A Playbook for Solo & Siloed Data Science Practitioners

Lauren Burke presented this material at the Columbus Data and Analytics Wednesday meetup on July 17, 2024

1. A Playbook for Solo & Siloed
Data Science Practioners
Lauren Burke-McCarthy
Senior Data Science Lead & AI Strategist
Further

2. About me
•Senior Data Science Lead & AI Strategist at Further
•Head of Community, Women in Analytics
•Instructor, Business Analytics, Denison Edge (Denison
University)
•7+ years in data science
https://laurburke.github.io
2Lauren Burke-McCarthy | laurburke.github.io

3. Who is this talk for?
The solo, siloed or initial data
scientist at a company
Stakeholders working with a
data scientist or data science
team
3Lauren Burke-McCarthy | laurburke.github.io

4. First
Data science is entirely
new to the organization.
And you’re building it
from the ground up!
Siloed
Data science exists in
other areasin the
organization.
But you are the only
one in your area
(embedded).
Solo
Your organization is
familiar withdata
science, previously had
other data scientists.
But you are currently
the only data scientist.
What this role could look like
Lauren Burke-McCarthy | laurburke.github.io4

5. This kind of role comes with challenges
Lauren Burke-McCarthy | laurburke.github.io5
High or
unrealistic
expectations
Lack of trust,
fear of the
unknown
Responsibility
to “own”
data science

6. …and opportunities
6Lauren Burke-McCarthy | laurburke.github.io
Set expectations for what data
science can and can’t doSet yourself up as the expert
Position yourself to inform
strategy Flexibility to explore options

7. PART 1
Set Your Expectations
7

8. Lauren Burke-McCarthy | laurburke.github.io8

9. What is a data scientist’s job?
Lauren Burke-McCarthy | laurburke.github.io9

10. But, what is your job?
•Formulate the right problems
•Clean and prep the data
•Build models and solutions
•Productionalize those models and solutions
•Communicate the results and impact with your
organization
10Lauren Burke-McCarthy | laurburke.github.io
data sciencev

11. What should already exist
•Defined processes and
tools for data collection,
storage, and governance
•Accessible, high quality,
suitable volume of data
Solid Infrastructure
•Leaders understand
core concepts of
analytics and data
science
•Data and analytics are
incorporated into
strategic plans
•Established use of:
•Descriptive/ diagnostic
analytics
•Reporting, dashboards
•Trust in data to inform
decisions over opinions
Data CultureData plays a role in
business strategy
11Lauren Burke-McCarthy | laurburke.github.io

12. Understanding your role
12Lauren Burke-McCarthy | laurburke.github.io
What are you
trying to achieve?
Organizational vision
for data science?
Where does the
data live?
Is the data ready for
data science?
Who’ll make decisions
from the results?
Who are your
“champions?”

13. MarketingFinanceManufacturingSales
OperationsHRSupply ChainProduct Customer
Support
IT
What will you support?
13Lauren Burke-McCarthy | laurburke.github.io

14. Mission
Statement
What you
want to
achieve
What areas
you’ll support
Who you’ll
work with
What impact
to expect
Who it will
benefit
Who will be
able to make
decisions
Define your public-facing goals
14Lauren Burke-McCarthy | laurburke.github.io

15. Why do I need something shareable?
•Investing in data science will mean doing things differently
•For this, you will need stakeholders and leaders on board
•To succeed, you’ll need to do this together
Investment is always easier to support when it comes
with defined, understandable goals and impact.
15Lauren Burke-McCarthy | laurburke.github.io

16. PART 2
Set Their Expectations
16

17. Starting and staying aligned
17Lauren Burke-McCarthy | laurburke.github.io
A shared vocabulary
maintains consistency
Align on a common set
of definitions
Always establish a
single source of truth

18. Set their expectations
18Lauren Burke-McCarthy | laurburke.github.io
Deliverables
What to expect & when
Align on next steps and
deliverables
Process
Demystify longer “time
to value” between
analytics and data
science projects
Goals
Establish a mutual
willingness to
reframe/reset if
necessary
...for the data science process itself

19. •A magic one-size-fits-all solution
•An immediate ROI on investment
•Always fast to implement
•Always successful
•Always necessary
What data science isn’t
19Lauren Burke-McCarthy | laurburke.github.io

20. Take a “road trip”
•Connect with folks in and
adjacent to your space
•Then ask them who else
should be involved
Strategies for engagement
Lauren Burke-McCarthy | laurburke.github.io20
•What are the problems?
•Who else knows about them?
•What value would solving them
provide?
•Who else needs to be involved to do
this right?

21. Share info along the way
•Data science vs. analytics
•Your goals, experience and the
value you’ll provide
•Example use cases
Strategies for engagement
21Lauren Burke-McCarthy | laurburke.github.io
Take a “road trip”
•Connect with folks in and
adjacent to your space
•Then ask them who else
should be involved

22. •What are their goals for the
next 6 months/ year?
•Any ongoing concerns?
•Any regular requests?
•What would you improve?
Your main goal: figure out the what/why
22Lauren Burke-McCarthy | laurburke.github.io
For suggested use cases:
Why and how would
this be valuable to
the org or users?

23. PART 3
Identify Opportunities
23

24. So, what projects do
we invest our time in?
Use cases should:
•Always be impact driven
•Be clear, achievable and
actionable
Avoid data science for
data science sake.USE CASES :)

25. “A human-centered approach to innovation that draws from the
designer’s toolkit to integrate theneedsof people, the possibilities
of technology, and the requirementsfor business success.”
Tim Brown
CEO, IDEO
Consider design-thinking principles
EmpathizeDefine TestIdeatePrototype
25Lauren Burke-McCarthy | laurburke.github.io

26. Why design-thinking applies
Results-orientedCustomer-focused
or user-centric
Iterative
methodology
InterdisciplinaryEmpatheticData-driven
26Lauren Burke-McCarthy | laurburke.github.io

27. •Figuring out the real problem we are trying to solve
•Understanding if we even need a model for the problem
•Making sure that what’s in our stakeholder’s head is
what we are going to provide
•Being willing to change direction and providing data to
support your reasoning
•Knowing how the solution will impact the users/business
What this looks like in data science
27Lauren Burke-McCarthy | laurburke.github.io

28. Start with quick wins, then think longer term
28Lauren Burke-McCarthy | laurburke.github.io
•Speed up or automate manual
processes
•Proactively alert errors by
introducing thresholds
How are they currently being made? By
who?
•Can you put data behind recurring
opinion-based decisions?
•Find opportunities to turn reactive
decisions into proactive ones
Operational EfficienciesInform Decisions

29. 29Lauren Burke-McCarthy | laurburke.github.io
Start with quick wins, then think longer term
•Biggest opportunities or
gaps in your industry or
organization
•How can you provide
unique value to clients
and users?
•Focus on repeatable,
reproducible use cases
over one-offs
•The timeline to the MVP
(minimal viable product)
Accelerate/DifferentiateAlways Consider

30. Also note:
•Perspectives will vary, everyone
may have different ideas on how
to achieve these goals
•You can push back and
encourage the reframing of a
problem if necessary
Use case building blocks
30Lauren Burke-McCarthy | laurburke.github.io

31. Expected Actions
What processes or
decisions will be
improved, updated or
done differently?
Expected Impact
Who will benefit from
this solution?
How will the business
overall benefit?
Expected
Supporting Data
What data will be used?
What data will be able to
prove the value?
Align proposed solutions to specific business outcomes.
Always define and align the value
31Lauren Burke-McCarthy | laurburke.github.io

32. Data science is kind of like improv…
Sometimes we say,
“Yes, and…”
And other times,
“No, but…”
32Lauren Burke-McCarthy | laurburke.github.io

33. •Expect to face questions you won’t
know the answer to
•Those with company, industry, or
user domain knowledge can help
provide context
•Data scientists aren’t the only ones
who make data science successful
•Rely on your partners in product,
operations, UX, engineering and
more to keep things moving
Subject Matter ExpertsStakeholders = Partners
Know where to find support
33Lauren Burke-McCarthy | laurburke.github.io

34. PART 4
Communicate and
Document
34

35. •Demonstrate value in status updates
•Find opportunities to deliver value incrementally
•Share impact > methodology
Value-first approach to communication
35Lauren Burke-McCarthy | laurburke.github.io

36. Without proper documentation…
36Lauren Burke-McCarthy | laurburke.github.io

37. •Establishes patterns
•Sets stakeholder expectations
•Increases familiarity with data
science process
Always strive for reproducibility
Builds trustEnhances productivity
•Similar projects get off the
ground faster
•Reduces future time to value
37Lauren Burke-McCarthy | laurburke.github.io

38. •Useor understand something
•Awarenessof changes or updates
•Receive outputs or reports
•Make decisions based on results
•Shareor explainkey information to
others
Who is the audience? What do they need/want?
Context
Overlap
Context
Overlap
How we think user-focused
38Lauren Burke-McCarthy | laurburke.github.io
Technical users
Stakeholders
End users

39. Make your data science presence known
Store documentation where:
•It can be easily accessed
•It is shareable and searchable
•Updates and changes can be tracked
•Stakeholders have a place to check-in
Increase visibility of data science in your org!
39Lauren Burke-McCarthy | laurburke.github.io

40. •Create a copyable template to share project details, track
status, tag team members and keep notes
•Link repos, decks, meeting recordings, external / internal
resources that add context
•Incorporate keywords for search visibility
•Keep note of project name, plus any and all nicknames
Use “SEO” to further enhance visibility
40Lauren Burke-McCarthy | laurburke.github.io

41. •Define goals that help you stay on track
•Start and stay aligned with your stakeholders
•Set expectations early
•Focus on the right problems
•Reframe problems together if necessary
•Communicate early, often and openly
Key takeaways
41Lauren Burke-McCarthy | laurburke.github.io

42. Thank you!
https://laurburke.github.io
linkedin.com/in/lauren-e-burke/