New Year's Resolution: Help Data Scientists Help You

Scaling AI Lynn Heidmann

Earlier this month, Forbes cited the key to getting value from data science as teamwork. That means in the coming year, with an increased focus on collaboration, lines of business will likely have more opportunities than ever to work together with data scientists. But this relationship will only work if business people help data teams help them by asking the right questions.


“If I had an hour to solve a problem and my life depended on the solution, I would spend the first 55 minutes determining the proper question to ask… for once I know the proper question, I could solve the problem in less than five minutes.”

- Albert Einstein

What Is the “Right” Question?

Asking the right question to be solved by data science, machine learning, and eventually by artificial intelligence (AI) can make the difference between building a one-time project that then must constantly be rebuilt or updated vs. a data product that can be built once and then continue to provide value over time.

Of course, what constitutes the right question depends largely on the industry, the specific business, and the role (for example, marketing will obviously have different questions than the risk team). But there are a few larger rules and things to consider to make sure you’ll get the maximum amount of value from the skills and tools data scientists provide.

Think Beyond Dashboards

Often times when lines of business get access to data team resources, they immediately ask for a dashboard that shows X (whether that is top products sold, highest-revenue customers, maintenance status for high-value assets, at-risk accounts, etc.).

While dashboards aren’t inherently bad and can be useful in specific cases and for certain audiences, they generally only do one thing: describe existing or past data. What they don’t do is predict future events or, perhaps more importantly, prescribe a course of action.

That means that oftentimes, teams ask for dashboards, but the dashboard doesn’t actually help them be more efficient or address the real root problem or goal (leaving the dashboard unused or rendering the data team’s time useless as they go back to the drawing board).

Get to the Root of the Problem

So instead of simply asking for a dashboard, ask yourself: what are you really trying to achieve? If you come to the data team with that problem or question instead, they likely will be able to think of a more creative solution that gets you where you need to be.

iceburg-below-surfaceEffective data science, machine learning, and AI projects tackle the root of the problem - not surface-level questions.

For example, instead of asking for a dashboard showing the customers most at risk of churning, consider asking the data team for help in being more efficient and effective in reaching at-risk customers. The solution might then be more than a static report, but an entire system that looks at data from all sources, analyzes it, automatically identifies at-risk customers, predicts whether they are likely to respond to a marketing campaign, and if yes, then automatically sends out an email with a special offer to those specific customers.

Much more useful than a dashboard, right?

Focus on Action

Identifying the root problem or question, and not an intermediary question, can be easier said than done. But one way to get there is to always focus on the end action; that is - what should be the result or the output?

This is best illustrated with examples. Let’s say you come to the data team with a question like one of these:

  • What is the health of all the machines on the shop floor?
  • Which patients are unlikely to show up to their appointments?
  • How can a manual review team correctly identify a fraudulent charge or claim?
  • Which advertisements are the most effective ones?

While these are all important pieces of a larger puzzle, it’s easy to see how a data team could deliver on these questions but not ultimately help your line of business be more efficient or more effective in the long term. In fact, the deliverable for most of these questions could be a static dashboard, and it would provide an answer. But so what?

Now think about what you could achieve by working with the data team to answer these questions instead:

  • How can we repair machines on the shop floor that are likely to break down before they actually do?
  • How can we suggest appointment times and days to likely no-show patients for which they are more likely to show up?
  • How can we accurately identify fraudulent charges or claims automatically so that we can rely less on the manual review team (or so that they only deal with very difficult cases)?
  • How can we connect audiences with the most effective advertisements for their segments?


These types of action-based questions are much more likely to deliver business-impacting results.

Run an Efficient POC

Choosing a good business question and use case can lead to effective proof of concepts (POCs) that then smooth the process of incorporating data into business processes in the future. For more on running an efficient POC, get the free POC white paper. 



You May Also Like

6 Ways AI Will Change Media & Entertainment

Read More

AI Projects Lifecycle: Key Steps and Considerations

Read More

Remote Data Science: How to Make it Work

Read More

What Is MLOps & Why It Matters More Than Ever

Read More