“The first decision in decision making is when to make a decision,”
proclaimed Mike Hayes, the chief digital transformation officer of
VMWare and former
commanding officer for SEAL Team. I was attending
Rev 3, a
conference on ModelOps, and, more generally, a conference for data
scientists to learn and network about their practices in experimenting and
delivering machine learning models that drive business impact.
It was at The Marriott Marquis on Broadway in NYC, and the ballroom was
filled with really smart, mostly young, and very eager aspiring
Digital Trailblazers.
But for me, I was actually following another one of Hayes’ recommendations,
“Slow down and carve out that think time.” Yes sir.
At the conference and long before it, I have been thinking about how more
people can improve their data literacy and, more specifically, become more
hands-on working in
citizen data science programs. I learned some recommendations at the conference and in speaking with
experts.
1. Learn from SMEs, ask questions, challenge the status quo
Two more great pieces of advice came straight during the opening session.
Hayes recommended to the audience, “What matters is who is closest to the
problem and suited based on experience to solve it.” He was referring to
where to enable decision-making and sharing his best practice – people who
understand the problem, risks, and ramifications are the ones who should see
the data to help guide decisions. In Hayes’ case, that may be pulling the
trigger on a special ops.
Later in the opening session, Linda Avery, chief data & analytics
officer at Verizon,
shared how they challenged their core operationally-minded culture. She
said, “With an execution-oriented operation comes a very directive culture.
We have to move away from that to a culture where it’s ok to question.”
This advice applies even more to aspiring data leaders, and it reminds me of
a post I wrote nine years ago on
asking smart big data questions. That post came after a town hall where several business leaders showed
off their data visualizations and other data tools to analyze customer and
operational data. That was my first citizen data science group, and their
journey didn’t start with learning new tools. It started by being
inquisitive and learning about what data was available to make more informed
decisions.
2. Bring your questions and data to a learning program that works for you
Once you have a question and some data, you are better equipped to learn
data practices and tools. It’s ok to start with the examples from tools,
books, and courses, but their exercises often gloss over the realities of
working through data quality and sourcing issues.
Speaking with Rosaria Silipo, principal data scientist and head of
evangelism at
KNIME, suggests
these five ways to learn data science practices and tools:
-
The practical way: Find a problem to solve, learn and apply different
techniques until you solve it -
The classic way: Get a book, study the algorithms, and do the exercises
recommended in the book -
The virtual way: Register for a quick online course. Follow the
explanations, do the exercises, and get your certificate (if any) -
The comprehensive way: Attend a data science program at a
college/university -
The mentoring way: Work with a colleague who is nice enough to introduce
you to the world of machine learning
My suggestion: Learning is an investment in time and often money, so pick an
approach that works best for your learning style and what you can afford. To
go from basics to mastery, you’re probably going to have to apply multiple
approaches. For example, you might apply “the classic way” to learn
statistics, “the virtual way” when learning a data viz tool, and “the
comprehensive way” when you’re ready to learn the math, science, and
implementation behind machine learning algorithms.
Lauren Clayberg, software engineer at
mabl, offers these
suggestions regardless of how you go about your learning. She says,
“Learning machine learning starts with understanding its core components:
the math behind optimization, how to find high-quality data, and the
benefits of different performance metrics. Since machine learning is just
automated optimization, building a solid understanding of these pillars
allows people to dive deeper into specific models and how to use them.”
3. Join an analytics team, deliver a business-impacting model
Rev 3 was filled with top data scientists that wanted to expand their
skillsets, and Nick Elprin, CEO & co-founder of
Domino Data Lab,
shared this key revelation with them. He said, “Playtime is over for data
science. If your work is relegated to an innovation lab, then you are
already behind.”
That’s also great advice to aspiring data scientists and citizen data
scientists. Bringing data visualizations, analytics, and machine learning
models to production, delivering business impact, and providing support
around the
modelops
is not a one-person effort. It takes a team of subject matter experts who
know the business, dataops engineers, citizen data scientists, data
governance specialists, and data scientists to collaborate on the problem,
experiments, and working solutions. To collaborate as a team, I shared why
data Science, dataops, and data governance need agile methodologies and
three ways to apply agile in data science and dataops in previous posts.
Going back to Hayes’ remarks, “The first decision in decision making is when
to make a decision.” It’s time to make a decision and develop your data
science skills!
