issue is echoed in the study of Shepherd et al. (2009). According to them, the
moment of failure is not always that straightforward. Sometimes entrepreneurs decide to
continue the business even though the situation is hopeless. However, in some other cases,
entrepreneurs do pivot and improve their entrepreneurial learning experience, but there is no
study that investigates factors that trigger pivots in software startups, to the best knowledge of
the authors.
2.3 Pivots in Software Startups
Pivot is often considered the synonym of change. However, it is not about introducing just any
change and making any decision. Several definitions of pivot are presented in literature in
recent years. Pivot is considered as validating a hypothesis related to a business model (Blank
2005;Maurya2012), even though it is not compulsory that a pivot can only be related to a
business model. According to Ries (2011), a pivot is a special kind of change designed to test
and validate the assumptions about a product, business model and the engine of growth. Based
on these definitions, for this study we define a pivot as a strategic decision which leads to the
significant change to one or more,but not all,elements of a startup:product,entrepreneurial
team,business model or engine of growth. When all of these elements change at the same time,
it is not considered a pivot but starting a completely new and different business.
Previous research on pivots in software startups is limited (Paternoster et al. 2014).
Bosch et al. (2013) offer an alternative to pivot or persevere i.e., to abandon the idea,
by presenting a software development model for early stage software startups. How-
ever, the study is not primarily focused on pivoting. The study by Van der Van and
Bosch (2013) describes pivots that software startups have made, and couples them
with architecture decisions. It compares pivots and software architecture decisions in
developing a new product, and presents the similarities and differences between these
two types of decisions. According to the study, both pivots and software architecture
decisions consider risk as a triggering factor in making a decision, while the focuses
Empir Software Eng

of pivots and software architecture decisions are different. This study considers a
pivot an example of business decisions only, and does not consider product related
pivots. Another study by Hirvikoski (2014) provides an overview of how software
startups pivoted historically through the examples of Twitter, Google and Facebook.
The study argues that most successful startups have made multiple pivots during their
journey. However the pivot examples are not based on empirical data, and the study
does not shed lights on why startups pivot. Moreover, it lacks rigorous scientific
argumentation. Terho et al. (2015) identify different pivot types (product zoom-out,
customer segment, business architecture etc.), and explain how they affect the differ-
ent parts of the lean canvas model. However, there is a lack of information on how a
pivot is identified and categorized under a specific pivot category.
There is a scarcity in the literature to identify major pivots that software startups
have made. In order to ground our research on some basis, we used the pivot types
reported in Ries (2011), with the awareness that these types are subject to systematic
and scientific validation. Ries (2011) presents ten different types of pivots that can
happen in startups:
&Zoom-in Pivot: A single feature of a product becomes the whole product, such as a chatting
feature of an online game becomes a stand-alone messager app.
&Zoom-out Pivot: Opposite to zoom-in pivot, a whole product becomes a single feature of a
much larger product. For example, a photo-sharing app is extended to an social media
platform for photographers.
&Customer Segment Pivot: It is to shift from one customer segment to another, e.g. a
training app orginally targetting at professional atheletes later on at amateurs, because a
product hypothesis is partially confirmed, solving the right problem but for different
customers than initially anticipated.
&Customer Need Pivot: As a result of getting to know customers extremely well,
sometimes one realizes that the problem they are trying to solve is not important
for the customers, but they often discover other related problems that are impor-
tant for them and can be solved.
&Platform Pivot: It refers to change from an application to its supporting platform
or vice versa, e.g., shifting from an online shop to a platform that hosts online
shops.
&Business Architecture Pivot: In this pivot, a startup switches business architecture e.g.
going for high margin, low volume instead of focusing on mass market.
&Value Capture Pivot: The methods that capture the value a company creates are
commonlyreferredtoasmonetizationorrevenue models. A startup can capture
value it creates through different ways. An example of value capture pivot can be
an online service changing from freemium price model to monthly subscription fee
model.
&Engine of Growth Pivot: Typical growth engines for startups are viral (through word-of-
mouth), sticky (attracting users to stay with a product/service as long as possible) and paid
growth models. A startup changes its growth strategy to seek rapid and more profitable
growth.
&Channel Pivot: It is a recognition that a startup company has identified a way to reach their
customers more effective than their previous one, e.g., from selling a product/service via
post mails to selling on online shops.
Empir Software Eng

&Technology Pivot: A startup delivers the same solution by using completely different
technology, e.g. an app shifting from iOS to Android platform.
In addition to these pivot types, Hirvikoski (2014) proposes a new one - social pivot,
Bwhere active changes in social factors,such as persons and environments,change the
direction of a company.^Similarly, there is a severe lack of scientific argumentation and
examples that support this new pivot type.
3ResearchApproach
Thisstudyisdesignedtobeexploratoryduetothenascentnatureofthesoftware
startup research area and the limited number of previous studies on pivots. In order to
quickly obtain useful data and understand the directions of inquiry in future primary
research, we decided to use secondary data on the software startup pivot examples
that we could find on different websites, to develop an initial understanding of the
phenomenon under the study.
Secondary data means that the research data is either collected by individuals other
than the researchers who conduct the study, or for any other purposes than the one
currently being considered, or often can be a combination of these two (Vartanian 2011).
There can be several sources of secondary data e.g. census, magazines, newspapers,
blogs, reports etc. The advantage of using secondary data is that data collection process
can be fast, and inexpensive (Vartanian 2011). Used with care and diligence, secondary
data can provide a cost-effective way of gaining an initial understanding of research
questions. Secondary data analysis is also considered a starting point for other research
methods, often helpful in designing subsequent primary research and can provide a
baseline with which to compare the primary data analysis results (Boslaugh 2007). The
use of secondary data is quite common in other disciplines, such as psychology
(Trzesniewski et al. 2011),anditisalsobecomingasuitableapproachinthesoftware
engineering research community (e.g. Wang et al. 2012).
The overall research methodology employed in this study can be considered the case survey
method (Yin and Heald 1975; Cruzes et al. 2015), since it Bworks best when the studies consist
of a heterogeneous collection of case studies^(in our case, a collection of software startup
pivot examples), and the researchers’main task is Bto aggregate the characteristics, but not
necessarily the conclusions, of these cases^(Yin and Heald 1975, p. 371). The case survey
method enables the researchers to note the various experiences found in each case and then to
aggregate the frequency of occurrence of these experiences, therefore ensures the analysis of
qualitative evidence in a reliable manner. Cruzes et al. (2015) list case survey as one of the
methods for the synthesis of qualitative and mixed-methods evidence that can be applied in
Software Engineering research. To further ensure the data collection and analysis process is
systematic and reliable, we adopted the systematic literature review guidelines by Kitchenham
(2007) and adapted them to our context. We developed a secondary data search and analysis
protocol. The protocol helps to reduce the researchers’bias, because without using it, it is
possible that the data collection or analysis could be driven by researcher expectations
(Kitchenham 2007).
The overall data collection and analysis process employed in the study is illustrated in
Fig. 1and explained in detail in the following text.
Empir Software Eng

3.1 Data Collection Steps
Step 1. Define and refine search keywords
The first step of the data collection was to define the search keywords used
to search the secondary data. Based on the main objectives and research
questions, we brainstormed the initial set of search keywords. The search string
was structured using the guidelines given by Kitchenham (2007). To ensure that
we captured the keywords related to software startups, we consulted the search
string used in a systematic mapping study regarding software startups
(Paternoster et al. 2014). We conducted several trial searches, observed the
Fig. 1 The data collection and analysis process
Empir Software Eng

search results, and refined the search string subsequently. As a result, the
following final search string was formulated:
}startup}OR }start−up}OR }early−stage firm}OR }early stage firm}OR
}early−stage compan*}OR }early stage compan*}OR }venture}
0
@1
AAND }pivot}
We us ed Bstartup^rather than Bsoftware startup^due to the fact that often
software startups are described as Internet startups, tech startups or simply
startups in online sources.
We are aware that we may miss some sources that represent pivots but do not
use this specific term. However this risk was mitigated by the fact that, together
with the Lean Startup movement, pivot became a commonly acknowledged and
used term in startup communities.
1.2
Step 2. Apply search keywords to Google search engine
To search online sources, Google search engine was used through Chrome
browser. To avoid the influence of geographical location on the search results the
website www.google.com was used. The search was conducted by one researcher.
Before starting the search process, the researcher deleted the search history in the
Chrome browser, cleared browser cache, and logged out from his personal Google
account. The intention of these steps is to ensure the least possible influence of
personal and historical data on the search results. In the Google search settings,
BGoogle instant predictions^was turned off, and B100 results/links per page^was
enabled.
The search was conducted on the first author’s laptop on February 26th, 2016.
The search resulted in 1,070,000 hits. However, Google search engine does not show
more than 1,000 results per search query (Jerkovic 2010). In addition, it omits the
results that it considers similar or duplicates by default. We disabled this omit option
to include those results to be analysed manually later. As a result, in total 783 results
were displayed and eventually accessible.
Step 3. Export search results
The search results needed to be exported in order to be analysed by multiple
researchers. For this reason, the first author installed SEOQuake
3
plugin to his
Chrome browser, which automatically exported the search results (in the format of
URLs) into an Excel file. This is the Search Results Collection Awhich contains 783
URLs and each points to a webpage.
Step 4. Apply inclusion/exclusion criteria to Search Results Collection A
To select the webpages that contain relevant and reliable content for this study, we
applied a set of inclusion/exclusion criteria to Search Results Collection A.
The inclusion criteria are:
&The URL is working, and freely available (or accessible)
&The topic of the webpage is about pivoting in startup context
&The webpage contains examples of startup pivots
1
https://www.startupgrind.com/blog/is-pivot-the-new-fail/
2
http://www.inc.com/alan-spoon/what-pivot-really-means.html
3
http://www.seoquake.com
Empir Software Eng

&The pivot examples are coming from software-based startups
&The webpage is in English
The exclusion criteria are:
&The webpage contains the duplicated content of a previously examined webpage
&The webpage is non text-based (e.g. videos, audios, or images)
&The webpage on Slideshare, Quora, LinkedIn, personal (or company) blogs
We excluded Slideshare because of the synthetic content and lack of contextual
information, whereas webpages from Quora, LinkedIn and personal (or company)
blogs are excluded for the potential subjectivity in the content.
To decide if a startup is software startup and if the pivot described is a real
pivot, we used the definition of software startup (defined in the beginning of
Section 1) and pivot (defined in Section 2.3) to guide the inclusion/exclusion
step.
The first two authors conducted this step separately. The evaluation results
from the two researchers were compared and the disagreed items (5 %, 39
were discussed out of 783 URLs) were discussed between the two researchers
until a consensus was achieved. This step resulted in the Search Results
Collection B which contains 138 URLs and represents 138 webpages.
Step 5. Identify cases from Search Results Collection B
We read through the content of the 138 webpages, and looked for the information
about the software startups that pivoted during their startup processes. We considered
each mentioned software startup a potential case for further analysis. Since this step
was relatively objective and straightforward, it was mainly conducted by the first
author. In the case of doubt, the second author was consulted. This step resulted in
the Case Collection A that contains 101 cases. The 138 webpages were re-organized
according to the identified cases.
Step 6. Apply quality assurance criteria to Case Collection A
To ensure that we have sufficient and adequate data on the cases for further
analysis, we evaluated the quality of data we had on the 101 cases in Case Collection
A based on the following quality assurance criteria:
&Does the data about a case startup allow the researchers to re-construct the pivoting
story of the startup in terms of what the startup was focused on before and after a
pivot, and why it made the pivot?
&Do the researchers have to make excessive guessing in order to understand the
pivoting type and the factors triggering those pivots?
A case is included if the answer to the first criterion is positive and the answer to
the second one is negative. The first two authors conducted this step separately. The
evaluation results from the two researchers were compared and the disagreed items
(36 out of 101 cases) were discussed until a consensus was achieved. This step
resulted in Case Collection B which contains 49 cases that were used in the data
analysis.
The data regarding these cases are contained in 47 webpages. The data on
one case may be spread in more than one webpage, and one webpage may
contain data on more than one case. The 47 webpages (represented by their
Empir Software Eng

URLs) used for analysis were documented and available at a permanent address
at https://figshare.com/s/152f52bb036cc6a67526.
3.2 Data Analysis Steps
Step 7. Extract the relevant data from Case Collection B
For each case (software startup) contained in Case Collection B, we were looking
for the following information on the case:
&Background information
–Name of the startup
–Location of the company
–Founding year and/or first product release date
–Business domain
&The main business/product/service before a pivot
&The main business/product/service after a pivot
&Description and explanation on how and why the startup pivoted
To get the background information, we first used the URL obtained through
our systematic search; if no information was found, we checked a startup’s
homepage or LinkedIn page; if still no information was found, we resorted to
Wikipedia. Wikipedia is used in six cases (Docker, Fab, Seesmic, Shopify,
Site59, Voylla). If there was more than one link that discussed the same
software startup (e.g. two links discussing Twitter as an example of pivoting),
we included all links along with the descriptions under the same startup name.
This step was conducted by the first author alone as it was mainly concerned
with data retrieval. The first and second authors discussed 3 unclear cases, to
resolve uncertain aspects regarding the background information of these cases.
Step 8. Coding the data to identify pivot types and triggering factors
The data extracted on each case was analysed qualitatively to identify the
pivot types and the factors that triggered the reported pivots. We relied on the
explanations given in the case material to identify the triggering factors of
pivots. The way we selected the cases ensured that the triggering factors that
led to pivots were described.
Acompletelyopencodingprocesswasusedtoallowtheemergenceofthe
triggering factors, meanwhile a seed category of pivot types as described in
Section 2.3 was used in the coding process to facilitate the identification of
the types of pivots these software startups have experienced. Table 1presents
an example of how coding was conducted:
This analysis step was conducted by the first and second authors separately.
The coding results from the two researchers were compared and the disagreed
items (the pivot types of 12 pivot instances and the pivot triggering factors of
5 pivot instances) were discussed between the two researchers until a consen-
sus was achieved.
This step resulted in 10 pivot types and 14 triggering factors.
Empir Software Eng

Step 9. Group pivot types and triggering factors
To provide a better structure of the pivot types, we classified them drawing upon
the four dimensions that are claimed to be vital for successful ventures by MacMillan
et al. (1987) and employed in other related studies (e.g. Giardino et al. 2014):
&Product dimension: startups are developing technologically innovative solutions
(Sutton 2003).
&Market dimension: it refers to identifying the essential need of the customers (Blank
2005).
&Financial dimension: it is related to the funding, investments, and return on invest-
ments and also the way a startup evolves sets the company growth and its place in
market (Yu et al. 2012).
&Team dimension: it is the main driving force behind several entrepreneurial activities
related to product and business development (Giardino et al. 2014).
The triggering factors were grouped into external and internal factors. External
factors are those that are beyond the control of a startup, whereas internal factors stem
from the decisions or activities of a startup itself.
4Results
4.1 Description of the 49 Pivoted Software Startups
The 49 software startups included in our case survey come from all over the world, however
the majority (37) are based in the United States, and 4 in Canada. Two case companies are
located in Israel, while the other 4 are located in United Kingdom, Australia, New Zealand and
India. For two companies we could not obtain the information on their geographic locations.
Social networks (30.61 %), e-commerce (24.44 %), and finance and business (12.24 %) are
the main business domains these software startups come from. The other domains include
digital government, operating system, health and travel industries. Most products developed by
these startups are market-driven and Internet-based. The targeted customers are either general
(such as Twitter, Yelp and YouTube) or from a specific segment (e.g., Ignighter targets at
Indian users primarily).
Twenty four out of the 49 cases are recent software startups, either being founded or releasing
their first products in the past five years (between 2010 and 2015), while 13 cases have launched
their products during 2005–2009. 7 startups released their products first time during 1998 to
2004, while for 5 cases the product release dates we could not obtain information.
Table 2lists the 49 software startups included in our study, including their company names
(at the time their pivots were reported in the webpage), the main business ideas before and after
Tab le 1 A coding example to identify pivot type and triggering factor
Software
Startup
Before Pivot After Pivot Pivot
Type
Reason of Pivot Triggering
Factor
BranchOut “Social
networking site for
professionals”
“Talk.co. A
chatting app
for employees”
product
zoom-in
“user acquisition slowed, people
were only using the site
occasionally”(retention low)
Negative
customer
reaction
Shaded columns contain codes
Empir Software Eng

their pivots. From these 49 cases we identified 55 instances of pivots, which means that some
cases contain more than one pivot example. These cases are marked in Table 2.
4.2 Overview of Pivot Triggering Factors
Table 3describes the major factors identified from the 49 cases that act as triggers for software
startups to pivot. These triggering factors are grouped under either External or Internal,as
defined in Section 3.2, Step 9. It is likely that a pivot instance is triggered by more than one
factor.
4.3 Overview of Pivot Types
The major pivoting types identified in the 49 cases are listed in Table 4, organized under the
dimensions of Bproduct^,Bmarket^and Bothers^. (Note that our findings did not reveal any
pivot that can be classified as financial or team related pivots.) One pivot instance is classified
under one pivot type only.
As shown in Table 4, market related pivots (45.40 % of the total 55 instances of pivots in 49
cases) are the most common types of pivots among the software startups included in this study.
Another 31 % of the pivot cases are product related pivots. The results also reveal several new
types of pivots (23.60 %) in addition to the existing pivot types presented in Section 2.3,
including market zoom-in pivot, complete pivot and side project pivot.
In the following sub-sections, the 49 cases are classified according to their pivot types and
corresponding triggering factors. Tables 5,6and 7list the cases that contain one instance of
pivot per case. Table 8shows the cases that have multiple pivot instances. For each pivot type
and triggering factor, we highlight the more illuminative and interesting cases in the text by
providing more detailed description and insights on these cases. The direct quotations from the
cases are also included while writing the description of each case, which are referred to in the
text as [casename] or [casename:link number] if a case has multiple URLs. The case names
and link numbering can be found at https://figshare.com/s/152f52bb036cc6a67526.
4.4 Product Related Pivots
In this and the following sub-sections, we use the pivot instances identified from the 49 cases
to further illustrate each pivot type and the corresponding triggering factors. Table 5shows a
list of software startups that each contains one pivot instance only and the instance can be
classified under the product related pivots. The triggering factors that caused these startups to
pivot are also shown in the table.
4.4.1 Zoom-In Pivot
Three startups - Flickr, Slack and Voylla - did product zoom-in pivot because their users
appreciated one particular feature rather than the whole product they offered. Flickr is a
representative example. It originally was an online massive multiplayer role-playing game
called Game Neverending. It failed to attract the customers’attention. However, the game
provided a photo sharing tool to allow players to share photos and save them on a webpage
while playing. This turned out to be the most popular aspect of the game. The founders
decided to leverage this popularity and pivoted towards a photo sharing application now
Empir Software Eng

Tab le 2 The list of 49 pivoted software startups
Software startups Before pivot(s) After pivot(s)
Android Operating system for cameras Operating system for smartphone (mobile
handsets)
appMobi Flycast: a mobile app selling creative
audio/video banner ads for iPhone,
Blackberry and Andriod devices
A set of tools for cross-platform develop-
ment of mobile apps
BranchOut Social networking site for professionals Talk.co: a chatting app for employees
BraveNewTalent Social recruitment platform Social learning palform
Carbon On demand valet parking Valet parking offering several services (e.g.
picked up laundry)
called delivery concierge service for car
ChartBeat Firef.ly: enabling website owners to see
how users are mousing around their
websites
Infographics about users visiting the
websites
Citivox Government enterprise software suite Community-organizing tool where people
can share what they like
Curios.me CrowdRally: Facebook fan site network Website for questions and answers
targetting at college students
Docker dotCloud: renting software and hardware
(platform as a service)
Open source development system
Eden Providing help to solve technical issues of
consumers
Solving technical issues especially
focussing on companies
Elto Tweaky: marketplace for developers to
make small changes in websites
Adding more functionalities and helping
small business to grow by connecting
them to marketers and growth strategists
Fab A social network targeted at gay
community
A daily flash sales website for modern and
latest fashion clothes, housewares,
accessories, clothing, and jewellery
Flickr Neverending: a massive multiplayer online
role-playing game
Sharing photos online
Groupize A travel startup focusing on consumers Group-booking solution for businesses
(travel management companies etc.)
Groupon thepoint.com: social goods fund raising site
based on tipping point
Group buying site working on same tipping
point
Handmake Me
a
Host My Portfolio: a portfolio service for
professional creatives
Reverse marketplace for handmade gifts
Hopper A travel discovery app to find where to
travel.
An app to suggest when is the best time to
fly.
Ignighter A group dating site for all A group dating site for Indian users
Instagram
a
Burbn: location based service Photo sharing app having different
filtration criteria
Jammber Social app with the goal to become
LinkedIn for the music industry
SaaS accounting platform for artists to
manage documentation (filling tax forms
etc.).
Keas Personalized care plans for customers Workplace wellness programs for different
businesses
Life On Air Air: video broadcasting Meerkat: a mobile app to stream live video
to their Twitter followers
MishGuru
a
Creating and printing 3D horse shoes A content management system
NextBigSound A website where music lovers could create
fantasy record labels to Bsign^artists
An enterprise data and analytics company
for artists, producers etc.
Empir Software Eng

Tab le 2 (continued)
Software startups Before pivot(s) After pivot(s)
Nextdoor Fanbase: wikipedia for sports A social network based on neighbourhood
Now in Store An app that allows brand to automatically
showcase their products in a portfolio
An app to help businesses to create
marketing content
Paypal
a
Confinity: helping Palm Pilot users
exchange money electronically
Online monetary exchanges
Pinterest Window shopping using mobile Collection of favourite items, and sharing it
with friends
RenentionScience
a
A platform for independent artists to
promote niche brand on social media
A platform that enables online businesses
to re-engage existing customers
Seesmic Video-based Twitter, enabling users to
broadcas t video clips
Social med ia client application to manage
multiple accounts on different social
platforms
Shopify Online snowboard business Online shopping cart for small businesses
Signpost Deal site like groupon Helping small and medium entreprises to
market online
Site59 Creating mini vacation packages from air
travel, hotel accommodation, and other
travel services, and selling them online.
Offer vacation packages by following the
B2B2C (business to business to
consumers) strategy
Slack Online role playing game Internal chat tool
Socrata A cloud-based database for
small-to-medium sized businesses
Cloud-based offerings for open data
governments
Streamline Streamlining customer service industry by
cutting waiting time during call-ins
Tools to help retailers in their decision to
expand
StyleZen Shopping site Technologies to help businesses leverage
Pinterest as a marketing vehicle
SymphonyCommerce A social shopping website that sells style,
home, beauty and living goods
E-commerce platform
Tagged A social network A social discovery product
Turntable.fm Stickybits: a mobile bar code scanning
startup
A social media website allowing users to
interactively share music
Twitter Odeo: personal podcasting service A microblogging platform
Vidyard A startup making marketing videos Providing analytics about videos
Voylla Online retailers for women’s apparel,
jewelry, and accessories
Online retailers for women’s jewelry and
accessories, excluding apparel from
intial offerings
Wix Flash website builder HTML5 based platform for website
development
Woot A electronics wholesale distributor A unique model for online shopping
(Internet retailers)
Wor ki ble
a
HireMeUp: a job site helping job seekers to
find jobs based on their availbility
Mobile based solution to find quality
people fast
Yelp Automated system for email
recommendations to friends
One-stop shop for local business reviews
YouTube Video dating platform Sharing videos online
Zealyst Helping people build meaningful new
connections and create stronger social
networks
Organizing events where people can make
new professional and personal
connections
a
case containing more than one pivot instance
Empir Software Eng

known as Flickr. In the case of Voylla which provides an e-commerce solution, the product
zoom-in pivot happens at the content level (offering less online products) rather than at the
underlying platform level. This is different from the software feature zoom-in pivot in pure
software startups like Flickr and Slack.
BranchOut and Shopify did zoom-in pivot for different reasons. Starting as a social
networking site for professionals, a so-called BLinkedIn meets Facebook^venture, BranchOut
had a good start but user acquisition was slow and users were only using the site occasionally.
New direction needed to be found, and it came to the attention of BranchOut that the
messaging service was used by a lot of users so it decided to break this feature out (called
Talk.co). Instead, Shopify narrowed their online snowboard business down to offering specific
Tab le 3 Major factors triggering pivots in software startup
Triggering factors Description # of pivot
instances
External
Negative customer reaction It refers to slow customer acquisition, slow customer
retention, no or negative response from customers etc.
15
Unable to compete with competitor Several competitors (e.g. big companies, other startup
companies) outplay the startup by working on the
same idea more effectively.
5
Technology challenge Several challenges related to technology, including
limitation with exisiting technologies (e.g.
performance issues), better technology availablity due
to emergence of disruptive technologies.
5
Influence of investor/mentor/partner Suggestion or pressure from investors, mentors or
partners to change the direction.
4
User appreciation of one particular
feature of the product
Uers appreciate one specific feature, rathen than showing
interest in the whole product.
4
Unanticipated use of product by users Users use product in an unexpected manner, which was
not foreseen before.
4
Wrong timing Providing a solution which market is not yet ready to
accept.
3
Positive response from an unforeseen
customer segment
Among different customer segments, one specific
segment shows more interest in the product.
3
Running into legal issue Legal problems with other companies (e.g., copyright
issues).
1
Side project more successful than main
project
Lack of interest from customers in the main project, but
they are interested in the side project.
1
Targetted market narrowing The initially targetted maket becomes smaller for the
startuptosurviveandgrow.
1
Internal
Flawed business model High cost of customer acquistion, or revenue model is
not working.
7
Identification of a bigger customer
need through solving an internal
problem
While developing a solution internally, to supprot the
core product, the startup realizes that the identified
internal problem is the real pain point for the
customers, compared to the problem their original
product solves.
5
Unscalable business Solving a problem in which not many people are
interested, resulting in unscalable business.
5
Empir Software Eng

online shopping cart solution, because in 2004 when they needed an online shopping cart for
their online business, they found no suitable choices, had to create their own, and realized that
it was the same issue many other small companies ran into.
In the case of Pinterest, multiple triggering factors were identified. It used to be a mobile
shopping app called Tote, allowed people to browse and shop from their favourite retailers, and
also sent them updates when their favourite items were available and/or on sale. The idea of
mobile shopping was ahead of its time in 2009, due to the fact that mobile payment solution
Tab le 4 Major pivot types in the software startups
Dimension Pivot type # of pivot
instances
Product Zoom-in: a single feature of a product becomes the whole product. 7
Technology: the same solution using comletely different technology. 5
Platform: a product becomes a platform or vice versa. 3
Zoom-out: a whole product becomes one feature of a much larger product. 2
Market Customer need: switch to a different problem that customers have 17
Customer segment: switch to a different customer segment than the one originally
conceived.
6
Channel: finding a more effective way to reach the customers than the earlier one. 1
a
Zoom-in: Focussing on one specific market sector rather than the whole market. 1
Others
a
Complete: Significant change in product, market and financial dimensions but the
entrepreneurial team remains the same.
11
a
Side Project: A different business idea parallel and unrelated to the main project
becomes the main project.
2
Tot al 55
a
new pivot type identified in this study
Tab le 5 Product related pivots and triggering factors
Pivot type Startup name Triggering factor
Zoom-in Flickr User appreciation of one particular feature of the product
Slack
Voylla
BranchOut Negative customer reaction
Shopify Identification of a bigger customer need through solving
an internal problem
Pinterest Unanticipated use of product by users
Wrong timing
Technology Wix Technology challenge
Android Targetted market narrowing
Platform appMobi Identification of a bigger customer need through solving
an internal problem
StyleZen
SymphonyCommerce Technology challenge
Zoom-out Elto Unscalable business
Carbon Flawed business model
Empir Software Eng

was not sophisticated at the time. Meanwhile, the founders also discovered that people were
more interested in sharing their favourite items lists with their friends and relatives rather than
Tab le 6 Market related pivots and triggering factors
Pivot type Startup name Triggering factor
Customer need Jammber Negative customer reaction
Docker
Citivox
ChartBeat Negative customer reaction
Unanticipated use of product by users
YouTube Negative custome reaction
Wrong timing
Signpost Unable to compete with competitor
Tagged
Yelp Unanticipated use of product by users
Hopper
BraveNewTalent Unscalable business
Vidyard
NowInStor e Influence of investor/mentor/partner
NextBigSound Flawed business model
Customer segment Eden Flawed business model
Socrata
Zealyst
Positive response from an unforeseen customer segment
Groupize Unable to compete with competitor
Keas Negative customer reaction
Channel Site59 Negative customer reaction
Influence of investor/mentor/partner
Zoom-in Ignighter Positive response from an unforeseen customer segment
Tab le 7 Other pivot types and triggering factors
Pivot type Startup name Triggering factor
Complete Twitter Unable to compete with competitor
Streamline Influence of investor/mentor/partner
Seesmic Negative customer reaction
Turntable.fm
Nextdoor
Curios.me Running into legal issue
Fab Negative customer reaction
Flawed business model
Woot Identification of a bigger customer need through
solving an internal problem
Side project Groupon negative customer reactoin
Flawed business model
Life On Air Side project more successful than main project
Empir Software Eng

doing window shopping at their favourite stores. The startup company apprehended an
opportunity through this unanticipated user behaviour and pivoted towards what we know
today as Pinterest.
4.4.2 Technology Pivot
Technology challenge is a triggering factor to technology pivot. In the case of Wix, it started as
a Flash-based website builder when Flash was the best option available for website develop-
ment before 2011. With the advent of smartphones, mobile devices and introduction of
HTML5, Flash was not anymore a viable option for their business because of its performance
problem with the smartphones. Due to this reason, Wix pivoted towards providing the website
development platform using HTML5.
Technology challenge is also behind the pivot of Android, presented as the emergence of
new technology. The original idea behind Android was to provide an operating system for
smart cameras that were linked to Personal Computers (PC), and to provide cloud storage to
store photos. However during that time, the smartphone and mobile devices industry witnessed
high growth, which led to the market of smart cameras shrinking. The camera market became
too small for Android business. The combination of the new emergent smartphone technology
and narrowing camera market triggered Android to pivot from an operating system for cameras
to provide mobile platform (operating system) focusing on handsets.
4.4.3 Platform Pivot
Platform pivot can be bi-directional by definition, either from a particular product to an
underlying platform or vice versa. However in our case sample, the three startups, appMobi,
StyleZen and SymphonyCommerce all pivoted from product to platform. For example,
appMobi, originally called Flycast, pivoted from a mobile app for iPhone, Blackberry and
Android to a set of tools that support cross-platform development of mobile apps.
Tab le 8 Software startups with multiple pivot instances, and triggering factors
Startup name Pivot instance Triggering factor
Instagram Product zoom-in User appreciation of one particular feature of the product
Technology Technology challenge
Handmake Me Customer need Negative customer reactoin
Flawed business modelCustomer segment
MishGuru Complete Unscalable business
Customer need Identification of a bigger customer need through solving
an internal problem
Paypal Technology Wrong timing
Customer need Influence of investor/mentor
Workible Technology Technology challenge
Customer need
RenentionScience Complete Negative customer reaction
Unscalable business
Complete Unable to compete with competitors
Empir Software Eng

The drive to the pivots of appMobi and StyleZen came from solving their internal
pain point first. Launched in 2011, StyleZen was an online shopping site. Since the
company was struggling with growing its user base without spending a huge amount
of money, they decided to use Pinterest to market about their business in order to
increase traffic and followers. The team experimented with different types of pinning
contents, and identified that there was a solution to increase traffic for their business
using Pinterest. The team implemented the solution as a set of tools, and realized that
their solution has more power to grow rather than the online shopping site itself, as
the co-founder described:
BOnce we figured out the technology, we had a very strong belief that it would be more
valuable if leveraged across multiple brands as opposed to my one startup brand. An
internal tool turned out to be more valuable than the shell.^[StyleZen]
Therefore, the main product of StyleZen pivoted to a Pinterest-optimization plat-
form called Ahalogy for other companies to use Pinterest as a marketing vehicle for
their business.
In the case of SymphonyCommerce, technology challenges are the factor behind its pivot.
The company initially was a social shopping startup, selling different goods (home, style,
living etc.) and requiring users to login through their Facebook account to access the website.
Its growth relied heavily on Facebook’s Open Graph. However, when Facebook changed
Open Graph implementation to make the connected apps less spammy, the viral effect and the
large audiences that some apps were enjoying, including SymphonyCommerce, were seriously
compromised. Due to this technological change from Facebook, the startup pivoted towards
providing a Symphony platform for e-commerce.
4.4.4 Zoom-Out Pivot
Two startups, Elto and Carbon, have pivoted in a product zoom-out manner. Tweaky
(now Elto) was launched as a marketplace for developers, providing a quick medium
for small businesses to make changes in their website with a certain fee. The founder
discovered that providing only small changes to the websites was not a scalable idea.
This triggered the company to decide to go broader by adding more functionalities
and providing more services to their marketplace, and to rebrand the startup’snameto
Elto (Bevery little thing online^). The co-founder commented:
BTens of thousands of customers later, we realized small tweaks to websites wasn’tthe
way we could add the most value to small businesses, who really want help to grow. Part
of that will always be adding functionality, but we’re now also focusing on connecting
them to marketers and growth strategists.^[Elto]
Carbon’s zoom-out pivot was due to the flaw in their original business model. It initially
started as an on-demand parking that rented space in garages and dispatched valet to pick up
cars. However, in order to run the business, they needed intensive investments from the
ventures, as the founder described:
BFor on-demand parking, every time you’re parking a car, you’re spending two, three,
even five times more. It’s very venture capital subsidized. Carbon didn’twanttoplay
that game anymore.^[Carbon]
Empir Software Eng

The founder realized that they should provide more services to earn profits. Hence, they
pivoted by adding several other services (e.g. laundry pickup) in addition to on-demand valet
parking, and became a delivery concierge service for the cars.
4.5 Market Related Pivots
Table 6shows a list of software startups that each contains one pivot instance only and the
instance can be classified under the market related pivots. The triggering factors that caused
these startups to pivot are also shown in the table.
4.5.1 Customer Need Pivot
The first type of market related pivot is customer need pivot. Negative customer reaction is the
triggering factor for several startups, including Jammber, Docker, Citivox, ChartBeat and
Youtube. Take the example of Jammber. The company initially started to provide a social
app where musicians could find different artists, stylists, photographs etc. Its goal was to
become LinkedIn for the music industry. However, while interacting closely with Nashville
artists, producers and the musician’s union, the co-founders discovered that the musicians
spent several hours on paper works to fill different documents etc. They realized this was a real
pain point for musicians. A decision needed to be made, as the founder commented:
BWe had to decide ourselves if we wanted to go the sexy route, or if we wanted to go the
money route…We decided to go the money route and do what’s best for Jammber. And
we’re now meeting investors who get that.^[Jammber]
As a result, Jammber pivoted towards a payment processing and document filing platform
between artists and their labels to communicate with each other, share documents and process
payments.
For ChartBeat and YouTube, apart from negative customer reaction, their pivots were also
driven by other factors. ChartBeat, initially called firef.ly, provided a way for website owners
to see what users were browsing around their websites, and let the same webpage users chat
with each other. However, instead of browsing the website, the users started chatting with each
other and discussing about firef.ly most of the time. This unanticipated use of the system,
together with low traction of the intended product, triggered the startup to pivot to ChartBeat
which provides different user information (how many people were on their site, where they
were coming from and what they were reading) to the website owners. In the case of YouTube,
it initially started as a video dating platform to find possible dates via videos. The idea was
ahead of its time in 2005. Moreover, the idea did not get much user attraction. Hence, it
pivoted towards a platform to share videos online.
The impact of competition can be crucial, as shown in the cases of Signpost and Tagged.
Both had to pivot the problem they initially wanted to solve for their customers. In the case of
Signpost, the startup originally was a website for daily deals, similar to Groupon. However
Groupon was a stronger player in this field, which forced the founder of Signpost to search for
new direction. Through talking with different local businesses, Signpost discovered that,
although there were many local sites available e.g. Yelp, Yahoo local, and Google local, there
was no tool available to update the business profiles on these sites. The founder realized this
was an unsatisfied need and pivoted towards helping small to medium sized businesses to
better market themselves online.
Empir Software Eng

Yelp and Hopper discovered real customer needs due to the unanticipated use of their
products by users. For example, Yelp intended their emailing system to be used by users to
connect with others for recommendation about local businesses. However, the users used the
system to write reviews about the local business. This emergent new mode of using the system
soon caught the attention of the founders, through which they identified a different yet more
promising need to be catered, and pivoted towards providing reviews about local businesses.
The realization that their businesses were unscalable pushed BraveNewTalent and Vidyard
to seek for new problems to solve. The initial idea behind BraveNewTalent was based on the
assumption that people wanted to follow the companies for whom they would like to work in
the future, and companies wanted to educate potential candidates on how they should work. It
intended to be a social recruiting platform. However, the startup did not figure out how to scale
their business idea as the founder described:
BWe realized we were trying to build communities around recruitment content. But
trying to build user engagement around transactional content like that doesn’t work. Job
seekers just wanted jobs, and recruiters just wanted to fill positions. So the model didn’t
scale just focusing on jobs.^[BraveNewTalent]
Consequently, BraveNewTalent pivoted from a social recruitment platform to social learn-
ing platform and primarily focusing on enterprises settings.
The story of NowInStore presents an example of customer need pivot triggered by yet
another factor. Initially, the startup developed an app that allowed different brands to showcase
their products in a portfolio. The startup was accepted in a New York based technology
accelerator. During their stay in New York, the founders had meetings with dozens of
investors. Due to the influence of investors, they realized what could be more effective, and
pivoted to a platform that helped businesses create marketing content by leveraging the data
from online stores. The founder commented:
BThe product has neatly positioned itself as a marketing platform for small and medium-
sized businesses (SMBs), leveraging smart data, and it’s going really well now.^
[NowInStore]
In the case of NextBigSound, it was flawed in their original business model, which
triggered the pivot. The original idea behind NextBigSound was to develop a website for
music passionate people, where they could create fantasy record labels to Bsign^artists.
Although they had thousands of users, they could not generate enough revenue from this user
base to survive. To find the real need of the musicians and artists, the founding team remained
the focus on the music industry, and tried to identify the mantra behind popularity of bands
from garage to a big hit. The new idea was to track social media to measure popularity of new
and up-coming music. Hence, they pivoted, and became a company which provided data
analytics to artists, producers and labels.
4.5.2 Customer Segment Pivot
Five startups, Eden, Groupize, Keas, Socrata and Zealyst, are classified under this category,
and each pivot is triggered by a different factor except Socrata and Zealyst, who have the same
triggering factor. Eden initially started with providing technical support to consumers to solve
their information technology related issues –similar to geek squad. The startup was mainly
focused on providing services to end consumers to generate revenue. However, they soon
Empir Software Eng

realized that their revenue model was not working as they expected, because the consumers
were more sensitive to price than they expected. Meanwhile the founder also discovered that
their main revenue was generated through the business customers (enterprises), so they pivoted
customer segment from consumers to business. The CEO commented on this pivot:
BIt’s tempting as a founder to have a consumer business idea, but once we noticed the
best part of our business was in B2B, we had to pivot. Serving businesses tends to have
higher frequency and higher profit.^[Eden]
Groupize originally targeted at consumer business by generating demands through creating
white-label agreements with multiple travel agencies. Unable to compete with their compet-
itors in the travel management service areas, they shifted their focus towards providing a
group-meeting solution for hotels, meeting planners and travel management companies, and
created partnerships with different hotel chains and independents. In contrast, the pivot of Keas
was due to negative customer reaction. Initially planning to provide personalized health care
places for consumers by leveraging their personal health data, Keas ended up providing
workplace wellness program for different businesses since their idea did not get much traction
in consumer market.
Socrata and Zealyst pivoted due to the same triggering factor. For instance, Socrata was
launched to create a cloud-based database for small to medium sized businesses. The original
idea was to put one’s database in a cloud and let someone else manage it. During the
presidential campaign in 2008, the president’s campaign team used this platform to put
contribution data online. This positive response from previously unforeseen customer segment
made the founder recognize that government can put data online by using a cost effective
solution like cloud computing. The startup pivoted accordingly towards a cloud based offering
for open data government.
4.5.3 Channel Pivot
Site59 presents an example of channel pivot. The initial idea of Site59 was to create mini-
vacation packages by combining last minute offers from different air travel, hotel accommo-
dation and other travel related services. However, the idea did not go viral as expected, and the
customer acquisition rate was low. One of the investors suggested Site59 to change the
distribution channel to reach customers, using the business to business to consumers
(B2B2C) model. Following the suggestion of their investor, Site59 pivoted the channel to
reach their customers and their new service was to prepare last minute vacation packages for
different airlines and travel portals which eventually reached the end customers they intended
to serve initially.
4.5.4 Zoom-In Pivot
Market zoom-in is a new pivot type emerged from our study. It is a type of pivot where a
startup narrows down its target market from a broader one to a more specific market segment.
Ignighter is an interesting case of such a pivot. The primary aim of Ignighter was to develop a
dating website for users. The targeted audience is general, however the founders expected to
receive positive response from the US, their home market. Unexpectedly, the idea got
promising attraction from customers in Asian markets, especially in India. The founders
carefully analysed the demographic data, and identified the promising user growth in India
Empir Software Eng

as compared to any other country. As a result the founders decided to focus on this market
segment and made the pivot, stating that Bwe are an official Indian dating site.^[Ignighter]
4.6 Other Types of Pivots
Table 7shows a list of software startups that each contains one pivot instance only and the
instance cannot be classified readily under either product or market category. The triggering
factors that caused these startups to pivot are also shown in the table.
4.6.1 Complete Pivot
A complete pivot is a pivot where an entrepreneurial team has to come up with a new
innovative idea after their initial innovative product/idea was outplayed due to different factors
(by their competitors) e.g. big companies started working on that idea and attracted their niche
markets. This pivot implies significant change in one or more aspects of a startup, including
product, targeted market and finance. The only unchanging element is the entrepreneurial team
that carries on the learning from the past experience to the new directions.
Twitter is an example of complete pivot. It initially started as a podcast service (Odeo) to
allow sharing and recording of podcasts. Then Apple iTunes started to fill this gap, leaving
behind the Odeo service. As they were unable to compete with Apple iTunes, the startup team
had to brainstorm to find a new direction, and came up with a new messaging service called
Twi tter.
The complete pivot of Streamline can be attributed to their mentors while they were in a
Techstars Seattle program. Initially the startup aspired to improve and streamline the customer
service industry by reducing the waiting time during call-ins to customer service centres. After
they were selected into the Techstars Seattle class, with the advice and guidance of Techstars
mentors, they completely changed their initial idea, and pivoted towards helping brick and
mortar owners in their decision making related to expansion. The founder described the new
idea as Bhelping retailers expand intelligently to the right physical locations^,becauseBone
mistake on a location could destroy a small retail concept. We’re solving this pain.^
[Streamline]
Three complete pivot cases, Seesmic, Turntable.fm and Nextdoor, were triggered by
negative customer reaction. Take Seesmic as an example. It started to be a video-based
Twitter. The founder realized that people would prefer to tweet in words rather than
contributing in a discussion by recording a video message, which is too much of a
hassle for engaging in a discussion. Based on the negative customer reaction (no
traction, difficulties in recording coherent video message), they changed their direction
completely, and pivoted towards developing a social media client application particu-
larlyfocusingonenterpriseservices.
Curios.me had to pivot completely because it ran into legal issues. Originally, it was a
Facebook fan-site network known as CrowdRally. The Facebook lawyer’s team identified
certain uses of CrowdRally as inappropriate. Curios.me pivoted towards providing a question-
answer site for college students due to the little chance of resolving the legal trouble.
Fab.com pivoted completely due to two factors. Starting as a social network targeting the
gay community, the startup did not get much traction, was stuck at one point, and unable to
reach the revenue point that they had projected. The other jobs the founders had in the design
field hinted them when they searched for new directions. Consequently Fab.com pivoted
Empir Software Eng

towards selling handpicked home goods, clothing and accessories. This new direction took off
and it has now become the well-known model fashion website.
Solving an internal problem sometimes can lead to a complete pivot. In the case of Woot, it
started as a wholesale electronic distributor and wanted to clear out their unsold inventory. While
solving own inventory problem, Woot discovered a unique model for online shopping website
combining both the need of urgency and scarcity, and pivoted towards software industry.
4.6.2 Side Project Pivot
Side project is a special kind of project that runs parallel to the main project of a software
startup, but may be based on a different even unrelated business idea and target at a different
set of customers. Groupon is a well-known example of side project pivot where side project
outshines the main one. However the deeper reasons were because of the issues with the main
project. Groupon initially started as ‘The Point’: social campaigns to collect fund for good
causes. Campaigns were only successful when a certain tipping point was reached. However
this project did not get much user traction, and there was no clear revenue model therefore it
was difficult to monetize the idea. However, the side project the team started in parallel, using
the same tipping point but for group buying and local deals, attracted more users. Eventually
the side project took off and it has now become the daily deal website famously known as
Groupon.
Life On Air is also an example of pivot where side project overshadows the other projects
and becomes the main project. The side project was called Meerkat, an app enabling users to
stream live video to their Twitter followers with a single click. It is different from the main
business of the company which was another app called Air. Meerkat’s usage shot up almost
immediately after launch and topped the list on Product Hunt, a site that allows technology
enthusiasts to surface and vote on new tech. Over 3 days roughly 15,000 people used the
Meerkat service. As a result the founder of Life On Air decided to focus the team on Meerkat
fully. He commented:
BMeerkat is the embodiment of the ability to run really fast, look up and see whether you
are going the right way, and if not redirect yourself.^[Life On Air]
4.7 Multiple Pivots
We identified six startups that evidenced multiple pivots. Table 8shows a list of software
startups that each contains two pivot instances together with triggering factors.
Instagram provides an illustrating example of a startup which evidenced two product related
pivots, zoom-in and technology, during their journey. Instagram originally was a location-
based service called Burbn, combining features of Foursqaure (photo share app) and
Mafiawars (game). Users could earn points for hanging out with their friends, and share
pictures inside of the app. However the users appreciated only one particular feature of the
product, photo sharing. The founders also realized the need of focusing on this specific aspect
(a product zoom-in pivot), as the co-founder explained this situation:
BI’ve heard that Plan A is never the product entrepreneurs actually end up with. I didn’t
believe it. In many ways, Burbn was getting a bunch of press, but it wasn’t taking off the
way we thought it would. We found people loved posting pictures, and that photos were the
Empir Software Eng

thing that stuck. Mike, my cofounder, and I sat down and thought about the one thing that
made the product unique and interesting, and photos kept coming up^.[Instagram:2]
The service was initially a browser based mobile app developed in HTML5. However due
to latency issues in HTML5, they pivoted towards providing an iOS only app for iPhones. The
main reason behind this technology pivot was the limitation of the existing technologies.
Handmake Me is a special case which evidenced two market related pivots: cus-
tomer segment and customer need. Their original idea was to provide a portfolio
service for professional creatives, called Host My Portfolio. However, the idea did
not bring any revenue. They realized that most of the products on their service were
craft-related. They pivoted towards hobbyists (customer segment) and created a
Breverse marketplace for handmade gifts^called Handmake Me. It allowed anybody
who wanted to buy an authentic handmade gift to request what they wanted and how
much they would pay, and craft-makers would then bid for the task. However, nobody
was requesting anything. In order to solve this, they changed their direction towards a
more conventional marketplace. They implemented various customized option for
buyers, and implemented a strict quality assurance criteria. BIt worked. People are
buying,people are selling and we are revenue generating at last.^[Handmake Me]
Another example of multiple pivots is MishGuru, which pivoted twice: complete
and customer need. The original concept was to develop a platform that allowed users
to custom design and print their own 3D horse shoes. During their work in a lab
accelerator program in New Zealand, they discovered that their idea was not scalable
because their targeted market (horse owners) was not conducive for rapid growth.
Hence, they pivoted in a complete new direction. The founder explained the new idea:
BWe started playing around with an idea for collaborative video making in between
friends. As part of that, we spent a couple of days building a really basic MVP using
Snapchat as the base platform to build it off since everyone already had the app.^
[MishGuru]
While working on the idea of collaboration video making, they stumbled upon
another challenge. They were unable to find a solution to manage content, and drive
user engagement on snapchat. They developed their own solution to manage these
kinds of activities. Now, they have become a content management system for Snapchat,
where users can create storyboards, build campaigns etc., and publish on Snapchat.
PayPal and Workible are examples, where both evidenced technology and customer
need pivots. Although they evidenced similar pivots, the triggering factors were differ-
ent. In the case of PayPal, timing issue was behind the pivots. PayPal pivoted from a
money exchange solution for Palm Pilot users to online payment, since mobile payment
was Bstill’the future’12 years later^[Paypal:2]. The new customer need that PayPal
pivoted to, online payment, was influenced by the merge with X.com and partnership
with eBay. In the case of Workible, it initially started as a job site, which helped job
seekers to find the job based on their availability. The solution was a desktop based
website initially. However, as the founders commented:
BWe had what we call now a blinding flash of the obvious, but it was a bit of an
epiphany that if we were going to build a world class tech business we could not do it
with a website we built out of India. And overwhelmingly we got this message that
everything was going mobile and everything had some social component.^[Workible]
Empir Software Eng

At the same time, the founders did a massive amount of research into the market and started
asking people what their biggest problems were. They learnt that the main issue was the
difficulty to find quality people fast. Mobile is the perfect solution for that because it allows
customers to connect with people instantly. After realizing the real need of the customers, the
founders pivoted towards providing a mobile platform that allowed companies to instantly find
the candidates that matched a set of criteria.
RetentionScience is an interesting case which evidenced two complete pivots before
founding RetentionScience. Initially it provided independent artists a platform where they
could promote niche brands and products via social media. Although the founders contacted
different channels (working with YouTube celebrities, sponsoring local concerts, etc.), their
business proved to be unscalable. They also discovered that the customers were reluctant to
appear as sell-out by promoting different brands. The unsalable business and negative cus-
tomer reaction triggered their first complete pivot. They pivoted towards providing a social
media-based analytics and referral platform for e-commerce businesses. The second complete
pivot happened due to the competitors. The founder discovered that there were many well-
funded startups working in the same area, and there were little chance that they could acquired
the funds to compete. Without funding, they could not accelerate their product development
and increase user growth, hence unable to compete with their competitors. Therefore, they
pivoted completely again towards a retention automation platform that used artificial intelli-
gence techniques, to engage existing customers and increase customer retention.
5 Discussion
When the conditions are uncertain and chaotic, failure is almost inevitable (McGrath 2011). It
is rightly applicable in the context of software startups that work under the conditions of
extreme uncertainty. A common element behind all the pivot cases we studied is validated
learning. Without it, the pivot decisions would be ungrounded (Ries 2011). Startups obtain
validated learning through failures, therefore making Bintelligent failures^(McGrath 2011). As
shown in the cases reported in Section 4, the studied software startups adopted this validated
learning process and used the acquired knowledge to set right directions. That is why in the
context of startups, failure is viewed positively, and failing fast and failing often is the mantra
of most lean startups. If startups do not learn from their failures, there is a high probability that
they would eventually fail permanently (Richardson 2011).
5.1 Reflection on Pivot Types
There are few studies related to pivots in software startups as described in Section 2. Our study
extends the scope investigated in Van der Van and Bosch (2013) and Hirvikoski (2014), and
adds new knowledge regarding pivots, pivot types and triggering factors. The study by Terho
et al. (2015) presents different types of pivots as described by Ries (2011), and how they affect
the business model. Our findings extend this study by providing further types of pivots (e.g.
complete pivot etc.), and also provide the categorization of different factors that trigger pivots
in software startups.
Among the pivots identified in our study, the most common is customer need pivot (17 out
of the 55 pivot instances). This is not surprising in the sense that it is consistent with the nature
of startups in general and software startups in particular. While working with highly dynamic
Empir Software Eng

and uncertain technologies and building innovative products, software startups are striving to
find the real and unique customer problems that are worth solving. In order to better
understand customer needs and identify real problems, software startups need to pivot
relentlessly.
The significance of customer need pivot reveals the importance of identifying the right
problem, first. In order to develop something valuable for their customers, startups need to
understand their problems (Blank 2005). However, the existing studies show that software
startups tend to ignore the identification of right customer problems, instead focus on
developing solutions and investing in product/market fit prematurely (Giardino et al. 2014,
2015). It is highly probable that the initial assumptions towards customers prove to be wrong.
This is one of the reasons of customer need pivot, where startups should pivot because the
problems that they identified are not real pains for customers. This is manifested in the case of
Yelp, Hopper and Jammber etc. where they pivoted subsequently according to different
customer needs they discovered.
In the course of better understanding the market, software startups often discover that, even
though the problem they want to solve is real but it is not the problem of the customer segment
they have initially presumed. 6 pivot instances made the customer segment pivot, the second
most common market related pivot type that our study has identified. Startups do not know
their potential customers in advance, and risk spending too many resources to come up with a
product that fails to achieve product/market fit. However, every failure has some lessons to be
learnt. This learning can be helpful to identify new targeting customers and then pivot towards
them, as manifested in the cases of Eden, Socrata, Groupize, etc.
Product related pivots are also important pivot types. 17 pivot instances in our sample are
related to product. Software startups have to reconsider their products and different features in
order to find the problem/solution fit and/or product/market fit. This often leads towards
product related pivots. Among different product related pivots, the zoom-in pivot is relatively
more common than other product related pivot types. It often happens that customers are more
interested in a particular feature rather than the whole product. Pinterest and Flickr are good
examples of product zoom-in pivot. Ideally, instead of wasting resources and building a
complex product with lots of features, it is better to focus on one feature that actually gained
the attractions of the customers and build it first. However it is not easy to understand which
can be the valuable feature to build first. Ries (2011) suggests to develop MVP in order to test
the hypotheses related to product, business model, and/or engine of growth. It may be
beneficial for entrepreneurs to use MVP to decide the candidate features to include in their
product offerings. By building MVPs, entrepreneurs have an initial set of features that are
appreciated by the initial users.
The opposite of product zoom-in pivot is product zoom-out pivot, which reflects the need
of achieving the problem/solution fit. It is possible that software startups have identified the
right set of problems, but their products are still incomplete. They need to expand their
solutions to add more features. Elto did zoom-out pivot by providing additional functionalities
to its original product, and secured a better place in the market.
Another important product related pivot type is technology pivot, second most common
within the product dimension, which reflects the role technology plays in software startups.
Software startups are prone to technology pivots due to the fact that they are building
technology intensive products. Often technology pivots are driven by the need of software
startups to be always at the cutting edge of technological advancement. This is manifested in
the cases of Wix, Workible, Android and Instagram. Sometimes existing technology has some
Empir Software Eng

performance issues, and technology pivots help startups to come up with an improved solution,
as shown in the case of Instagram.
In order to support their products and make solutions complete, software startups often
develop both products and supporting platforms. The platforms support their core products.
Sometimes it happens that the platforms solve larger problems than their original products do.
Therefore platform pivots are desired. StyleZen is one such example. The startup pivoted
towards a platform to support online shopping businesses instead of becoming yet another
shopping website. It is worth mentioning that platform pivot can be also from a platform to a
product running on the platform. However, we could not find evidence in our case collection.
One possible explanation can be that this direction is not as frequent as the product to platform
direction.
In terms of the scope of change and the amount of effort and resource needed, no pivot type
is more demanding than complete pivot. This is a new pivot type identified in our study. It is
the second most common among all pivot types (11 out of 55 pivot instances). We term it
complete pivot since it is related to almost all the aspects of a startup, including product,
market and financial, with only the original team as the rooting element in the pivot, which
ensures the learning from previous failing experience is maintained. Famous companies such
as Twitter, Fab.com, and Turntable.fm all went through significant changes in their business
before they found successful and sustainable business model to scale.
Side project pivot is another interesting new pivot type. Even though working under a high-
pressurized and extreme chaotic environment, many software startups run one or more side
projects simultaneously that are generally not related to their main ideas. A side project is a
project that runs parallel to the main project, but may target at a different set of customers.
These side projects may become main projects when outshining them. Groupon is a good
example that was initially started as a side project. Therefore it is arguably beneficial to have a
side project parallel to the main product development project. Further studies need to be
conducted in order to explore the importance and implication of side project, and the cost
associated to running such parallel project, especially in the software startup context.
The third new type is market zoom-in pivot, which is demonstrated by the Ignighter case. It
is a reflection of striking the product/market fit. It is often suggested that, to start with, a startup
should find its focus and niche market, identify the early adopters of their product. This type of
pivot shows the need to do so. However, since we only found one instance in our case
collection, the robustness of this new pivot type is yet to be tested.
Another finding worth mentioning is multiple pivots which can happen either simulta-
neously or separately. Some pivots may be closely linked and there is a possibility that chain
reaction occurs, which means one pivot triggers several other pivots, known as Bthe domino
effect^(Terho et al. 2015). This chain reaction is manifested in the case of Instagram (product
zoom-in and technology pivot) and Workible (customer need and technology pivot), where
two pivots occurred simultaneously. One startup may have several pivots spread across their
courses of development too, which is revealed in the case of RetentionSciene (two separate
complete pivots). This indicates the importance of constantly checking and correcting the
directions until a startup obtains a sustainable business model.
A pivot can also be related to business architecture, value capture, and engine of growth or
social aspects. However our study does not find any case that evidences these kinds of pivots.
One reason could be that these pivot types may be less common in comparison to other types,
therefore not appearing in our case collection. Why these are less common types, however, is
beyond the speculation. Broaden the scope of cases may surface these pivot types in software
Empir Software Eng

startups, verify if they are indeed less common, and unveil why they appear less frequently as
pivot types.
A study by Giardino et al. (2015) reveals that building an entrepreneurial team is one of the
prominent challenges for software startups. Our results show that, however, there is a
knowledge gap regarding the pivots related to the team dimension. The same is true for the
financial dimension. Even though several pivot types are suggested by Ries (2011), our results
did not yield any related cases. Both team and financial are important dimensions of successful
software ventures (Macmillan et al. 1987), and further investigation is needed in order to gain
any insights in these aspects.
5.2 Reflection on Triggering Factors
The majority of the triggering factors listed in Table 3are considered external factors, which
are events occurring beyond the control of a startup. This implies that for many of the studied
startups, the major pivots they made were more reaction to what happened externally rather
than purposefully design change as suggested by Ries’definition of pivot (Ries 2011).
Table 3shows that negative customer reaction is the most common factor triggering pivots.
Slow user acquisition, low user retention rate and no growth are some manifestations of
negative customer reaction, and show that startups are unable to achieve the product/market fit.
Negative customer reaction works as a first litmus test for the startups to decide whether they
are solving the right problem for the right set of customers or not, and consequently whether
they should pivot or not. In the case of Seesmic and Jammber, both startups reacted to negative
customer reaction and pivoted subsequently.
In order to come up with innovative and cutting-edge products, software startups have to
compete with other competitors and especially with big companies. It is the second most
common external factor that triggers software startups to pivot. The big companies have much
more resources than what software startups can wield. They can implement innovative
products rather quickly as compared to startups. Twitter stumbled upon this challenge when
their initial idea of offering podcast services was outplayed by Apple with the launch of
iTunes. They pivoted drastically.
Thriving in technology uncertainty is the top challenge faced by early stage software
startups (Giardino et al. 2015). Accordingly, technology challenge is a common factor that
triggers pivot, and the pivot types are generally related to technology or product, such as
shown by the cases of SymphonyCommerce and Wix.
One of the contextual features of software startups is that they are heavily influenced by
stakeholders and investors (Paternoster et al. 2014). The suggestions from the investors/
mentors/partners greatly affect the development processes of software startups, and may
eventually change their course. It is probable that a software startup has a good technology
idea, but their investors, mentors or partners have a different vision, which affects the overall
direction of the startup. Investors/mentors may have valuable suggestions that are worth
listening. NowInStore and Streamline are good examples of opening up to the suggestions
from investors/mentors, and changing their directions accordingly.
Meanwhile, our findings also reveal that a flawed business model is a prominent internal
factor that triggers software startups to pivot. Scaling a business that has flaws in their business
model, or in other words, Bpre-matured scaling^(Giardino et al. 2014), may lead to the
eventual failure of a startup. Software startups may avoid failure by identifying the flaws in
their business model earlier and pivot accordingly. Low or no revenue, or high acquisition cost
Empir Software Eng

indicates a flawed business model and consequently the need to pivot, as demonstrated in the
cases of Groupon and Fab.com.
As shown in Tables 5,6,7and 8, there is no linear and one-to-one relationship between
pivot types and factors triggering pivots. Different factors trigger different types of pivots in
software startups. For example, unanticipated use of product by users is a factor that triggers
Pinterest to product zoom-in pivot, while in the case of Yelp, the same factor causes it to pivot
to different customer need. Similarly, it is probable that multiple factors collectively trigger a
pivot in a software startup. Instagram, Fab.com, and Groupon are prominent examples of
startups where pivot occurred due to several factors. It needs to be emphasized that the list of
triggering factors in Table 3is not exhaustive. There may be other triggering factors yet to be
discovered.
Last but not least, our findings on the major types of pivot and triggering factors causing
pivots allow us to reflect upon the role of the unique nature of software product plays in
software startup pivoting. For example, due to the flexible and modifiable nature of a software
product, product zoom-in and zoom-out pivots should be relatively easy to implement for
software startups than for startups that produce physical products, such as hardware or medical
devices. Similarly, since it is common that the use of software is always appropriated by users,
which is generally beyond its designed use (Mendoza et al. 2010), software startups should be
more watchful to the triggering factors such as unanticipated use of product by users.
5.3 Validity Threats and Mitigation
5.3.1 Threats to Construct Validity
This aspect of validity threat refers to what extent the operational measure really represent
what is investigated related to the research questions (Runeson and Höst 2009).
One threat to construct validity stems from the use of Bpivot^only in the search string to
find pivot cases. There may be cases that represent pivot but do not use this specific term. They
may rather use general terms such as Bchanging direction^,Bchanging strategy^or Bstrategic
change^. As a result, we may have missed some important pivot cases. However, pivot has
become a popular and common term used in startup communities, together with the Lean
Startup movement. In addition, changing business direction and strategy has a broad meaning
and does not reflect the specific implication of pivot which emphasizes on the change that
allows Bto test a new fundamental hypothesis about the product, strategy, and engine of
growth^(Ries 2011). Therefore, the probability of using these broad terms to get relevant
cases is low.
One limitation of the study is the exclusion of known pivot cases because of lack of
information about their pivoting story e.g. Facebook and Nokia are two world famous
companies, but we did not get sufficient information to find the triggering factors for their
pivots.
5.3.2 Threats to Internal Validity
Internal validity threat refers to the broader problems of making inference (Yin 2003).
One internal validity threat to our findings is also related to the secondary data we
employed since we have no control over the quality and accuracy of the data, which leads
to a weak basis for data analysis. In some cases the pivot instances were described in the
Empir Software Eng

interviews of the founders, in other cases were recounts from the writers of news articles who
did not have direct working experience in the startups being reported. This validity threat was
partially mitigated by data triangulation, using multiple sources of the same cases wherever
possible. The benefits allowed by secondary data, e.g., obtaining a large number of real world
pivot cases (49 software startups representing 55 pivot instances) with relatively less effort in a
short timeframe, out-weighted this potential validity threat.
Another threat is the quality of the selected cases. To reduce the personal bias in case
selection, we defined a review protocol with quality assessment criteria (as described in
Section 3). This review protocol was discussed and agreed among the research team to cover
maximum studies. Researcher triangulation was applied to increase the internal validity. Two
researchers conducted the key review tasks independently. As for the selected cases, not all of
them were equally informative. There were few pivot cases which provided more details (e.g.,
RetentionScience) as compared to others (e.g., YouTube). However we ensured that all the key
information about each pivot case was available.
To ensure the replicability of our study, we provided as many details as possible and
necessary in Section 3. The data collection and analysis steps reported are replicable. However,
due to the dynamic nature of the Web and Google search engine, the search results using the
same search string reported in Section 3may not be replicated completely with the lapse of
time. To mitigate this potential threat, we published the cases we obtained from the search
results in the online repository (the link is given in Section 3Step 6).
5.3.3 Threats to External Validity
This aspect of validity refers to what extent it is possible to generalize the findings (Runeson
and Höst 2009).
The first potential threat to external validity is posed by Google search engine on the
accessibility of search results. In our case we had access to 783 links, which were the top-
ranked query results according to its search algorithm. We could not know if these 783 links
were representative of the total search results (1,070,000).
A pivot action taken by a startup is not guaranteed to be successful and lead to eventual
successful startups. However, our sample of pivot cases may be biased towards successful
pivots and startups due to the fact that these pivots were published online, many times as the
examples of successful pivot. This bias may pose a threat to the generalizability of our findings
to the general population of software startups.
Another potential threat to generalization is the representativeness of the resulting
collection of cases. The 49 software startups come predominantly from the United
States (37 out of total 49 software startups). Since the US is arguably the world leader
in innovation and technology as represented by Silicon Valley and other startup
ecosystems, it can be argued that our sample is representative to certain extent. As
far as pivot is concerned, we believe that the demographic distribution should not
impact on the ways startup directions are changed.
The limited number of cases on different types of pivot did not allow us to conduct any
quantitative analysis, such as meta-analysis, to identify the relationship between consequence
and antecedence of pivots. Besides, much case context information was reported in a large
deviance in term of business stage, company size, product type, industry domain and etc. This
presents a potential threat to the generalizability of the pivot types and triggering factors to
different contexts. However the main objective of this study is to explore software startup pivot
Empir Software Eng

phenomenon qualitatively. We targeted at theoretical generalizability rather than statistical
generalizability.
One limitation of the study is regarding the generalizability of the identified pivot types and
triggering factors. The 49 software startups in the sample do not cover all the business domains
(e.g. missing real time application, safety critical application etc.). Moreover, not all the pivot
types and triggering factors are manifested in the equal amount of cases, some only have one
or two corresponding cases, such as market zoom-in and side project as pivot types and
running into legal issues as a triggering factor. Future work is needed to further explore these
pivot types and triggering factors to increase their generalizability.
6 Conclusions and Future Work
Software startups are producing innovative software products and solutions in dynamic
markets using cutting-edge technologies. To achieve success, they need to continually make
the decisions of pivoting. This paper provides an initial classification of pivot types that
software startups have made through a case survey study of 49 software startups that pivoted,
and the factors that triggered software startups’pivots.
The study results show that customer need pivot is the most common type of pivot among
all types, while zoom-in and technology pivots are major product related pivot types. Our
findings extend the existing knowledge of pivot by introducing three new pivot types: market
zoom-in, complete and side project. A knowledge gap regarding team and finance related
pivots is also underlined. Among the multiple external and internal pivot triggering factors
identified, negative customer reaction represents the most common reason why software
startups pivot, followed by the flawed business model of software startups.
Our findings have implications to both software startup practice and research. The
empirical evidence from the analysed cases suggests that software startup teams should
gather maximum knowledge and consider failure as an opportunity to obtain validated
learning. Considering the chaotic and unpredictable environment of software startups,
the validated learning will be crucial to drive business and product decisions in order to
proceed in the right direction. Both the identified pivot types and triggering factors can
be utilized by startups to make more informed decisions on when and how to pivot. It
also implies the use of different metrics to track customer reaction, product usage, etc.,
to support informed decisions regarding pivoting. To get the actionable metrics is
important especially in the case of product zoom-in pivot, where data analytics can
help to identify the usage of different features.
As for the implication to research, our study provides a set of empirically validated or
derived pivot types that can serve as the conceptual basis for future studies on the pivot
phenomenon. Similarly, the identified triggering factors and their definitions can be utilized as
the conceptual basis to formulate hypotheses in explanatory studies that explore the linkable
between the trigger factors and pivot types.
Our study opens up several new avenues for future research on software startups and
pivoting. The first direction is to collect primary data to validate the pivot types and
triggering factors identified in this study. Primary studies, especially case studies, can
also allow us to collect more contextual information on software startups, such as
which markets they targeted at when pivots happened, at what stage of product
development, in order to achieve more in-depth understanding of the pivot types and
Empir Software Eng

triggering factors. Further studies can also explore the actual process of a pivot and its
consequences (both technical and business). Another future direction is to study
pivoting in software startups in a longitudinal manner. Our study took the snapshots
of pivot instances of the studied startups. Future studies can consider the business and
product development life cycles of a startup, and investigate at which stage pivot most
probably occurs, is most beneficial, and costs least. Taking a longer time span, future
work can also compare how software startups today pivot differently than those, e.g.,
40 years ago, and how triggering factors vary over the time. In addition, an entrepre-
neurial team is one key element of a startup and future studies could be conducted in
order to address how team size and structure are related to different types of pivots.
From a product perspective, it could be interesting to explore how the characteristics
and attributes of software, such as complexity and modifiability, may influence pivoting
decision and process. Moreover, these cases may indicate the need to identify and track
the actionable metrics, and role of these metrics while making pivoting decision. Last
but not least, our findings open up an interesting direction of research regarding
relationship between triggering factors and pivots, to investigate if there is any
causal-effect patterns between triggering factors and pivots. Further studies need to
be conducted to identify other triggering factors causing pivots in software startups, and
investigate how same triggering factors cause different types of pivot.
References
Blank S (2005) The four steps to the epiphany. 1st edn. CafePress
Blank S (2013) Why the lean start-up changes everything. Harv Bus Rev 91(5):64+
Bosch J, Olsson HH, Björk J, Ljungblad J (2013) The early stage software startup development model: a
framework for operationalizing lean principles in software startups, Lecture Notes in Business Information
Processing (LNBIP), pp 1–15
Boslaugh S (2007) An introduction to secondary data analysis. In: Secondary data sources for public health: a
practical guide. Cambridge University Press
Carlson (2011) The real history of Twitter. http://businessinsider.com/how-twitter-was-founded-2011-4/.
Accessed 2 Nov 2015
Cruzes DS, Dybå T, Runeson P, Höst M (2015) Case studies synthesis: a thematic, cross-case, and narrative
synthesis worked example. Empir Softw Eng 20(6):1634–1665
Eisenmann, T, Ries, E, Dillard, S (2012) Hypothesis-driven entrepreneurship: the lean startup. Harvard Business
School Entrepreneurial Management Case No. 812–095
Giardino C, Wang X, Abrahamsson P (2014) Why early-stage software startups fail: a behavioral framework.
ICSOB, pp 27–41
Giardino C, Bajwa SS, Wang X, Abrahamsson P (2015) Key challenges in early-stage software startups. XP,
Helsinki, Finland, LNBIP, pp 52–63
Giardino C, Paternoster N, Unterkalmsteiner M, Gorschek T, Abrahamsson P (2016) Software development in
startup companies: the greenfield startup model. In: IEEE Transactions on Software Engineering
(forthcoming)
Hirvikoski K (2014) Startups pivoting towards value. Data and value-driven software engineering with deep
customer insight. Proceedings of the seminar no. 58314308. Jürgen Münch (ED.) University of Helsinki,
Finland, pp 1–7
Jerkovic JI (2010) SEO warrior: essential techniques for increasing web visibility. O’Reilly Media
Kelly SM (2013) Gaming empire supercell: we pop champagne every time we fail. http://mashable.
com/2013/11/13/supercell-apps-success/#FkbCo9dnpkqZ. Accessed 14 Nov 2015
Kitchenham B (2007) Guidelines for performing systematic literature reviews in software engineering. EBSE
Technical Report. EBSE-2007-01. Software Engineering Group, School of Computer Science and
Mathematics, Keele University, Keele, UK and Department of Computer Science, University of Durham,
Durham, UK
Empir Software Eng

Linberg KR (1999) Software developer perceptions about software project failure: a case study. J Syst Softw
49(2):177–192
Macmillan IC, Zemann L, Subbanarasimha P (1987) Criteria distinguishing successful from unsuccessful
ventures in the venture screening process. J Bus Ventur 2(2):123–137
Marmer M, Bjoern LH, Dogrultan E, Berman R (2011) Startup genome report a new framework for under-
standing why startups succeed
Maurya A (2012) Running lean: iterate from plan a to a plan that works, 2nd edn. O’Reilly Media
McGrath A (2011) Failed by design, https://hbr.org/2011/04/failing-by-design. Accessed 8 April 2016
Mendoza A, Carroll J, Stern L (2010) Software appropriation over time: from adoption to stabilization and
beyond. Australas J Inf Syst 16(2)
Nazar J (2013) 14 Famous business pivots, http://www.forbes.com/sites/jasonnazar/2013/10/08/14-famous-
business-pivots/. Accessed 2 Nov. 2015
Nobel C (2011) Why companies fail and how their founders can bounce back. Working Knowledge, Harvard
Business School, Boston. http://hbswk.hbs.edu/item/6591.html. Accessed 14 Nov 2015
Paternoster N, Giardino C, Unterkalmsteiner M, Gorschek T, Abrahamsson P (2014) Software development in
startup companies: a systematic mapping study. Inf Softw Technol 56(10):1200–1218
Richardson A (2011) STOP CELEBRATING FAILURE, Harvard Business Review, http://www.businessinsider.
com/stop-trying-to-celebrate-failure-2011-3?IR=T. Accessed 8 April 2016
Ries E (2011) The lean startup: how constant innovation creates radically successful businesses. Penguin Group,
London
Runeson P, Höst M (2009) Guidelines for conducting and reporting case study research in software engineering.
Empir Softw Eng 14(2):131–164
Savolainen P, Ahonen JJ, Richardson I (2011) Software development project success and failure from the
supplier’s perspective: a systematic literature review. Int J Proj Manag 30(4):458–469
Shepherd DA, Wiklund J, Haynie JM (2009) Moving forward: balancing the financial and emotional costs of
business failure. J Bus Ventur 24(2):134–148
Sutton SM (2003) The role of process in software start-up. IEEE Softw 17(4):33–39
Terho H, Suonsyrjä S, Karisalo A, Mikkonen T (2015) Ways to cross the rubicon: pivoting in software startups.
SSU’15 workshop, PROFES’15, pp 555–568
Trzesniewski KH, Donnellan MB, Lucas RE (2011) Secondary data analysis: an introduction for psychologists.
Washington, DC, US: American Psychological Association Secondary data analysis: An introduction for
psychologists
Van der Van JS, Bosch J (2013) Pivots and architectural decisions: two sides of the same medal? What
architecture research and lean startup can learn from each other. In Proceedings of Eight International
Conference on Software Engineering Advances (ICSEA’13), Venice, Italy, pp 310–317
Vartanian TP (2011) Secondary data analysis. Oxford, New York
Wang X, Conboy K, Cawley O (2012) BLeagile^software development: an experience report analysis
of the application of lean approaches in agile software development. J Syst Softw 85(6):1284–
1299
Womack JP, Daniel TJ, Roos D (1990) The machine that changed the world. Free Press, USA
Yin RK (2003) Case study research: design and methods. Sage Publications Inc
Yin RK, Heald KA (1975) Using the case survey method to analyze policy studies. Adm Sci Q 20(3):371–381
Yu YW, Chang YS, Chen YF, Chu LS (2012) Entrepreneurial success for hightech start-ups - case study of
taiwan high-tech companies, Palermo, Italy, pp 933–937
Empir Software Eng

Sohaib Shahid Bajwa is a PhD student in Computer Science at Free University of Bozen –Bolzano, Bolzano.
His research interests include software startups particularly focusing upon pivots, empirical software engineering
and agile software development. He received his M.Sc. in Computer Science from Blekinge Institute of
Technology (BTH), Sweden. Contact him at bajwa@inf.unibz.it.
Dr. Xiaofeng Wang is a researcher at the Free University of Bozen-Bolzano. Her research areas include software
startups, software development process and methods, agile and lean development. She can be reached at
xiaofeng.wang@unibz.it.
Empir Software Eng

Dr. Anh Nguyen Duc is a post-doc researcher at the Department of Computer and Information Science,
Norwegian University of Science and Technology. His research interests include data mining, human factors
in distributed teams, agile software development and software startups. He is active in different roles, reviewers,
PC members and co-chairs in several journals, conferences and workshops, including IST, JSS, ESEM, ICGSE,
REFSQ, PROFES, ISEC and IWSS.
Pekka Abrahamsson is a full professor in Software Engineering at Norwegian University of Science and
Technology, NTNU at Trondheim. His research interests are in empirical software engineering, software startups
and innovation in software engineering. He is the chairman of the global Software Startup Research Network. He
can be reached at pekkaa@ntnu.no.
Empir Software Eng