More companies have embraced citizen data science and have deployed tools
like Tableau, Microsoft Power BI, Qlik, Domo, KNIME, Sisense, Google Data
Studio, Amazon QuickSite, and other self-service BI to business teams.
It’s common to see these tools deployed to individuals scattered across the
organization, and each focused on their department’s business objectives.
For example, digital marketers are likely to analyze campaign performance,
operations managers review supply chain risks, and financial analysts
evaluate profitability.
Other times, a centralized group takes on dashboarding, data visualization,
and lightweight data science work as a service. Business leaders expect this
team to maintain reporting tools and respond to special requests.
To centralize or decentralize data scientists?
While centralized citizen data science groups are more likely to drive the
creation and support of best practices, it can also be a barrier for
organizations looking to become more data-driven. Having a centralized data
visualization or reporting team is effectively a punt – allowing business
teams to stay an arm’s distance from analytics work. They can sometimes
formulate unrealistic requests to the centralized data science team and are
likely still using complex spreadsheets for their most strategic analysis.
As a CIO, I’d rather push the citizen data science tools to business
departments where they are more likely to ask the right questions,
understand the underlying data, and translate insights into faster
data-driven decision-making. I seek data owners who will improve data
quality and find innovative ways to use analytics to grow business or drive
efficiencies.
But I’m also worried about creating a new generation of data governance
issues, where citizen data science work goes unchecked, and business users
create an unmanageable number of complex dashboards and data visualizations.
I’m concerned about duplicate reports, outdated dashboards, and data
visualizations with no one supporting them – all examples of
data debt.
Avoid centralization by establishing a center of excellence
My answer to this conundrum is to create
citizen data science centers of excellence (CoE)
that promote best practices, drive skills development, and establish
self-organizing standards. While they may focus on data visualization, their
responsibilities extend to dataops and data governance.
You can read some of my previous posts on these topics, including these
three top articles:
Also, please read some of my stories on being buried in bad data, chapter 7
of
Digital Trailblazer, and some of my
best data-driven organization best practices in chapter 5 of
Driving Digital.
Start developing a citizen data science CoE
.jpg "StarCIO Centers of Excellence")
Here are a few things to focus on in developing a citizen data science
CoE:
-
Drive collaboration among practitioners – Identify tools and
communication mechanisms for knowledge sharing. For example, the CoE can
use an Atlassian Confluence Space to document best practices, share
reusable data visualizations, or ask questions. -
Take ownership of data catalogs and dictionaries – When a citizen
data scientist adds a new data source, creates a new calculated measure,
or builds a rule to create dimensions, are these documented in the data
catalog? CoEs should have checklists and establish naming conventions for
managing derived data and calculations. -
Establish release standards – When is a data visualization ready to
be released to end-users and decision-makers? What are the quality
standards, and what testing is required? Self-service BI tools often
support real-time editing, which is powerful when developing dashboards
but makes it easy to deploy costly errors. What’s your organization’s
definition of done? -
Incentivize internal training – Requiring a citizen data scientist
to take a tool’s basic online training is an important first step but
insufficient training for working smart and safely with enterprise data
sources. A CoE should set up business-specific training that includes labs
with proprietary data sources and covers data governance practices.
Leadership should incentivize CoE members to develop course materials and
teach classes, while employees should be required to take the training. -
Celebrate and market success – The best way to grow support and
develop a data-driven culture is by showcasing and marketing the CoE and
analytics programs’ successes. Empowerment is contagious; the more leaders
celebrate wins, the more people seek involvement.
When it comes to data and analytics, having a few go-to people or a
centralized data science department doesn’t go far enough to drive a
data-driven culture.
Reach out to me
if you have questions or want me to write more about data-driven centers of
excellence.
