You’ve heard about systems of intelligence, you're collecting lots of data, and you’re ready to start getting value from it. You’re well versed in the seven steps to complete a data project. The only remaining question is: what should you tackle first?
Choosing your first data project, or your first system of intelligence, is a fine balance between impact and feasibility, a tightrope act where failure means losing the confidence and trust of your audience (whether that’s other teams or management), and success means applause and encouragement to carry on, tackling larger and more high-profile projects.
Today, most companies would say they prioritize the use of data and analytics for decision making (after all, in the era of big data, companies that don’t will not survive). Yet despite this, studies show that there is still an underlying lack of trust. One survey revealed that only about 50 percent of respondents believe that C-suite executives fully support their organization's data and analytics strategy, and 60 percent of organizations say they are not very confident in their data and analytics insights. Even more worrisome, only 16 percent of respondents believe they perform well in ensuring the accuracy of models they produce.
Given this the success of your first data project is critical in building the trust and confidence you need to build the team and be entrusted to take on more work. So you’ll need to choose wisely... (no pressure).
We’ve come up with a five fundamental considerations when choosing your first data project to ensure you’re set up for success:
- Choose a project that already matches your team’s skillset; experimenting with new, unfamiliar things is too risky.
- Consider the visibility of the project; will the results be noticed by the right people?
- Keep it relatively small; don’t bite off more than you can chew - be sure you can complete the project in a timely fashion.
- Set clear and impactful goals; the clearer the goals, the easier they are to meet.
- Show repeatability and/or scalability; show that the team’s ability to deliver results isn’t an anomaly.
1. Match Your Team’s Skillset
Whatever first project you choose, ensure you have the right skills on the data team in order to execute. Banking on hiring someone down the road who will be able to contribute with specific skills or learning new skills along the way are risky when it’s critical to deliver results quickly. While upleveling skills is something the data team can (and should!) work toward later on, for initial projects, it’s best to leverage existing knowledge.
Instead, choose a project with which the team already feels comfortable. Ideally, someone on the team will have worked on a similar project in the past to provide valuable insight on the realistic timeline for completion and potential pitfalls to avoid along the way.
2. Consider Visibility
Perhaps one of the most important considerations when choosing your first data project is knowing who you will ultimately need to impress - who has the power to give your team the resources and support needed to continue if the first data project is a success?
Depending on your answer to that question, you might choose a project that more directly impacts that person’s work or one of his/her quarterly objectives. For example, if the person ultimately responsible for the fate and direction of your data team is the CIO, it might not be wise to choose an initial project that delivers results to the marketing team. It will be harder to illustrate the return on investment (ROI) of the project and the team overall on initiatives with which the stakeholder is not familiar. In this case, it would be better to choose a project directly impacting the CIO’s key results.
3. Keep It Small
All of the above being said, while you do want to go for impact of key players when choosing your first data project, you don’t want to get in over your head (classic advice) - this is where the balancing act really comes in. Keep your first project manageable and ensure you can execute in a relatively short timeframe. You don’t want to be in a place where you don’t have the right data, the right access to data, or the right resources to deliver on your goals when you’re already weeks into the project.
Keep your first data project small and simple (but impactful) for quick results.
In other words, save your lofty goals for later. First, prioritize a project with high enough impact, but that's low enough on time commitment and hurdles to ensure success. And make sure you’re realistic and honest with yourself about the timeline and how long your first data project that you’ve chosen will feasibly take you!
If you can successfully deliver results quickly with an important but short data project, it’s highly likely your team will be entrusted in the future to execute on larger and more important tasks that do allow the team to grow their skills.
4. Set Clear, Impactful Goals
Once you’ve decided on your first data project, don’t immediately dive in. Before getting started, make sure there are clear end goals so that the project doesn’t drag on and so that everyone is working together toward the same result.
Also, don’t confuse prototyping and experimentation with actual results in production. The goal of your first data project should almost always be getting a model into production, not just producing results in a sandbox or dev environment. When talking about and presenting the results of your first data project to your peers and management, you want to be able to show the end-to-end product. If you’re not setting your sights on production, you’re not setting realistic expectations for future data projects.
5. Show Repeatability, Scalability
The final consideration? Make sure right off the bat from your first data project that you can prove your success was not a one-time act.
The ability to replicate success and scale are paramount in showing your data team's abilities.
You can walk the tightrope again (and again and again). You can do this by building in automation from the start, ensuring that processes are reproducible, models can be easily monitored and re-trained even while in production, and that the data team itself is set up to take on multiple projects at once. Hint: starting off with a data team tool can help with all of this, and it can also help with data governance and privacy, which will become a bigger concern the more things your team takes on.