ArticlePDF Available

Object KPIs for the digital transformation

Authors:

Abstract

KPIs (key performance indicators) are currently widely used in the industries at management level and in the toolkit of the consulting companies. However, they are interpreted by humans, and humans act on the results based on the experience of an individual. What is good,bad or underperforming is determined by fixed setpoints based on recognised industry benchmarks. Dynamic setpoints that are based on individual company or market circumstances are not common or even unheard of. KPIs are not automatically fed back into the control cycle of managing a company or an operational plant by a computerized business model. In general, they are high level in nature and do not go down to the nucleolus of the production process and operating plant equipment. Therefore, simplifications and reduction of data are necessary to make it manageable for decision-makers. However, in the time of cloud computing, deep learning, and AI science, it is possible to analyse the performance of infinitly small parts/equipment of a processing plant. The resulting data can be amalgamated from the bottom up to give precise results, the possibility to act instantaneously and the ability to identify the root cause of any issues. This article intends to offer potential solutions to how KPIs can be utilized for the digital transformation of any industryfor improving processes and business opportunities.
Object KPIs for the digital transformation
Jordan, Frank, [Lecturer] PhD.-Student at Comenius University, Bratislava)
Abstract
KPIs (key performance indicators) are currently widely used in the industries at management level and
in the toolkit of the consulting companies. However, they are interpreted by humans, and humans act
on the results based on the experience of an individual. What is good,bad or underperforming is
determined by fixed setpoints based on recognised industry benchmarks. Dynamic setpoints that are
based on individual company or market circumstances are not common or even unheard of.
KPIs are not automatically fed back into the control cycle of managing a company or an operational
plant by a computerized business model. In general, they are high level in nature and do not go down
to the nucleolus of the production process and operating plant equipment. Therefore, simplifications and
reduction of data are necessary to make it manageable for decision-makers.
However, in the time of cloud computing, deep learning, and AI science, it is possible to analyse the
performance of infinitly small parts/equipment of a processing plant. The resulting data can be
amalgamated from the bottom up to give precise results, the possibility to act instantaneously and the
ability to identify the root cause of any issues.
This article intends to offer potential solutions to how KPIs can be utilized for the digital transformation
of any industryfor improving processes and business opportunities.
Key Words:
KPI, Object KPI, digitalization, digital transformation, deep learning, AI science,
dynamic KPI set points, automate feedback, adaptive KPI, automatic process KPI
generation, KPI tracking, and KPI evaluation.
Introduction
Key performance indicators (KPIs) or scorecards have been used since the 1950s and
enjoy great popularity in all industries today. Most consulting companies are using KPIs
as their base toolkit to advise clients on how to monitor, track and improve their
business. It has become a culture to expose working processes and company success
rates to employees, and the walls of office entrances are plastered with scorecards
and KPI charts.
However, an active self-automated optimization process utilising those KPIs has not
yet been implemented. Instead, it depends mainly on the goodwill of the people and
the interpretation of the managers to improve a process. KPIs are abstract and do not
give any direction on what the employees or managers should do to improve them.
Furthermore, they do not indicate where the root cause of the problem is.[7,8]
For instance, if one KPI indicates a negative trend, this trend can be understood by
everyone. However, the root cause and the corrective actions required are not
apparent and depend on the interpretation by the individual. Generally it takes a
significant amount of time to realize that the trend is negative, and an even longer
period of time to initiate a corrective response.
Consulting companies can be hesitant to agree with this and could argue, they have
appropriate processes and methods in place to discover those trends in a timely
manner. However, this is again reliant on human-resources and very time-consuming.
The usage of their “tools” is binding resources and thus delaying the opportunity for
the management to take corrective action.
The objective of this article is to analyze how KPIs can be enhanced and used as part
of an automated (computerized) algorithm to interact with the business directly. The
definition of these indicators, whether it is a lead, lag, RI, PI, KRI, KPI, etc. [1,2,3], is
not relevant because this article will demonstrate how attributes can be attached to
indicators to give a machine a holistic view into the business process. Therefore, all
indicators become no key performance indicators and will be evaluated in real-time
and are part of a new reality.
Classic KPIs
Measurement initiatives are often coupled together without proper knowledge of the
organization's critical success factors and an understanding of the behavioral
consequences of certain measures.[9]
In the process industries, KPIs are associated with equipment packages.
As an example, we will investigate an instrument air system installed in a plant.
The instrument air package consists mainly of a compressor which produces
compressed air for a plant as auxiliary air to drive valves. Although the system is not
very complicated, it is critical to the business, as without it, there is no production.
Consequently, the availability should be more than 98%, and if it is an excellent
performer, it should be above 99%. In order to reach this availability, there is usually
a redundancy measure provided; in other words, a second compressor.
A plant can have hundreds of performance indicators. Therefore, it is beneficial to
give them self-explanatory names like IAS1 and IAS2 (instrument air system 1 and
2).
As timebase, this example uses one year which is equal to 8760 hours per year.
The KPI IAS1 will measure the time during which no alarm or fault has been activated
against the total time in operation, which is an indicator for the healthiness of the
system.
Upper limit 99% of hours per year equals 8672h
Lower limit 98% of hours per year equals 8584h
Slope = 0.011363636
Intercept= -97.5455
This KPI is considering the running hours, which can be counted by a SCADA system
and feed into the linear formula.
As a result, the compressor is reliable and performing well if its KPI is close to 99%
working hours per year, which corresponds to a KPI value of 100%.
If the running hours are closer to 98%, the compressor needs too much maintenance
and has a bad performance KPI which is 0%.
The KPI IAS2 will measure the operational scope of the compressor design. Under
normal conditions, one compressor can replenish the instrument air vessel and
maintain the system pressure. In a scenario where the second compressor is needed
more, it can be considered that the instrument air system is running out of the design
scope due to leakages or compressor problems.
IAS2 will measure the running time of compressor one, two or both against the
running time of both compressors.
𝑅𝑢𝑛𝑛𝑖𝑛𝑔 𝑟𝑎𝑡𝑖𝑜 𝑖𝑛 % = (𝑇(𝐶1 𝑜𝑟 𝐶2) 𝑇(𝐶1+ 𝐶2)) (𝑇(𝐶1 𝑜𝑟 𝐶2)
(C1,C2 = Compressor 1 / 2)
Therefore, the limits are defined as:
The upper limit is 90% of the time one compressor is sufficient
The lower limit is 100% of the time one compressor sufficient
Slope = -0.1
Intercept= 10
IAS 2 will be recorded after each trigger of the compressor until both compressors
are stopped.
As we can see with IAS1 and IAS2 we will get a KPI which is measurable and clearly
defined. However, how can we decide that the compressor needs to be replaced
because the maintenance cost is too much? Even if two compressors are similar over
the lifetime of a plant they will act differently.
We can decide based on a holistic, fact-based criteria wether we will let one
compressor run more than the other compressor in order to reduce the overall cost of
maintenance and increase the uptime of the plant.
The result is precisely measurable without any manipulation, but the final decision in
the context of the entire business is subject to personal interpretation, rather than
objectively considering the whole enterprise.
For a machine or an artificial intelligence, it is not feasible to make any conclusion if
there is no reference or rule available which determines what should be done in case
of good or bad performance.
The upper and lower limit is fixed and chosen based on benchmarks, that are used in
similar industries. There is no possibility to individually tailor the classic KPI based on
business needs, CAPEX (capital expenditure), OPEX (operational expenditure),
available spare parts, logistics and supplies, and available human resources.
Therefore, classic KPIs have a lesser value for the digital transformation of the
industries.
Object orientated KPIs
As we have shown, the classic KPIs have no attributes which makes it hard for a
computer system to evaluate, define or recommend strategies to improve the business.
There is no feedback loop to adapt to set targets or limits.
The goal is to have a KPI with associated attributes which can be evaluated for a
business model. The KPI attributes can be fed into a computer model which represents
the business of the company. The business model is purely a mathematical model
where all known factors or KPIs should be fed into. With rules and formulae compared
against a knowledge base, the output will guide the company on what would be the
course of action at a given time to prevent degradation of the business.
Figure 1: KPI controlled business model
As we can see (Figure 1), multiple parameters and attributes will be required to feed
into a business model to represent a company in detail from bottom up.
Computers should collect the data required in order to accureately represent the status
of the company.
For instance:
most manufacturing plants already have an asset management system in place,
but it depends on specialists to make the decision on what is essential.
We need attributes for each equipment for:
meantime between failure
spare part availability
maintainability
cost
meantime to repair
redundancy
process conditions
importance for the business
the assets have different values, but the KPI of the equipment does not take care
of the importance of this package.
We need attributes for each asset for:
priority for the business
value contributes to the company
maintenance schedule
turnarounds
It becomes evident that the number of data multiplied by all the equipment parameters
of a company are huge and cannot be tracked, and analysed by a human being.
It would be an immense effort to maintain constant repetitions and analytics manually
to predict anything out of this data.
Therefore the performance indicators are recommended to be reduced to 10 [9] in
order to make it manageable and be called key performance indicators.
However, with modern computer capabilities, it is not a problem to handle extensive
data. With cloud computing, the performance is almost unlimited and affordable.
Once the data is evaluated, the result could be fed into a business model to
directly steer:
the day to day operation
maintenance activities
repetitive and planned works
investments of the company
reporting focussed on hot spots
continuous improvements required
The key performance indicator becomes an object KPI and is measured in real-time.
Digital transformation
Now let us assume we have all data available in a digital format, what infrastructure
would we need to evaluate the data?
Figure 2: KPI data collection, evaluation by AI
Similiar to a human brain, the information needs to be available in one location which
is accessible for the artificial intelligence. (Figure 2)
The work can now be segregated between:
repetitive or known work done by computers
creative work done by humans such as:
o define what would be the ideal case or best practice
o rules of the business
o experience or restrictions
The artificial intelligence (AI) is searching for known events in the data lake. The AI is
comparing the data with the knowledge base.
Once there is a known scenario identified, it triggers an event with fixed rules and
measures to be executed.
Therefore, we can assume that the actual work will be executed:
best practice based on all known information
no violation of the rules
consistent quality of work, no mistakes, no human error
no complacency, no time delay
no health and safety compromise
full transparency
Based on the history and trends of the KPI attributes, they should be continuously
updated and refined to make better predictions and corrective actions. The AI will feed
back and modify attributes to enhance the efficiency of the business.
Object KPI for fully automated maintenance process
For a better understanding of the benefits, we shall demonstrate how the Object KPIs
are implemented into the maintenance strategy of an operating plant.
The IT infrastructure of a plant is broken down into several layers:
1. field layers consisting of all field instruments, process value, and diagnostics
2. control layer consisting of SCADA, HMI, and IT networks
3. asset management level (AMS) and maintenance management system (MMS)
The SCADA (supervisory control and data acquisition) or HMI (human-machine
interface) system provides the interface for all data or process values to humans to
act on and evaluate manually. If there is any corrective action required, it can be
manually entered into the MMS (maintenance management system), and the work-
order is generated to be executed. Depending on the understanding of the
maintenance planner and the breakdown of the maintenance structure the equipment
will be repaired.
How the plant should work and what to do in an emergency is described in an
operating manual. The main work of the operation staff is to monitor the process,
read alarms and find anomalies that are occurring. Once an abnormality is found, it
should be acted upon based on the operating manual or according to the experience
of the operators. This process is not free of human errors, and it is hard to expect
humans to prevent an accident. Due to cost-cutting and reduction of personnel, the
pressure of the individual has increased, which contributes to an increased risk of
severe accidents.
Due to disasters in the industries, humans are no longer considered as a safeguard
for a plant.[Shell DEP] Safeguards are a higher level of automation and protection,
but there are still operators involved.
For ideal operation, failures should be eliminated, and work should always be
executed in the same safe way and to the best benefit to the business.
With full automation,the majority of an operator’s involvement can be reduced.
However, in a failure scenario, the situation still depends on humans to decide.
The analytics and diagnostics in modern equipment is very advanced, and it sends
an error message with detailed information to the AMS about the status of the
machine. However, a human has to look at the condition and make a decision about
the necessary action.
In an ideal unmanned plant (no operators), the equipment which is going to fail,
triggers a process which decides on the criticality of the action to be taken:
1. the machine is so essential that it has to be replaced, repaired immediately
2. the repair can be done at the next planned maintenance schedule
3. the repair can be done at the next maintenance campaign
For improving the performance, the emergency case should be minimized because it
has a direct impact on the business and the company can only react to the effects.
High plant availability should be planned and maintained by a computer-supported
maintenance regime.
Usually, the maintenance schedules are fixed, and they do not consider how the
equipment has performed in between. Therefore, unnecessary maintenance work will
be done. With Object KPIs it can be decided according to which category the
equipment should be maintained. For instance, if a transmitter is stable working and
no anomalies are detected, it is unnecessary to perform maintenance on it. The risk
of causing a failure by disassembling and reinstalling it is higher than the risk of doing
nothing at all. There is a huge potential in cost-saving if the individual part’s
performance decides if maintenance is required.
Figure 3: Digitized work process for AI-controlled maintenance
The Object KPI of the individual instrument or equipment will include all parameters
which a computer system needs to know in order to understand what the importance
of this equipment for the business is. With its holistic view of the situation (access to
all data) it will automatically select the best solution which has been pre-determined
by experts for this particular equipment.
The reaction of the plant and the object KPI will be trended, so that the individual
object KPI can be continuously adjusted based on real-time data from the plant.
Consequently, there is no delay, no HSSE violation and no mistake with the
execution. The AI system will improve its knowledge and enhance its prediction
capability as well as the business opportunities with each event.
As a practical example, let us assume we have a pressure transmitter installed in a
plant from the same manufacturer and same type.
Depending on the location where the transmitter is installed, it might fail more
frequently than at another site.
The vendor will give all transmitters the same MTBF (mean time between failures),
disregarding any differences in temperature, vibration, and process conditions.
Therefore, if we want to predict the maintenance interval of the transmitter, we must
consider all these parameters and adjust the maintenance interval as per the
conditions of the individual transmitter. Any opportunity should be used to keep the
plant uptime high. Therefore, each shutdown or unplanned stop should be used for
maintenance.
With a flexible, individual equipment-oriented maintenance schedule determined by
the object KPIs of each equipment, the handling of the maintenance schedule will be
very complex. If the business is to benefit from predictive maintenance, a machine has
to make these calculations and decisions.
The AI system will learn from experience and improve the transmitter KPI attributes
accordingly as well as determine the individual maintenance schedule.
The acceptance of flexible maintenance schedules that are based on predictive
maintenance controlled by object KPIs requires a radical, new way of thinking.
Possible benefits
The shareholder will benefit from the new concept in the following ways:
1. Reduced OPEX
a. significant reduction of operators and experts over the lifetime of the plant
b. personnel is focussed on maintenance and assisted by an AI system
c. predictive maintenance is based on real-time data and not statistical
benchmarks. Therefore, it is best-practice for the specific plant and not
just for an average plant.
d. knowledge is machine-based and continuously available over the lifetime
of the plant
e. improved uptime of the plant’s business
2. Clear facts and priorities for the investment funnel.
3. Best practice for the plant with the lowest possible CAPEX.
4. Optimally adjusted business model to ensure the best profit.
5. Leading indicators which come directly out of the process ensure the fastest
possible reaction times.
Conclusion
The enhanced object KPI model, in combination with modern computer science, has
the potential to improve businesses significantly. It can be integrated into a fully
automated plant and, together with an online evaluated business model, steer OPEX,
CAPEX, investment funnel and reporting actively.
Consequently, object KPIs for individual equipment business elements are the
prerequisite for deep learning (machine learning) and thus required for artificial
intelligence to be integrated into a business process for decision making.
Object orientated KPIs are not fixed to a certain business, they can be applied to all
kinds of industries which are undergoing the evolution of digital transformation.
Companies that follow the way of the digital transformation can anticipate an
improvement of their business by more than 5-10%. It is a substantial improvement to
beat the market and survive the coming age of the digital transformation.
However, only companies that can utilize the new object orientated KPIs to their
maximum potential and integrate it into their business model will have success with
this technology. Naturally, experts and out of the box thinkers are also required to make
it successful..
Top management can focus more on the business und will be less involved in curing
symptoms. Based on hard facts gathered in real-time, management can act instantly
to the best of the current situation.
The consulting industry of the future will focus on how to teach the machines, drive the
plant or business instead of writing procedures for humans.
Bibliography/ References:
1. Carol Taylor Fitz-Gibbon (1990), "Performance indicators", BERA Dialogues (2),
ISBN 978-1-85359-092-4
2. Key Performance Indicators What Are Key Performance Indicators or KPI (Wiki)
3. Key Performance Indicators: Establishing the Metrics that Guide Success, accessed 23
April 2016
4. Palffy, Georgina. How Business Works (1st ed.). DK Publishing. p. 146. ISBN 978-1-
46542-979-7.
5. "Key Performance Indicators" (PDF). Colleges Ontario. Retrieved 2013-05-25.
6. Daddis, Gregory (June 1, 2011). No Sure Victory: Measuring U.S. Army Effectiveness
and Progress in the Vietnam War. ISBN 978-0-19974-687-3.
7. Robert D Austin, "Measuring and Managing Performance in Organizations"
8. Martin Fowler (2003-08-29). "Cannot Measure Productivity". Martinfowler.com.
Retrieved 2013-05-25.
9. DAVID PARMENTER Key Performance Indicators ISBN 978-1-119-01984-8
10. Bernard Marr, Key Performance indicators, ISBN-13: 978-0273750116
11. Yokogawa middle east, pictures of instruments, PLC picture
Contact: Frank Jordan, MSC. Electrical
[Lecturer/] PhD.-student at Comenius University Bratislava
Faculty of Management
Odbojárov 10,
820 05 Bratislava 25,
Slovak Republic
E-mail: FJ1808@web.de
Peer Reviewed by
prof. RNDr. Michal Greguš, PhD.,
Dean
Faculty of Management,
Comenius University in Bratislava,
Odbojárov 10,
820 05 Bratislava 25,
Slovak Republic
ResearchGate has not been able to resolve any citations for this publication.
Article
This study analyzes how the United States Army, particularly the Military Assistance Command, Vietnam (MACV), attempted to measure its progress and effectiveness while conducting counterinsurgency operations during the Vietnam War. In short, in a war without front lines, how did the army know if it was winning or losing? White House advisers, Pentagon officials, MACV staff officers, and army field commanders all faced immense challenges in identifying useful metrics for gauging success in an unconventional environment. Throughout the war, they often came to contradictory conclusions. Political, economic, and cultural factors influenced daily the course and conduct of the army's counterinsurgency operations. In such a complex environment, how did American officers and soldiers know whether or not they were making progress over the course of a decade-long war?
Measuring and Managing Performance in Organizations
  • Robert D Austin
Robert D Austin, "Measuring and Managing Performance in Organizations"
  • Martin Fowler
Martin Fowler (2003-08-29). "Cannot Measure Productivity". Martinfowler.com. Retrieved 2013-05-25.