Want to know why so many minimum viable products stop at one-hit wonders rather than going on to be hit-making machines? It's all about having the right mindset and achieving scale by deploying the appropriate team structure for your time and place on the project spectrum.

1. Cognizant 20-20 Insights
February 2021
Bridging the Chasm: From Data
Science MVP to Impact at Scale
Want to know why so many minimum viable products stop at one-hit wonders
rather than going on to be hit-making machines? It’s all about having the right
mindset and achieving scale by deploying the appropriate team structure for
your time and place on the project spectrum.
Executive Summary
Machine learning, data science, predictive modelling, AI,
data and analytics are all hot topics right now. But as often
as you hear the phrase “data is the new oil” you’ll also
hear data science (disparagingly) referred to as a “cottage
industry”. For every story about a “game-changing
innovation” there are many more untold stories about
projects that didn’t quite meet their end goal or had some
kind of fatal flaw that prevented them from being fully
adopted by the business. And many of the projects that
reach production are not sustainable or cannot be scaled.
What does industrialisation mean? It means achieving the
same outcome (e.g.,increasing sales,optimising production
or predicting failures) across potentially very different parts
of a business (i.e., different divisions or different countries).
For example, Project 1 (typically a minimally viable project,
or MVP) is focused on optimising production in a particular
plant. To replicate the same process and outcomes at a
second plant,we launch Project 2. And then to rapidly scale
to additional plants, Project 3, Project 4, etc. are undertaken.

2. 2 / Bridging the Chasm: From Data Science MVP to Impact at Scale
Cognizant 20-20 Insights
How do we do this so that scalability is not only
achieved but returns results? How do we move
beyond executing everything as a single project to
a robust and replicable process? Project 1 might
take 10 weeks but we want Project 3 and beyond to
deliver value after only one or two weeks.
This white paper addresses why so many successful
MVPs fail to bridge the enormous chasm between
one-hit wonder and hit-making machine. It then
offers recommendations on actions and structures
that increase the odds of success.
MVPs for data science
Let’s start by exploring the mind of a data scientist.
Generally, data scientists are smart, creative
people who enjoy problem-solving. And solving
big problems is easier than solving little problems.
Indeed, data science projects often revolve around
large-scale complex challenges – with the relentless
hunt for ever-increasing accuracy a major motivating
driver (see Figure 1).
This is not necessarily a bad thing. Throw in
the capacity for a lot of hard work in a small
amount of time and top it off with great two-way
communication with the stakeholders and you’ve
The returns on data science: A balancing act
Many stakeholders believe that increased accuracy increases value, but value often occurs at lower-
than-expected levels of accuracy; highly accurate black box models can often impede adoption, thus
reducing value.
Figure 1
Source: Cognizant
BELIEF
REALITY
VALUE
ACCURACY

3. 3 / Bridging the Chasm: From Data Science MVP to Impact at Scale
got an excellent recipe for a successful MVP. But
how do MVP projects really run? The best word is
“nimble” (to avoid confusion with “Agile”). A lot of
interesting data science is essentially unknowable in
advance: Is my data suitable for modelling? What
kinds of models can I, should I and do I have time
to try? How do I decide among them? A successful
data science project may change directions a couple
of times a day.
This also means that many data science projects are
essentially bespoke. If you peel back the covers on
a typical data science MVP, it is often thousands of
lines of investigational code obscuring the couple
of hundred lines required to actually produce the
solution. Compounding this is the fact that whilst
data scientists are often excellent programmers, they
are often poor developers (and may not even know
what it takes to be a great developer).
So the successful MVP that delivers enormous
value and wins awards is likely run on code that was
developed late at night to solve a particular problem
and was never really meant for anyone else.
And this is the chasm. How do you take the spirit of
the MVP and scale it up into something repeatable
and industrialisable?
The mindset for success
Every journey starts with a single step but it is critical that for each and every step you have the right mindset
across the team. Here’s a breakdown of the key categories:
Worker mindset
❙ I will build this like I’m going to be forced to run the process
every week for the rest of the year.
❙ I will build this like someone else is going to run this every
week and ask me questions.
Designer mindset
❙ How will future problems
be different from the
current problem?
❙ How much customisation
to the current solution will
be required?
❙ How do we streamline the
process – is there a simpler
approach?
❙ How do I maintain a
coherent vision with other
projects?
Management mindset
❙ Project 1: Deliver high-
quality, impactful results
without overly worrying
about profitability.
❙ Project 2: Do it the way it
should have been done
the first time. Break even
on cost.
❙ Project 3: Deliver at
scale. Recoup all previous
investments.
❙ Project 4 and beyond:
Transition to the run team.
The successful MVP that delivers enormous value and wins awards
is likely run on code that was developed late at night to solve a
particular problem and was never really meant for anyone else.

4. Cognizant 20-20 Insights
The interaction and balance of these mindsets
is critical. Management needs to recognise that
scalability may not happen instantly but that they do
need to push for this in the second or third project.
Workers need to focus on solving the problem at
hand but also have one eye on the future so that the
current solution is not too specific. The designer
role can be the trickiest, as it requires having and
enforcing a vision for the entire program of work,
while being flexible enough to adapt.
MVP teams are often small, which can cause
problems. In a small team, everyone wears
multiple hats but it is very hard to simultaneously
maintain a worker + designer + management
mindset. A structure that adds more people and
compresses the timeframes can actually help.
Additionally, having oversight from the designer
and management mindsets to course-correct on a
periodic basis is critical.
4 / Bridging the Chasm: From Data Science MVP to Impact at Scale
MVP teams are often small, which can cause problems.
In a small team, everyone wears multiple hats but it is
very hard to simultaneously maintain a worker + designer +
management mindset.

5. 5 / Bridging the Chasm: From Data Science MVP to Impact at Scale
Cognizant 20-20 Insights
Structure for the journey
One thing to realise upfront: This is a journey, and
different teams are required at different times. In the
management mindset above, Project 1 is the MVP.
This is the project where you just get it done. Success
of the project is critical (see Figure 2).
Project 2 is the first attempt to scale – taking the
project and applying it in a different context. Often
Project 2 requires throwing away all of the previous
code and starting again. Not because the MVP
“wasn’t right” but because those projects are nimble
and once you reach your destination you realise that
there was a much more straightforward way to get
there.
Project 2 is about taking the ultimate ideas from
the MVP and making sure they are implemented in
a more straightforward way. The MVP team is still
Teaming along the value journey
Each project delivers similar value but the improvement comes from the increasing
efficiency that is gained by switching to the right team at the appropriate time.
Figure 2
Source: Cognizant
VALUE
EFFORT
MVP Team
Dev Team
Run Team
P
R
O
J
E
C
T
3
P
R
O
J
E
C
T
2
P
R
O
J
E
C
T
1

6. 6 / Bridging the Chasm: From Data Science MVP to Impact at Scale
Cognizant 20-20 Insights
the right team to do this: They are the most familiar
with the initial project and best placed to determine
what they should have done the first time. There is
no shame in this; it is just a function of the nimble
nature of data science.“Obvious” solutions are really
only obvious once you know the solution.
Project 3 is the real gap. This is where you have
to switch from solving the big problems to the
little ones. This is not nearly as much fun; you
might double your code base just to deal with
“edge conditions”. This is the place where a strong
development ethos really shines. If you can succeed
here then the transition to the run team and
execution at scale is significantly easier.
One way to structure the teams is with a core
team – or centre of excellence (COE) – that focuses
on MVPs. This core team lives within the digital
development team and transitions work to the wider
team. The core team needs:
❙ Very strong relationships with the business SMEs.
❙ The ability to rapidly deliver real results.
The wider digital team must:
❙ Be able to generalise the models/approaches to
different contexts.
❙ Write production-ready code.
One way to structure the teams is
with a core team – or centre of excellence –
that focuses on MVPs. This core team lives
within the digital development team and
transitions work to the wider team.

7. 7 / Bridging the Chasm: From Data Science MVP to Impact at Scale
Cognizant 20-20 Insights
Looking forward: No shortcuts
One final cautionary point: Whilst many projects
stall at Project 1 or Project 2 because they can’t
bridge the mindset gap, those stages are important.
Most attempts to jump straight to Project 3 often
fail to deliver any value. The first two projects are
an important part of the learning process. Without
them, the final solution often is overdesigned, is
way over budget and doesn’t meet anybody’s
requirement. Even if you have to throw away
everything, those initial projects add significant value,
both in their deliverables and to the ultimate quality
of the programme of work.
The three key points
Let your data scientists roam free and create, but make sure you have the team and the structure to guide
them on the path to success:
❙ Begin with a repeatability mindset, but do not enforce repeatability from the beginning.
❙ Value is the ultimate goal, not accuracy.
❙ The MVP team is unlikely to be the team that successfully scales the solution.

8. 8 / Bridging the Chasm: From Data Science MVP to Impact at Scale
Cognizant 20-20 Insights
About the author
Michael Camarri
Head, Cognizant Data Science Team, APAC
Michael Camarri leads Cognizant’s Data Science Team across the APAC region and has 22 years of
experience consulting in data science (a.k.a., decision science). At marketRx (which Cognizant acquired
in 2007), Michael developed and designed the mathematics, consulting processes and tools that enabled
the company to grow from six employees to over 500 in seven years. Michael has a number of global
roles at Cognizant, focusing on building repeatable analytical processes for major global corporations
across a wide range of industries. He is currently based in Melbourne and has a PhD in statistics from the
University of California, Berkeley, and an honours degree in pure mathematics, statistics and computer
science from the University of Western Australia. He can be reached at Michael.Camarri@cognizant.com
| www.linkedin.com/in/michael-camarri/.

9. World Headquarters
300 Frank W. Burr Blvd., Suite 600
Teaneck, NJ 07666 USA
Phone: +1 201 801 0233
Fax: +1 201 801 0243
Toll Free: +1 888 937 3277
European Headquarters
1 Kingdom Street
Paddington Central
London W2 6BD England
Phone: +44 (0) 20 7297 7600
Fax: +44 (0) 20 7121 0102
India Operations Headquarters
#5/535 Old Mahabalipuram Road
Okkiyam Pettai, Thoraipakkam
Chennai, 600 096 India
Phone: +91 (0) 44 4209 6000
Fax: +91 (0) 44 4209 6060
APAC Headquarters
1 Changi Business Park Crescent,
Plaza 8@CBP # 07-04/05/06,
Tower A, Singapore 486025
Phone: + 65 6812 4051
Fax: + 65 6324 4051
About Cognizant’s Artificial Intelligence Practice
As part of Cognizant Digital Business, Cognizant’s Artificial Intelligence Practice provides advanced data collection and management expertise, as well as
artificial intelligence and analytics capabilities that help clients create highly personalized digital experiences, products and services at every touchpoint
of the customer journey. Our AI solutions glean insights from data to inform decision-making, improve operations efficiencies and reduce costs. We apply
Evolutionary AI, Conversational AI and decision support solutions built on machine learning, deep learning and advanced analytics techniques to help
ourclients optimize their business/IT strategy, identify new growth areas and outperform the competition. To learn more, visit us at cognizant.com/ai.
About Cognizant
Cognizant (Nasdaq-100: CTSH) is one of the world’s leading professional services companies, transforming clients’ business, operating and technology
models for the digital era. Our unique industry-based, consultative approach helps clients envision, build and run more innovative and efficient businesses.
Headquartered in the U.S., Cognizant is ranked 194 on the Fortune 500 and is consistently listed among the most admired companies in the world. Learn
how Cognizant helps clients lead with digital at www.cognizant.com or follow us @Cognizant.
© Copyright 2021, Cognizant. All rights reserved. No part of this document may be reproduced, stored in a retrieval system, transmitted in any form or by any means, electronic, mechanical,
photocopying, recording, or otherwise, without the express written permission from Cognizant. The information contained herein is subject to change without notice. All other trademarks mentioned
herein are the property of their respective owners.
Codex 6228