You've probably heard of it before, but what is it, why is it important, and how is it relevant to me?
Gartner defines Data Strategy as "a highly dynamic process employed to support the acquisition, organization, analysis, and delivery of data in support of business objectives."
While there are many strategy frameworks such as Defensive vs. Offensive, it should always point to the business' goals & objectives (as the definition points out). However, creating and implementing a Data Strategy is much easier said than done - that's why only 1/3 of organizations have a Data Strategy and only 13% are actually delivering on them.
Knowing what you want to do is very different than knowing the best way to do it.
I often hear variants of the following scenario: "Daniel! We want to leverage an AI-based [product/feature/capability] which will read X data and output Y! We're just starting and need an AI/ML expert to build it." Usually, when I ask about X data, what it looks like, how it's being acquired, etc., I'm given a brief 1000-mile stare followed by an optimistic "Well, that's why we need an AI/ML expert!"
Most of the time, this founder actually needs help with Data Engineering to build the company's "data irrigation system" before doing anything with AI/ML; and before that, they need someone to build a holistic Data Strategy that's cohesive with the overall business goals, challenges and roadmaps. Knowing what you want to do is very different than knowing the best way to do it. That's why investing in strategy is important - because it could save months of wasted time, money, and energy spent moving in the wrong direction.
You must walk before you can run, and you must crawl before you can walk.
I like to reference Monica Rogati's "Data Science Hierarchy of Needs" when starting to talk to companies about their Data Journey. Before you do anything, it's critical that you understand where you are in your Data Journey. Where are you?
Are you ingesting any data? If so, what does it look like? Will it change? If so, in what ways and how frequently? Does the data flow through your entire system? Can you analyze any of it? All of it? Do you have ideas as to which features are the most critical to training future models? Will you have the ability to collect these features? Do you have the ability to run simple experiments with your data? Why do you think an AI-based model or feature is the answer? Do you have a performance baseline for your goals/tasks? Do you have success criteria for your model? As you can see, the questions can go on and on...
I remind any team starting a new data initiative: You must walk before you can run, and you must crawl before you can walk.
I say all of this not as an excuse to spend ages building infrastructure, not developing that MVP/prototype/proof-of-concept, but as a reminder to founders and their teams to avoid letting the tail wag the dog. Start with your business goals and work backwards, build a plan for what needs to be in place to achieve those goals, and work with the right people to help you execute on those plans.
Good luck to anyone out there building! Closing with a quote I totally ripped off Ben Franklin: an ounce of Data Strategy is worth a pound of tweaking your AI-thingamajig.
________________________________________________
If you or someone you know would like to talk about Data Strategy, Data Science & Engineering, or Applied AI & ML, please feel free to reach out (daniel.c.leblanc94@gmail.com) - I'd love to chat!
Comments