Conference PaperPDF Available

TweetSight: Enhancing Financial Analysts' Social Media Use


Abstract and Figures

Financial analysts utilize information from heterogeneous sources (e.g. corporate filings, economic indicators, news, and tweets) to generate unique trade ideas through a sensemaking process. In this paper, we seek to understand the role of social media in this process. We conducted a semi-structured interview and identified essential benefits and barriers for the primary social media platform used by the analysts - Twitter. Analysts use Twitter as a query exploration tool, as a bellwether to understand sentiment, and to gauge knock-on effects. Drawing from our findings, we developed four scenarios to guide the design of TweetSight. Finally, we evaluated the design of TweetSight by walking analysts through the prototype. Analysts responded positively to anchoring contextual tweets in news articles to facilitate discovery and exploration of Twitter. Our findings and design implications can be applied more broadly to leverage social media for sensemaking, benefiting various business communities.
Content may be subject to copyright.
TweetSight: Enhancing Financial Analysts’ Social Media Use
Rama Adithya Varanasi
Benjamin V. Hanrahan
Pennsylvania State University
State College, PA 16801
Shahtab Wahid
Bloomberg LP
New York city, NY 10022
John M. Carroll
Pennsylvania State University
State College, PA 16801
Financial analysts utilize information from heterogeneous sources
(e.g. corporate lings, economic indicators, news, and tweets) to
generate unique trade ideas through a sensemaking process. In this
paper, we seek to understand the role of social media in this process.
We conducted a semi-structured interview and identied essential
benets and barriers for the primary social media platform used
by the analysts - Twitter. Analysts use Twitter as a query explo-
ration tool, as a bellwether to understand sentiment, and to gauge
knock-on eects. Drawing from our ndings, we developed four
scenarios to guide the design of TweetSight. Finally, we evaluated
the design of TweetSight by walking analysts through the prototype.
Analysts responded positively to anchoring contextual tweets in
news articles to facilitate discovery and exploration of Twitter. Our
ndings and design implications can be applied more broadly to
leverage social media for sensemaking, beneting various business
Human-centered computing Empirical studies in HCI
Financial analysts; Prototyping; Twitter; Human centered design;
ACM Reference format:
Rama Adithya Varanasi, Benjamin V. Hanrahan, Shahtab Wahid, and John
M. Carroll. 2017. TweetSight: Enhancing Financial Analysts’ Social Media
Use. In Proceedings of 8th 2017 International Conference on Social Media &
Society, Torronto, Canada, July 2017 (SMSociety’17), 10 pages.
In the current age of extensive digital connectivity, social media has
become an important medium of communication. As such, use of
First Author
Permission to make digital or hard copies of all or part of this work for personal or
classroom use is granted without fee provided that copies are not made or distributed
for prot or commercial advantage and that copies bear this notice and the full citation
on the rst page. Copyrights for components of this work owned by others than the
author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or
republish, to post on servers or to redistribute to lists, requires prior specic permission
and/or a fee. Request permissions from
SMSociety’17, Torronto, Canada
2017 Copyright held by the owner/author(s). Publication rights licensed to ACM.
social media in the enterprise has been studied to various degrees
]. Researchers have observed that employees use social
media for various work related activities, such as keeping updated
with peers, nding solutions to work problems, and discussing
work-related information with their community [11].
Financial analysts represent an interesting case of social media
use in the enterprise. Financial analysts are responsible for gener-
ating nancial reports and providing stock recommendations for
buying and selling various nancial instruments [
]. A primary
output and revenue generator of these reports are the unique trade
ideas. In generating these ideas, analysts depend heavily on external
information from various sources e.g, corporate lings, company
fundamentals, news from traditional media, and opinions on social
media [
]. However, there has been very little research on how
these nancial analysts make use of social media to accomplish
their work. Therefore, we present the following research question:
: How do nancial analysts leverage social media in sense-
making to generate trade ideas and research reports?
To gain insight into this question, we performed semi-structured
interviews with nancial analysts to understand their perceptions
and usage patterns of social media. Financial analysts are partic-
ularly of interest in this aspect because their work is extremely
time sensitive and high stakes. They need to generate unique and
relevant trade ideas at an extremely fast pace, as these ideas lose
value over a period of time [
]. They work in a high stakes envi-
ronment where their earnings are directly tied to their trade ideas
and research reports [26].
In this investigation we uncovered some of the problems analysts
experience using Twitter as an information source. We derived
four scenarios from these problems, which were built to further
probe and check these problems around past nancial events. These
scenarios were then used to design TweetSight - a prototype which
allows nancial analysts to explore important and contextual tweets
while reading a news article.
: How can we support the practices of nancial analysts in
leveraging social media (Twitter) for sensemaking?
We conducted design analysis interviews with the analysts to
validate the eectiveness of TweetSight. Our study can help re-
searchers and designers understand how social media is used in
sensemaking by organization professionals whose work is time
critical and involves high-stakes knowledge creation.
In this section we present prior work related to our main contri-
butions: How nancial analysts use social media; how to support
sensemaking practices with social media.
1 of 10
SMSociety’17, July 2017, Torronto, Canada Varanasi et al.
2.1 The work of a Financial Analyst
Financial analysts are an important part of the nancial market
ecosystem as they collect, study, explicate, and disseminate infor-
mation to various market participants [
]. They are remunerated in
direct proportion with the performance of their recommendations.
Therefore they keep abreast by gathering information from various
inputs, making sense of key trends from news and interacting with
management [
]. They seek to be responsive to their clients
and gain their approval [
]. There is extensive research on nancial
analysts from nance and accounting perspective [6, 7, 26].
On the other hand, research in the eld of Human Computer
Interaction has largely focused on stock market movement. Studies
have leveraged information (e.g, social media) to predict market
movements using data mining techniques [
]. Particularly
as breaking insights on these platforms have a signicant eect
on stock prices [
]. We seek to ll this gap in the literature by
examining how nancial analysts leverage social media to provide
accurate forecasts.
2.2 Sensemaking research in specialized elds.
Sensemaking is often dened as people "making sense of their
worlds" [
]. In other words, it is how an individual processes
information in order to come to a deeper understanding [
]. Even
though sensemaking has been used in various research areas related
to information [
] and organizational studies [
], it was brought
to the eld of Human Computer Interaction (HCI) by Russell et al.
]. Russell et al. framed sensemaking as the process of forming
meaningful representations and using them to facilitate insights
and actions.
The most established sensemaking model is Pirolli and Card’s
]. In Pirolli and Cards’ model the overall process is divided into
two loops - Foraging and Sensemaking [
]. The foraging loop
involves activities such as seeking, ltration, and extraction of in-
formation, and the sensemaking loop involves the development
of representational schemas used to draw conclusions and form
hypotheses. One active research direction in sensemaking is how
to support creating, organizing, and shaping such external repre-
sentations of knowledge [16].
A lot of work has been done around sensmaking processes and
systems for various specialized domains. For instance: Citesense is
a tool which assists academics in making sense of vast amounts of
literature [
]; Wahlstrom et al. examined sensemaking in safety
critical decisions in car racing [
]; and various other areas such as
web searching [
], intelligence, business analysis [
], and
education [
]. Our study provides new insights into how nan-
cial analysts leverage social media in sensemaking and proposes a
social media tool to support these practices.
2.3 Social Media in the Enterprise
With the rise of social media in recent years, research into how it is
used in the enterprise has increased. Researchers have found that
social media provides opportunities for colleagues to form better
relationships in the organization [
]. One dierence between work
and personal use of Twitter, is that work tweets are often limited
to knowledge sharing [
]. Similarly, updates on Facebook’s wall
updates are used extensively for connecting with employees on a
personal level [
]. On explicitly professional social networks, e.g,
LinkedIn, users are interested more in the current professional lives
of their social network.
However, there are comparatively fewer studies which have
investigated how social media facilitates more primary work pro-
cesses. One example is Mena et al., who found that Twitter is used
by many e-retail platforms to engage and solve customer problems
]. Facebook and Twitter is also used in the academic domain to
recruit participants [
]. Most relevant to this study, are studies
into the growing use of social media by the nancial domain [
We seek to add to this growing body of literature, through inves-
tigating how social media is directly used by nancial analysts in
their work.
To investigate nancial analysts’ existing practices with social me-
dia, we conducted semi-structured interviews with six participants.
Our research partner belongs to a large, privately-held nancial cor-
poration, helped us to select and schedule participants. Our partner
reached out to candidates with extensive experience in the nancial
domain and specic knowledge into how social media is used. In
our pool of participants we had three application specialists experts
and three product managers. Participants had a minimum of 10
years (mean 16 years) of experience in the nancial eld. The inter-
viewees spent an average of nine hours in their work environment
per day.
3.1 Procedure
We conducted the interviews over the telephone and they lasted
anywhere from 45-60 minutes. We provided interviewees with a
brief introduction about what we were doing and how it might ben-
et their organization. Our interview prompts were purposefully
high-level, open-ended questions, such as: “What is your typical
work session like?”, and “What are your deliverables?” We then tran-
sitioned to more specic questions: “How do you use social media
in your deliverables?” Since the interviews were semi-structured,
the questions were only used to guide the conversation. All the
interviews were audio recorded after obtaining permission from the
participants and later transcribed. We used open coding to analyze
the material with a particular focus on our various exploratory
goals: (1) Understanding the common deliverables and practices of
nancial analysts; (2) Usage of social media in their analysis. We
present our ndings in the next section.
3.2 Overview - What does a nancial analyst
As per our interview analysis, the job of a nancial analyst (also
known as an equity, research, or rating analyst) is to generate trade
ideas that create value for their clients’ nancial portfolios. These
trade ideas aim to minimize risk and generate revenue. Financial
analysts are futher divided into two main categories, sell-side and
buy-side. Sell-side analysts perform regular, in-depth research re-
ports about specic companies to give recommendations to buy-side
analysts. In addition to these reports, buy-side analysts conduct
their own research to generate unique ideas and recommendations
2 of 10
TweetSight: Enhancing Financial Analysts’ Social Media Use SMSociety’17, July 2017, Torronto, Canada
to sell to their investors, referred to as ‘trade ideas’. These trade
ideas aim at:
[...] understanding what business is doing, what it
is capable of doing and also understanding where
the street is positioned around the story. As an
analyst, your job is to pitch the story and the trade
idea. - I2
According to I3 and I5, irrespective of the category they belong
to, they start as junior analysts, covering one or two companies in
their portfolio within a sector. As their experience increases, they
start covering more companies (as many as twenty) and sectors.
The set of companies that an analyst covers are referred to as their
‘target list’.
As nancial analysts’ end goal is to recommend whether or not
to buy or sell shares of a company, a company is the single most
important entity for them. Naturally, their research and deliverables
revolve around particular companies in their target list. As such,
the analysts day-to-day activities involve reading news, opinions,
and other information about the companies in their target list. We
found that the information that the analysis gather can be broadly
classied into two types: Events and Themes. In events based re-
search the analysts are interested in how a specic event impacts a
particular company. An example event provided by I5 was when
Tesla announced the release of Model 3; a new budget electric car
which could revolutionize inexpensive and eco-friendly car produc-
tion. In contrast to events, themes are more broad, and refer to a
particular phenomena that may involve multiple events. Themes
also contrast with events in that they are an indicator for a particu-
lar sector (such as automotive), instead of a particular company. For
instance, in 2016 there were multple competing announcements
of aordable electric cars from Tesla, Chevrolet, Audi, and Ford.
In this example, similar events occurred across various companies
which represents a theme across the automotive sector. Based on
our interviews there are a few salient characteristics of the work
which nancial analysts perform.
3.2.1 High stakes job. Interviewees mentioned that analysts
earn most of their income from sales commissions on their trade
ideas and research reports. Therefore their income is directly linked
to the quality and the frequency of their ideas. Analysts need to
predict the price at which a company’s stock will trade as accurately
as possible. This constant pressure to frequently generate high
quality, accurate ideas makes for a high stakes job.
3.2.2 Time constraint. Financial market prices move fast and
unpredictably. As such, trade ideas must be generated with a similar
speed. As time passes, these ideas quickly lose their uniqueness and
ability to generate income. Therefore, analysts have to capitalize
on their unique ideas before other analysts catch up (I2 called this
the crowding aect).
3.2.3 Unique trade ideas. Lastly, according to I1, it is not enough
for analysts to just generate ideas. If everyone has the same idea
at the same time as the analyst, the ideas are not useful. Therefore,
they need to present unique ideas through a deeper understanding
of the market. Analysts gain this insight through meeting with the
management of the companies they are covering, attending impor-
tant conferences in the sector, and reading news and breakthrough
[...]So he starts to develop kind of pieces in his
head. Pieces in his head come from all sort of
sources but a lot of them come from walking down
the street, attending conferences, meeting man-
agement and also reading news articles. - I4
3.3 How are unique trade ideas generated?
The unique nature of nancial analysts’ work means that they
must be on constant lookout for new information from various
information sources. This information is used as input into their
models, allowing them to predict trade-prices and build various
trade-ideas for their research and recommendations.
3.3.1 Locating new information (Input). We found that the basis
of any trade idea lies in understanding the current events or themes
which impact a company or sector respectively. For example, release
of information in a quarterly earnings report by Microsoft can
impact the company’s stock price. To gauge the impact of such an
event or theme, the analyst is always on the lookout for information.
Irrespective of specialization, analysts are constantly scouring
for information to understand events and themes that impact their
target list. Common sources include nancial metrics released by
the company (such as price-to-earning ratio), ocial press releases,
news (e.g, Wall Street Journal), and social media (e.g, Twitter, Face-
book, and blogs). We understood that the information gathered
plays a critical role in the analyst’s job.
News often contains information that can impact their research
model and feed into unique, protable trade ideas. As I2 points out:
“News is the precursor to either unlock value or create an investable
idea. News drives the market.” Analysts start each day reviewing
news articles. They look at the companies in their target list which
have undergone price changes and prioritize those companies when
reading the news. If the analyst nds that a company’s stock has
moved 20%, they would then read news on what is causing the price
uctuation. Based on this understanding, the analyst can then go
and update the portfolio security details or come up with a buying
or selling rationale.
When I would start with the beginning of the day,
I would come in and look my portfolio and I see
which news is associated with each stock. I would
focus rstly on stock price changes. So something
was up and down a lot, I would focus on that news
rst. - I3
To facilitate these practices, many analyst use nancial tools
like Terminal (by Bloomberg) and Platform (by S&P Global Market
Intelligence) which aggregate news from various sources into a
consolidated platform.
3.3.2 Modeling and Analysis. I1 communicated that a model is
a representation of how a particular company is going to perform
nancially in the near future. Analysts take various approaches to
come up with these models. In the top-down approach, the analyst
starts from the entire universe of their investable space and lter it
down to the names that have fundamental, quantitative metrics and
3 of 10
SMSociety’17, July 2017, Torronto, Canada Varanasi et al.
perform analysis at that level. Professionals who use the bottom-
up analysis method look at larger themes in the marketplace by
gleaning insights at the company level. Irrespective of the approach,
the analyst is always adding variables and data to the model to
calculate a more accurate output. In I1’s own words, analysts are
trying to use the model to understand the answer to the following
question: “Do I understand industry well enough to put money in
these trade ideas?” Each model is unique to each analyst because
the variables and strategy which each analyst uses is always specic
to him or her.
3.3.3 Calculating Target Price (Output). The output of the ana-
lyst’s model is a unique price at which a company will be valued in
the near future. It is often called the ‘target price’. Analysts’ income
is directly dependent on how well they can predict this target price.
They present these price estimates in form of research thesis or
trade ideas.
Every time a major event occurs within a particular company,
the analyst has to update the target price predictions as the inputs
of the model change. For example, if a company comes out with
quarterly earnings, the analyst will need to revise their target price
numbers based on this updated data.
3.4 Usage of social media by nancial analysts
Like news, social media is utilized by analysts as another informa-
tion source. Opinions on social media are often used to complement
the information presented in the news. A major dierence between
news and social media is that the information available from social
media is not something on which analysts put their money on, but
rather see as a “head start opportunity” in performing research into
newly opened avenues. Within social media, the most frequently
used platforms are discussed below.
3.4.1 Blogs. Blogs provide a place for readers to interact with
authors, to collaborate with peers, and to share the content with
various other readers [
]. Analysts tend to develop trust in par-
ticular experts and fellow analysts. They follow these professional’s
steps and advice, including information from informal writings
such as blogs. These blogs provide important information, however,
there is an obvious presence of personal judgment involved when
analysts develop the list of blogs they follow - these may not be
same across analysts even within the same domain. The following
quote by I1 shows the extent to which trust plays a factor.
For example, Wall Street Journal can be good for
one industry. Some of the columnists might be
good. They are experienced in writing, for exam-
ple, say one industry as opposed to something else.
They have a track record and analysts follow that.
Sometimes, they follow just other analysts. If the
analyst changes from...these subscribers change
too - I1
3.4.2 Collaborative communities. I2 and I5 mentioned collabo-
rative communities as another source of information. These mostly
take the form of internet forums where a group of analysts pitch
their nancial ideas and discuss them, e.g, Yahoo communities. In
communities like these, users often post with pseudo-handles to
create anonymity as the ideas and recommendations they provide
might impact their personal reputation. Due to this, accountability
is not as prevalent. At the other end of spectrum there are high-
prole forums referred to as ‘Investor Clubs’ where access is limited
to specic groups of professionals to share investment ideas for
couple of reasons. I2 mentioned that one reason is for neophyte
analysts to build their personal brand. Analysts who present an
idea which predicts the future markets well get recognition in their
community. Other times, professionals, like investment managers,
want everyone else to see their personal view on an issue and help
others understand what the market is currently missing.
3.4.3 Twier. Interviewees unanimously cited Twitter as the
most commonly used social media tool by the analysts. One thing
that is unique to Twitter usage among the nancial industry is
the use of cashtags, which are special hashtags that use the ticker
symbols for each company (e.g, $AAPL and $MSFT). Tweets about
nancial content often contain these cashtags. I3, I4, and I5 men-
tioned that Twitter is used for both quantitative and qualitative
analysis. In quantitative analysis, text mining programs are used to
provide directional and predictive information. In our interviews,
we uncovered a few specic reasons for how and why Twitter is
used their qualitative research, which we are more interested in.
Twitter for breaking information - There are times when social
media will pick up an event before traditional news wires, as there
are individual experts who are tweeting about an event in an un-
ocial manner. These unocial tweets by experts hold potential
value to the analysts as they give them a head start.
It is ‘rst-to-market’. You have the advantage of
rst knowing something. If you are the rst to
know, you can take the opportunity before anyone
else can take it. In that sense, it became primary
source for most of the fund managers and analysts
I know. - I1
Sometimes these tweets, especially those by established experts
in the eld, can also impact the market and move the price of stocks
in a particular direction. I6 provided an example where a famous
fund had aected the movement of Apple stock.
What you get out of social media is that often you
will have people who have very specic opinions
on a stock that they tweet. For example, the fa-
mous fund manager like Carl Ichan (@Carl_C_Ichan)
he might say Apple is worth $200 a share and in
fact he said exactly that. He published a tweet, he
said it is vastly undervalued. Once the investor can
say that it is Carl Ichan who is doing the tweeting
then the stock starts rallying tremendously - I6
Query expansion - In addition to providing access to breaking
news, tweets also serve as a mechanism to understand the ques-
tion of ‘why’ something is moving in the market. There are times
when stocks move tremendously, but the traditional news does not
provide an immediate reason. Tweets by knowledgeable people
in that industry can help analysts ll these lapses and explain the
movements of nancial markets.
For instance, I4 mentioned there are instances when there is a
big move in the pharma stock industry and stock of a particular
company moves 3%. The analyst does not know why the stock is
4 of 10
TweetSight: Enhancing Financial Analysts’ Social Media Use SMSociety’17, July 2017, Torronto, Canada
moving - even the news media is not covering the event. In those
situations, tweets by experts in the Pharma industry can possibly
provide an answer for the analysts.
There are times when social media will pick up
before the traditional news wires will and in those
instances you are involved in the trade in a stock
and the stock is moving you don’t really know
why - I2
Bellwether - I5, who specializes in the retail market sector, re-
ported that many companies use Twitter in the retail and fashion
sector to engage customers and inform them about their new prod-
uct lineup, interesting oers, etc. This engagement is used by ana-
lysts as they often provide a bellwether as to how well a company
is attracting customers.
Tweets from reliable sources are also used in research reports
as a potential reference point. When Bill Gross, a famous nan-
cial manager, tweets about American infrastructure falling short of
1.4 trillion by 2025, it has a signicant impact and analysts might
reference this tweet. Two interviewees explained that data gath-
ered from Twitter is often used as input into the analyst’s model.
Additionally, nancial companies like theirs have created special
teams who work exclusively on Twitter feeds. As part of their job,
they often manually vet important tweets to push through nancial
applications like Terminal.
3.5 Problems discovered with Twitter usage
Even though there are denitive benets to analysts using Twitter,
there was a hesitancy by analysts in relying on social media to reap
these benets. In this section, we present the major problems which
lead to hesitation to use Twitter.
3.5.1 Problem-1: Diicult to Assess Knock-on Eects. Incidents
related to one company often have repercussions on other com-
panies in the same sector, and can even extended to companies in
other sectors at times. This phenomenon is often called the “knock-
on eect.” According to I4, the Mexican government’s decision to
deregulate oil drilling o the Gulf Coast had a big impact on the ce-
ment industry. This is because this type of drilling increases demand
for cement for the lling process. Moreover, it is not cost-eective
to ship cement from the US or China, therefore, the obvious choice
for many oil companies was Cemex - a major Mexican cement
supplier. The deregulation eect had a strong knock-on eect on
share prices of Cemex.
Successfully predicting tertiary eects such as these creates a
rich nancial opportunity for the analyst. The trade ideas gen-
erated from these knock-on eects are highly ecacious in that
they require a deep understanding of events happening around the
company, and opinions of a variety of subject matter experts. As
a complement to traditional news, Twitter can provide an analyst
with additional information about the event or theme in the form
of expert analysis, recommendations, insider information to deduce
such knock-on eects. Currently, analysts consume these tweets in
an out-of-context fashion because they are presented based on sub-
scription and chronology. Therefore, it is prohibitively dicult and
time-consuming to eectively leverage Twitter for this purpose.
3.5.2 Problem-2: Diicult to Discover Relevant Tweets. There are
a number of reasons that it is dicult for analysts to nd the correct
experts currently tweeting important information.
Subscription structure - In Twitter, a user can have many followers.
When you follow a user, you get all of the tweets that they write.
Therefore, to eectively utilize Twitter to nd the next breaking
news analysts would need to predict which important people will
break the story.
Too much noise - An alternative strategy, is to search for specic
keywords related to a current event or theme. However, this also
proves dicult due to the large amount of tweets and it takes a
prohibitively large amount of time to locate the few important
3.5.3 Problem-3: Small Impact Window. As the market moves
quite fast, any breaking tweets lose their value once the the informa-
tion is known widely thereby causing ‘crowding-aect’. Therefore,
the manner in which analysts currently consume tweets takes too
long to be useful.
Based on our ndings from our interview study, we present our
design study of TweetSight - a tool that assists nancial analysts by
situating relevant tweets within news articles. As part of this design
study, we conducted walk-through sessions of the initial prototype
with nancial analysts to determine the potential ecacy of our
The aim of TweetSight (as show in the Figure 1) is to situate
contextually relevant tweets alongside news articles. We chose news
articles as the anchor point as these articles represented a common
resource that nancial analysts organized their information seeking
and sensemaking around. As soon as the analyst opens a news
article, TweetSight displays the most relevant tweets alongside it.
As an example, if the user is reading an article titled “People are
already lining up in tents to pre-order the Tesla Model 3,” they will
see tweets about Tesla’s Model 3 on the right-hand side. These
tweets are ltered based on additional parameters, such as key
people who tweet frequently in the sector, individuals whom the
analyst follows, and the presence of Cashtags within the tweets.
Once the tweets are loaded, the user can rene the tweets based on
either keywords in the news or in the co-occurring statistics.
In our Scenario-Based Design process, we developed ve dier-
ent problem scenarios, each of which was grounded in our previous
interviews. Out of these problem scenarios, we developed four cor-
responding interaction scenarios. We then constructed a prototype
based on these scenarios. While our prototype was interactive, in
order to receive quick feedback on the design, we only supported
the activities presented in the scenario.
4.1 Design Objectives
4.1.1 Objective-1 Discovering Knock-on Eects. The co-occurring
statistics bar lets the user to observe related companies and helps
them understand how events or themes relate with other compa-
nies or people. Additionally, the user can rene these tweets in
order to explore tweets which discuss the co-occurring entities.
These tweets can assist analysts in understanding the eects a
focus company is having on the co-occurring elements.
5 of 10
SMSociety’17, July 2017, Torronto, Canada Varanasi et al.
Figure 1: User Interface components of TweetSight Application: a) News article section, b) Contextual Tweets section, c) Co-
occurrence statistics of the news article.
4.1.2 Objective-2 Leveraging the short term market impact. The
option to subscribe and receive notications for future contextual
tweets using the subscription feature allows nancial analysts to
get to important tweets much faster. This process can provide them
with an opportunity to take advantage of such tweets before the
crowding-eect takes place.
4.1.3 Objective-3 Reducing the eort required for query explo-
ration. Using the highlighted keywords in the article, the user can
re-formulate the tweets present on the right hand side. It allows
users to skim through an article by looking at the highlighted key-
words. Additionally, as they are reading the news in detail, the user
can perform query exploration by loading relevant tweets using
the keywords present in the article.
4.2 Design of TweetSight
The interface is divided into three main components (see Figure1);
(a) news article section, (b) contextual tweets section, and (c) co-
occurrence statistics of the news article.
4.2.1 News article section. The news article section (1) contains
the actual article where the names, organizations, and companies
in the article are highlighted as keywords. In the Tesla example
mentioned above, Elon Musk is one of the highlighted keywords.
The classication and uniqueness of these keywords is determined
using Named Entity Recognition (NER) and TF-IDF, two common
algorithms which are used in textual analysis of the data. As the
user scrolls through the article, they can click on these highlighted
keywords to rene and reformulate the tweets on the right hand
side, making them more specic. While the user reads through
the Tesla article, they may observe the company name - ‘General
Motors’ as a highlighted keyword. If the user adds General Motors
to the current query, we display tweets that mention both Tesla
and General Motors.
4.2.2 Contextual Tweets. The contextual tweet section displays
the most relevant tweets regarding the current news article. The
tweets are extracted from Twitter using keywords from the article.
In order to restrict tweets to the nancial domain, the tweets are
further ltered by veried accounts and Cashtags. We further re-
stricted the tweets by categorizing them into prioritized and other
tweets. Tweets are labeled and shown in the prioritized section if
they are published by curated tweet handles who have consistently
provided valuable opinions in the sector or by the accounts the
user personally follows. The list can be edited and maintained in
the settings panel. For the purpose of our study design, we used
the list provided by the partnered nancial company. The user has
the option to add tweet handles to this list by clicking on the add
person icon beside the tweet. Curated accounts are shown with
a star beside the tweet. The above ltration process reduces the
volume of tweets, making it easier for the analysts to nd valuable
and relevant information. Users are also able to sign-up for future
contextual tweets related to the current article. For instance, if the
user subscribes to the Tesla news article, they will receive noti-
cations for any tweets published after they subscribe. In this way,
analysts do not need to actively monitor an article to see how it
4.2.3 Co-occurring Statistics. Co-occurring statistics presents
an expandable information bar at the bottom to show frequently
co-occurring companies and entities. These co-occurring values are
computed by performing text mining on the news corpus within the
past month. For instance, in Tesla’s article, the most co-occurring
companies are GM, Ford, and Nissan. Alongside each keyword, the
level of co-occurrence is shown besides each result. Similar to the
highlighted keywords in the news article, these results can also be
used by the user to rene and reformulate their tweets.
6 of 10
TweetSight: Enhancing Financial Analysts’ Social Media Use SMSociety’17, July 2017, Torronto, Canada
4.3 Design evaluation of TweetSight
4.3.1 Method. In order to understand how TweetSight may help
to address the problems discovered in our rst study we invited
subject matter experts to provide feedback on our tool. For the
design evaluation study, we recruited six dierent participants from
the same nancial corporation. To maintain similar composition to
our previous study, we recruited three application specialists, two
nancial analysts, and a nancial product manager. The average
experience in the company for these participants was six years (max
= 12 years, S.D = 3.3 years). The participants who were application
specialists and product managers had more than 5 years of working
experience as a nancial analysts before joining the company.
4.3.2 Procedure. The design evaluation sessions were conducted
virtually using the WebEx application. Each evaluation session
lasted between 60 - 80 minutes. In the sessions, each participant
was presented with three interaction scenarios covering the prob-
lems discovered in the previous study. An example of an interaction
scenario is reproduced below:
Bill is a sell-side analysts who is covering Apple in his portfolio.
He comes across the news item labeled “Apple, The FBI And iPhone
Encryption: A Look At What’s At Stake” on Mar17, 2:18pm. As
Apple is in his portfolio item, he is interested in the news item
which is becoming a big story. The news item talks about how the
FBI vs. Apple is going big about the issue of privacy about which
all the tech giants are concerned. He is curious in understanding
what stance are other technology giants going to take.
In this scenario, as Bill is navigating through news, his attention
is grabbed by Google which is highlighted as one of the key word
in the body. He becomes curious by the presence of the competitor
company in the news article. He then clicks on the keyword to
re-formulate tweets. The resultant tweets contain nexus of ‘Apple’,
‘FBI’, ‘encryption’, and ‘Google’. User can observe a tweet released
by Sundar Pichai (CEO of Google) showcasing support and declar-
ing stance with Apple. Tweets of support like these can help Bill
realize the positive support of other technology companies on Ap-
ple in relation to the FBI feud. Additionally, he can click on the
button on the top to subscribe to the future tweets which arising in
context to the news article.
For each scenario, the participant was presented with the prob-
lem, and understanding of it was ensured. Then, we presented a
step-by-step walkthrough of how an analyst could arrive at the
solution using TweetSight. We chose to do a walk through as we
had a limited time with each participant instead of observing how
they use the tool, we were more interested in the concepts of the de-
sign as opposed to its usability. After completion of each scenario,
participants were asked a series of questions. For example, one
question for the aforementioned Tesla scenario was: “How does an
analyst perceive negative perception (as tweeted by few tweet han-
dles) occurring towards the Bolt while navigating through the Tesla
news article? What kind of impact does it create on their analysis
of the trade price?” All of the interviews were audio-recorded with
participant permission and transcribed for analysis. We examined
the data to nd how participants responded to our questions.
4.4 Findings
4.4.1 Helping to Discover Knock-on Eects. Interviewees found
it useful to view highlighted keywords within the news article
and using these keywords to nd new social media opinions, as
it allowed them to create new exploration threads without losing
focus on the current one. They felt that this would allow nancial
analysts to recognize knock-on eects by helping them nd tweets
that best informed their research thesis. I7 felt that this method
of showcasing tweets was much more ecient and productive
for analysts compared to the traditional method of subscribing to
Primarily for our users, they are always thinking
about what is the investment thesis here. So, let’s
say the story was negative for Apple (talking about
scenario-1). If I know that Apple is in trouble, I
would want to think of investment thesis I can
generate from that. Maybe I don’t want to short
Apple, because the news is already out and the
whole world is shorting Apple. You are not going
to make any money by shorting Apple. Maybe
there is a supplier of Apple or a competitor of
Apple that has been mentioned in article whose
related opinion might be a good information thesis
- I9
4.4.2 Finding New Perspectives. Based on the interviewee’s in-
sights, we discovered that nancial analysts are primarily concerned
in understanding the development of a particular event from new
perspectives. They felt that this understanding was easier as Tweet-
Sight improved the discoverability of tweets. They felt that using
our tool’s design, the analyst could discover tweets of support or
disagreement by important tweet handles such as Sundar Pichai
(in scenario-1), the CEO of Google, a Wall Street Journal reporter
or another respected nancial analyst. They were more skeptical
of tweet handles that they did not know.
The rst tweet is... (referring to a tweet which
has nancial numbers) this tweet has more impact
than the other two because of getting to fundamen-
tal numbers that might change someone’s analysis.
The analysis would themselves know that but if
they don’t, that is the tweet from the article to pull.
But again, my point is that the content like that is
more impactful - I7
They also felt that TweetSight would allow analysts to form their
own perception. However, I3 felt that the analyst might need to
verify these opinions with fundamental numbers before presenting
these perspectives in a research resport.
But if you had a very technical article about a very
niche portion of cyber-security of semi-conductors,
then using this technology would help to nd the
smart people whom you did not know existed
on Twitter that aren’t in the social consciousness
rather than big news event where a lot of content
is being produced where most of it is probably
going to be regurgitating what they have in the
news article. - I8.
7 of 10
SMSociety’17, July 2017, Torronto, Canada Varanasi et al.
Figure 2: User Interface components of TweetSight Application.
4.4.3 Additional Context. They also thought that TweetSight
will be useful when an analyst reads articles containing jargon or
technical nuances beyond their understanding. In this case, having
the opinion of a subject matter expert on the right hand side can
help the analyst to decipher the news article as a part of query
exploration process.
I don’t think it improves the reliability, it certainly
improves the discoverability. It makes it much eas-
ier for the customers to nd tweets that they care
about - I9
4.4.4 Keeping Abreast. The interviewees felt that subscribing
to get future tweets would allow them to follow developments on
the y. Analysts must follow any developing themes, however, it is
dicult to constantly monitor numerous themes. Therefore, they
appreciated the ability to receive notications about new tweets
as notications without having the need to go back the the news
That is pretty handy. It is saving me the time which
I will spend you know putting in keywords or
looking for it myself. If I can get relevant alert
setup which noties me pro-actively it is helpful. -
4.4.5 Making Connections. Interviewees found the idea of co-
occurring statistics most relevant and useful as a design feature
in the tool. I11 felt that presenting co-occurring companies and
entities below the news allowed the analysts to gain an overall idea
about the direction of themes associated with the company. He
saw this happening as an analyst tries to link opinions on Twitter
with the companies and entities mentioned in the co-occurrence
I think it is useful specically if you are talking
about co-occurring products or those kinds of
things. For example, if I want to see what people
are saying about the new IPhone, and I am read-
ing an article on Apple; I think I can quickly add
Iphone co-occurring keyword to see what people
are saying exclusively about Iphone - I7
I8 felt that this feature will be more useful when digging deeper
into research and creating new perspectives, versus uncovering
tweets which break information on social media. Overall, analysts
talked about its potential use case in understanding the develop-
ments between rival companies in a particular sector. They felt that
this feature has the potential to uncover valuable tweets that cover
the competition space between two companies but are not very
easy to nd.
Based on the results of our rst study, we feel there is a strong scope
of increasing the potential of Twitter in nancial analysis. We felt
that there are many attributes of Twitter which make it very useful
as an information source for qualitative analysis. TweetSight’s ini-
tial design was aective in improving the overall discoverability of
relevant tweets.
Additionally, the comprehensive work process followed by -
nancial analysts to come with novel trade insights follow a notional
model of sensemaking loop [18].
Information gathering
- Our ndings show that analysts
gather information from various external sources such as
news, social media, nancial metrics, press releases etc.
Based on the companies in their portfolio they search and
lter information from these sources. They collect relevant
evidences in ‘shoeboxes’ [
]. Analysts often use notes or
electronic documents to collect these pieces of evidence.
Representation in schemas
- Analysts use selective evi-
dences from the shoebox to build their schema or analysis
model. This process happens in an iterative manner as the
analyst constantly adds new pieces of evidences to their
model from the shoebox. They can also re-represent the
nancial models to improve their analysis results.
Insight Development
- The schema is often used to build
insights and support hypotheses. In our ndings, analysts
often used models to develop their theories and research
Knowledge product
- A presentation or publication is
made from the insights generated for their audience. Fi-
nancial analysts reveled that they do the same by pitching
trade idea presentations and suggest investment strategies.
8 of 10
TweetSight: Enhancing Financial Analysts’ Social Media Use SMSociety’17, July 2017, Torronto, Canada
6.1 In Twitter we Trust?
I7 and I9 raised the reliability of the authors of dierent tweets
as an issue. Despite the assurance of veried accounts, they were
hesitant to use tweets that they did not know. We plan to improve
this design by showing additional information such as previous
historical tweets, facts presented in Wikipedia Infobox about the
user, and an option to give endorsements. These design features
mimic some of their existing practices in their online communities
to improve the provenance of tweets.
6.2 Notication Overload
Another challenge which I12 mentioned is to control the number of
alerts which a user receives when subscribing to multiple contextual
tweets of multiple news items. This challenge can be overcome by
providing digests at specied intervals, instead of a notication for
each tweet.
6.3 Rened Renement
Interviewees provided the potential use of rening news items in
reverse using the contextual tweets. They felt that using interesting
tweets they can rene and nd more contextual news items related
to it, thereby saving time.
6.4 Beyond Cashtags
Interviewees also mentioned that cashtags are mostly used by com-
panies listed in American stock exchanges which will limit the
tweets results to a particular geographical location. To overcome
this, additional lter options such as number of retweets and likes
are required to limit the tweets to minimum.
Financial analysts have a limited amount of time that they can
dedicate to activities which do not contribute directly to their work
]. Therefore, recruiting several experts, each with extensive
experience, was a major challenge in conducting the two studies.
We also used medium delity prototypes as a medium to receive
quick feedback on our design solutions. Even though they were
quite eective, they are limited in their range of activities they pro-
vide to the users. Therefore, we plan on performing a longitudinal
study after further developing the tool.
Our design research study revealed how nancial analysts utilize
social media in various ways to perform sensemaking. We also
discovered essential problems around one of the social media tool
Twitter, decreasing its potential to be used as a major information
source by these analysts. Based on this information, we showcase
initial design prototype of TweetSight to reduce the problem and
assist in the sensemaking process. We feel our ndings provide a
signicant impact on the society of nancial analysts by providing
them quality ideas and reducing the overall time for their work
Saleema Amershi and Meredith Ringel Morris. 2008. CoSearch: a system for
co-located collaborative web search. In Proceedings of the SIGCHI Conference
on Human Factors in Computing Systems. ACM, 1647–1656.
Antonio R. Anaya, Manuel Luque, and Manuel Peinado. 2016. A visual recom-
mender tool in a collaborative learning experience. Expert Systems with Applica-
tions 45 (March 2016), 248–259.
Michelle Q. Wang Baldonado and Terry Winograd. 1997. SenseMaker: an
information-exploration interface supporting the contextual evolution of a user’s
interests. In Proceedings of the ACM SIGCHI Conference on Human factors in com-
puting systems. ACM, 11–18.
Jyotirmoyee Bhattacharjya, Adrian Ellison, and Sonali Tripathi. 2016. An explo-
ration of logistics-related customer service provision on Twitter: The case of e-
retailers. International Journal of Physical Distribution & Logistics Management 46,
6/7 (July 2016), 659–680. DOI: 2015-0007
Elizabeth Blankespoor, Gregory S. Miller, and Hal D. White. 2014. The Role of
Dissemination in Market Liquidity: Evidence from Firms’ Use of TwitterâĎć.
Accounting Review 89, 1 (Jan. 2014), 79–112.
Lawrence D. Brown, Andrew C. Call, Michael B. Clement, and Nathan Y. Sharp.
2015. Inside the Black Box of Sell-Side Financial Analysts. Journal of Accounting
Research 53, 1 (March 2015), 1–47.
Nerissa C. Brown, Kelsey D. Wei, and Russ Wermers. 2014. Analyst Recommen-
dations, Mutual Fund Herding, and Overreaction in Stock Prices. Management
Science 60, 1 (Jan. 2014), 1–20. DOI:
Hailiang Chen, Prabuddha De, Yu Jerey Hu, and Byoung-Hyoun Hwang. 2014.
Wisdom of crowds: The value of stock opinions transmitted through social media.
Review of Financial Studies 27, 5 (2014), 1367–1403.
David Cutler, James Poterba, and Lawrence Summers. 1988. What Moves Stock
Prices? Technical Report w2538. National Bureau of Economic Research, Cam-
bridge, MA.
Brenda Dervin. 1983. An overview of sense-making research: Concepts, meth-
ods, and results to date. The author.
Kate Ehrlich and N. Sadat Shami. 2010. Microblogging Inside and Outside the
Workplace.. In ICWSM.
Lucy L. Gu, Dorothy Skierkowski, Paul Florin, Karen Friend, and Yinjiao Yi. 2016.
Facebook, Twitter, & Qr codes: An exploratory trial examining the feasibility
of social media mechanisms for sample recruitment. Computers in Human
Behavior 60 (2016), 86–96.
Michael Hagenau, Michael Liebmann, and Dirk Neumann. 2013. Automated
news reading: Stock price prediction based on nancial news using context-
capturing features. Decision Support Systems 55, 3 (June 2013), 685–697.
W. Glynn Mangold and David J. Faulds. 2009. Social media: The new hybrid
element of the promotion mix. Business Horizons 52, 4 (July 2009), 357–365.
Sharon Meraz. 2009. Is There an Elite Hold? Traditional Media to Social Media
Agenda Setting Inuence in Blog Networks. Journal of Computer-Mediated
Communication 14, 3 (April 2009), 682–707.
Sharoda A. Paul and Meredith Ringel Morris. 2009. CoSense: enhancing sense-
making for collaborative web search. In Proceedings of the SIGCHI Conference
on Human Factors in Computing Systems. ACM, 1771–1780.
Peter Pirolli and Stuart Card. 1999. Information foraging. Psychological review
106, 4 (1999), 643.
Peter Pirolli and Stuart Card. 2005. The sensemaking process and leverage
points for analyst technology as identied through cognitive task analysis. In
Proceedings of international conference on intelligence analysis, Vol. 5. 2–4. https:
Peter Pirolli and Daniel Russell. 2011. Introduction to this Special Issue on
Sensemaking. Human-Computer Interaction 26, 1 (Jan. 2011), 1–8.
Eduardo J. Ruiz, Vagelis Hristidis, Carlos Castillo, Aristides Gionis, and Alejandro
Jaimes. 2012. Correlating nancial time series with micro-blogging activity. ACM
Press, 513. DOI:
Daniel M. Russell, Mark J. Stek, Peter Pirolli, and Stuart K. Card. 1993. The
cost structure of sensemaking. In Proceedings of the INTERACT’93 and CHI’93
conference on Human factors in computing systems. ACM, 269–276. http://dl.acm.
Meredith M. Skeels and Jonathan Grudin. 2009. When social networks cross
boundaries: a case study of workplace use of facebook and linkedin. In Proceedings
9 of 10
SMSociety’17, July 2017, Torronto, Canada Varanasi et al.
of the ACM 2009 international conference on Supporting group work. ACM, 95–104.
Eugene Soltes. 2014. Private Interaction Between Firm Management and Sell-
Side Analysts. Journal of Accounting Research 52, 1 (March 2014), 245–272.
Ward van Zoonen, Joost W.M. Verhoeven, and Rens Vliegenthart. 2016. How
employees use Twitter to talk about work: A typology of work-related tweets.
Computers in Human Behavior 55 (Feb. 2016), 329–339.
Mikael Wahlstrom, Antti Salovaara, Leena Salo, and Antti Oulasvirta. 2011.
Resolving Safety-Critical Incidents in a Rally Control Center. Human-Computer
Interaction 26, 1 (Jan. 2011), 9–37.
Rina R. Wehbe, Shahtab Wahid, Siddharth Gupta, and Edward W. Ishak. 2016.
Tackling User Research Challenges within the Finance Industry. ACM Press,
796–803. DOI:
Karl E. Weick. 1995. Sensemaking in organizations. Sage Publications, Thousand
Feng Xiong, Acklesh Prasad, and Larelle June Chapple. 2016. The economic
consequences of corporate nancial reporting on Twitter. (2016). http://eprints.
Steve Y. Yang, Sheung Yin Kevin Mo, and Anqi Liu. 2015. Twitter nancial
community sentiment and its predictive relationship to stock market movement.
Quantitative Finance 15, 10 (Oct. 2015), 1637–1656.
P. Zhang and D. Soergel. 2016. Process patterns and conceptual changes in
knowledge representations during information seeking and sensemaking: A
qualitative user study. Journal of Information Science 42, 1 (Feb. 2016), 59–78.
Xiaolong Zhang, Yan Qu, C. Lee Giles, and Piyou Song. 2008. CiteSense:
supporting sensemaking of research literature. In Proceedings of the SIGCHI
Conference on Human Factors in Computing Systems. ACM, 677–680. http:
Dejin Zhao and Mary Beth Rosson. 2009. How and why people Twitter: the role
that micro-blogging plays in informal communication at work. ACM Press, 243.
10 of 10
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
Purpose – The success of e-retailers is intrinsically linked to the effectiveness of their logistics processes which, inevitably, involve third party service providers. As the most tangible representative of the e-retailers it is inevitable that customers expect the e-retailer to resolve delivery queries, including on social media platforms. The purpose of this paper is to investigate the effectiveness of e-retailers’ logistics-related customer service interactions on Twitter with a view towards identifying effective and ineffective social media customer service strategies. Design/methodology/approach – The design and public nature of Twitter encourages organic conversations between e-retailers and customers as well as between customers and other customers. The methodology applied here accounts for this by collecting and analysing interactions within and as part of conversations, not as independent observations. In total, 203,349 tweets were collected from 22 of the most popular e-retailers. A random sample of 5,000 logistics-related conversations (16,998 tweets) is used for the analysis presented here and forms a foundation for future research. Findings – Conversations are initiated by customers on the basis of 24 event triggers which can be categorised as occurring either before or after an order is delivered. These can be general queries or related to a specific order or delivery issue. The paper identifies a number of significant findings such as the extent to which e-retailers and logistics providers redirect customers to other channels to resolve queries, ignoring the implicit preference by customers to use Twitter to resolve their problem. Similarly, the lack of interactions between e-retailers and their logistics providers within the Twitter platform to help resolve customer queries results in ineffective customer service. Practical implications – The study identifies the way in which e-retailers can substantially improve the effectiveness of the customer service they provide on Twitter by ensuring that customer queries can be resolved within the platform and by working with their logistics partners to do the same. This is critical since problems may be directed to the e-retailer or the logistics provider but both companies jointly suffer the consequences of poor customer service. Originality/value – The study examines a hitherto underexplored aspect of retail logistics – the social media-based customer service activities of e-retailers. Methodologically, the study is rooted in the acknowledgement that interactions on Twitter form conversations and analyses should take this into account. This is a distinctly different approach from existing Twitter-related studies which conduct an automated sentiment analysis of tweets. This approach reveals a rich picture of interactions and, importantly, identifies where conversations between e-retailers begin, how they develop and how they conclude.
Conference Paper
Full-text available
We present a case study illustrating how a user experience (UX) team performs user research in the finance industry. In particular, we focus on the impact of salespeople and financial professionals on how the research is conducted. Challenges stemming from this--such as recruitment, time constraints, and conflicting expectations--and potential ways to mitigate them are discussed. Our work contributes to an understanding of how to do research in time-sensitive, high pressure environments while also working with gatekeepers to accessing users.
Full-text available
The construction of knowledge representations during sensemaking resembles meaningful learning in which conceptual changes to knowledge structure take place in various forms. Guided by a cognitive process model of sensemaking expanding prior models with ideas from learning and cognitive psychology, we conducted a qualitative user study of 15 participants with news writing and business analysis tasks to investigate the evolvement of their knowledge structures. We collected and analysed think-aloud protocols along with recorded screen activities, intermediate work products including notes and concept maps, and the final reports. Findings suggested that: (a) the sensemaking process can be viewed as composed of several iterations that fall into nine slightly varied common patterns, which make up the components of sensemaking; (b) conceptual changes fall into three broad classes – accretion, tuning and restructuring; and (c) changes in forms of representation seem to assist in sensemaking. These findings provide insights for system design that assists in sensemaking and intelligent use of information.
Full-text available
Twitter, one of the several major social media platforms, has been identified as an influential factor to financial markets by multiple academic and professional publications in recent years. The motivation of this study hinges on the growing popularity of the use of Twitter and the increasing prevalence of its influence among the financial investment community. This paper presents an empirical evidence of the existence of a financial community on Twitter in which users' interests align with the financial market related topics. We establish a methodology to identify relevant Twitter users who form the financial community, and we also present the empirical findings of network characteristics of the financial community. We observe that this financial community behaves similarly to a small-world network, and we further identify groups of critical nodes and analyze their influence within the financial community based on several network centrality measures. Using a novel sentiment analysis algorithm, we construct a weighted sentiment measure using tweet messages from these critical nodes, and we discover that it is significantly correlated with the returns of the major financial market indices. By forming a financial community within the Twitter universe, we argue that the influential Twitter users within the financial community provide a better proxy between social sentiment and financial market movement. Hence, we conclude that the weighted sentiment constructed from these critical nodes within the financial community provides a more robust predictor of financial markets than the general social sentiment.
Full-text available
As more and more information becomes available in larger, ever more rapid flows, the skills of sensemaking are no longer required just of specialists in intelligence analysis, but increasingly of everyone. We all live in a data world with continually flowing streams of information. These articles are the beginning of a coordinated effort to build better tools and better theories of how to cope. Sensemaking is increasingly a part of all our lives; this collection shows a few ways to understand what's happening to us.
This study examined the effectiveness of three social media based recruitment channels for sampling rural adolescent populations for online health research. At present, there is no consensus on the optimal social media based vehicle for recruiting adolescents due to limited research. This exploratory study compared Facebook ads, Twitter, and QR code postcards at three different but demographically similar rural high schools. The results showed that QR codes had the highest response percentage and the lowest cost per recruited participant, whereas Twitter had the lowest response percentage and Facebook had the highest cost per recruited participant. Although this is the first time QR codes were examined in this context, it seemed to show potential in online health research. The findings are interpreted from a variety of theoretical and conceptual frameworks. Applications of each recruitment channel are discussed and suggestions are provided for future research.
Collaborative learning incorporates a social component in distance education to minimize the disadvantages of studying in solitude. Frequent analysis of student interactions is required for assessing collaboration. Collaboration analytics arose as a discipline to study student interactions and to promote active participation in e-learning environments. Unfortunately, researchers have been more focused on finding methods to solve collaboration problems than on explaining the results to tutors and students. Yet if students do not understand the results of collaboration analysis methods, they will rarely follow their advice. In this paper we propose a tool that analyzes student interactions and visually explains the collaboration circumstances to provoke the self-reflection and promote the sensemaking about collaboration. The tool presents a visual explanatory decision tree that graphically highlights student collaboration circumstances and helps to understand the reasoning followed by the tool when prescribing a recommendation. An assessment of the tool has demonstrated: (1) the students collaboration circumstances showed by the tool are easy to understand and (2) the students could realize the possible actions to improve the collaboration process.
Our objective is to penetrate the “black box” of sell‐side financial analysts by providing new insights into the inputs analysts use and the incentives they face. We survey 365 analysts and conduct 18 follow‐up interviews covering a wide range of topics, including the inputs to analysts’ earnings forecasts and stock recommendations, the value of their industry knowledge, the determinants of their compensation, the career benefits of Institutional Investor All‐Star status, and the factors they consider indicative of high‐quality earnings. One important finding is that private communication with management is a more useful input to analysts’ earnings forecasts and stock recommendations than their own primary research, recent earnings performance, and recent 10‐K and 10‐Q reports. Another notable finding is that issuing earnings forecasts and stock recommendations that are well below the consensus often leads to an increase in analysts’ credibility with their investing clients. We conduct cross‐sectional analyses that highlight the impact of analyst and brokerage characteristics on analysts’ inputs and incentives. Our findings are relevant to investors, managers, analysts, and academic researchers.