Building Innovative Product, DataScience and Beyond.
What is innovation and what is it not ? How to approach it so you succeed ? What about DataScience Projects ?
2. What’s Innovation ?
1. Problem you care about
2. Idea Generation
3. Idea Validation
4. High Speed and
Low Cost of Iteration
and enough runway
5. Empowered People who love
what they do and care about the problem
1.1 Closeness
with User
3. What’s the outcome ?
6. Learning What Other’s Don’t
Know (Problem and Solution)
7. Successful Idea to Scale
4. Detour - What
Innovation is NOT
Keep the innovation map we saw in mind. We will
come back to it shortly.
6. Neither is it careful
planning
You don’t know what result of your experiment would be.
You don’t know the learning you would have after next experiment.
7. It’s not the
particular
technology
Start with the problem you care about first. Figure out the right tech to apply.
Never the other way around ! You can’t “experiment” with outcome in mind already.
Also, If you don’t care about the problem, you would not survive the initial failures.
8. Not Enough Runway
Before innovation flies, it crawls, walks and runs. Too short a runway and no idea will fly.
9. Not a work
for large team
You need small, self sufficient, fast moving team. Large team is only needed to scale the
final solution.
10. Why Bother ?
Innovation is tough, why bother ?
Execution is fast.
In shorter run the better
executing team wins.
Exploration is slow and
wasteful. In long run those
who explore overtake
So you need both
12. 1. Problem
You care about
• Innovation is a winding path
• People put more effort for
things they care about
• How to enable this ? We’ll see
13. 2. Idea Generation
For idea generation you want to do following
• Improve quality of ideas
• Improve quantity of ideas
14. Idea Quality & Quantity
•Be cloooose to
users
• Ideally - be your own target
demographic
• E.g. if an investment product, be user of
that investment product to invest
• Entire team should feel as
close to users as possible
• Otherwise, you solve the
wrong problem or fail to
learn
15. Idea Quality and Quantity
•Explore
• Experiment a lot (cheaply).
Each experiment improves
your next idea quality.
• Have a long term process,
with progress measured by
learning done, not success
of every step (don’t
measure innovation in
similar way as execution/
delivery projects)
16. Idea Quality and Quantity
• Be aware of and leverage
Open Source
• Don’t reinvent the wheel,
leverage open source to
solve common problems
well, contribute back to it
as well
• Be aware of what is
possible with bleeding
edge open source
17. 3. Idea Validation
• Never drive blind
Build analytics into your product
so you know how well is your
product doing? how is it being
used? Did your last improvement
improve retention or referral ? Did
the last idea deployed become
popular ?
• Talk to users who used your new
idea
• Take your product/new ideas for
test drive yourself on a new road.
Often you will notice any pain
more directly.
18. 4. Rapid Cheap
Experimentation
• If each experiment takes
months or hundred thousand
dollars to conduct , you would
not experiment much
• Invest in building infrastructure
to enable rapid
experimentation using devops
and any tool to that effect
19. 5. Empowered People
Who love their work and care about the problem
• People are very important component of
innovation
• Hire people who just love what they do.
People who do work just to pay bills never
build ground breaking products
• Empower them, trust them. Use agile
approaches where you don’t stand behind
the shoulders or try to control people.
• Agile daily stands should be just 5-10 min
of what each person is doing today and
not a question and answer or control
system. Note this requires people who
love what they do.
• Teams should be small (5-10) and self
sufficient
20. Distributed Side Projects
• Using Devops and Other Automation Tool
Make it easy for any team to experiment
in company as a cheap side project
• Keep cost of side projects low
• Allow anyone to start one with minor
approval (not require higher
management involvement)
• Snowballs into lot of innovation
• Every experiment succeeds (if learning
is the target)
• Some more financially successful
• Other improve quality of ideas in
future
22. All previous still
applies
Problem You Really Care About
Empowered People Who Love Work & Care about The
Problem
Being Close to User
Generating Ideas -> Validating Ideas -> Quick Iterations
Learning and Sufficient Runway
23. Adopting Latest Innovation Research In House
Using Open Source Models,
Libraries & Platform
Early on concentrate on
learning from each small
project, not success or
inevitable failures
As you figure things out,
scale and execute
You must do a bit of this to
develop appreciation of
nuances & solve problems
not yet solved
Identify core pain of users/
customers and focus on
each of those with research
too (longer runway)
Small self sufficient and multi
disciplinary (DS, DE, API, UI,
Devops etc) teams
Small self sufficient and multi
disciplinary (DS, DE, API, UI,
Devops etc) teams
24. 2. Idea Generation
3. Idea Validation
4. High Speed and
Low Cost of Iteration
and enough runway
Get This Right
Tools to speed up experimentations
25. Start with simple solutions
Start with simple rules
Once you have lot of data, introduce ML
Replace any complex rules with ML
26. Open Data
Wikipedia, Wikidata,
Twitter etc
Apps and API
Your own UI, API etc
would provide data
DS Strengthens With Data
Data Augmentation
Use Sensible
Augmentation of
Available Data
28. Why Bother ?
Innovation is tough, why bother ?
Execution is fast.
In shorter run the better
executing team wins.
Exploration is slow and
wasteful. In long run those
who explore overtake
So you need both