If I imagine myself waking up to a notification showing statistics that I should have slept more, it summons feelings of hopelessness. Inevitably, I should have slept more but the way to enable that involves witchcraft on a toddler, and I don’t possess a wand.
It’s tempting to track sleep simply because you can. The data might be interesting. I could be convinced if I were actively experimenting: do I sleep better after a cup of tea? What about going to bed earlier? But if you’re like me, you already know both the problem and the solution — and no amount of data is going to change that. In fact, any data that doesn’t align with my theory that the toddler is the problem is going to be rejected as bad data.
This brings me to my topic: actionable data.
Data projects only become meaningful when the data inspires action. Action doesn’t have to be immediate, but it does have to be plausible. If the data can’t realistically change what someone does, now or in the future, it’s unlikely to be worth the time spent collecting or analyzing it.
So before starting a data project, it’s worth asking a few questions:
• What would I do if the data confirmed what I already believe?
• What would I do if it told me the opposite?
• Who will see this analysis, and how will it influence their decisions?
I was recently asked to re-run an analysis I’d done several years ago to see if we were in a better position. While I understood why it might be interesting to revisit, we already believed we knew the answer, and more importantly, the analysis wasn’t going to provide new information or unlock new action.
Given time is finite, it’s worth asking: why are we spending time confirming something we already think we know? If the result doesn’t change a decision, reduce uncertainty, or enable a different conversation, then even a correct answer has limited value.
Instead of saying that directly, I asked questions like: Why might things be different now? and Who will see this analysis that didn’t the first time? The goal was to step back from the question itself and focus on what the answer would actually enable.
Asking these questions can feel like waking someone abruptly from a dream. It forces you to imagine a future where, after a lot of work, you realize you haven’t solved the actual problem (the toddler with FOMO who rejects sleep on principle).
But if imagining that future reveals a different decision, a different framing of the problem, or even a different question you’d ask next, congratulations – you have actionable data. Sometimes the most valuable action is realizing the problem isn’t what you thought it was.
Collecting more data in the hopes that any extra information will eventually be useful is a form of hoarding. More data is not better, and bigger data is not better. This isn’t an argument for less curiosity; it’s an argument for intentional curiosity.
If time were unlimited, we could collect all the data and run every analysis. But time is finite. Focusing on projects that go beyond “interesting” and have the potential to influence future decisions is where effort pays off. I could have logged every diaper change, but as much as I enjoy a good spreadsheet, it violates the principle of opportunity cost: time spent on one thing is time not spent on another.
By helping people think through whether their project will lead to actionable data, you’re not just protecting analyst time. You’re helping your data customers avoid disappointment later when their project doesn’t deliver the impact they imagined.
Before gathering more data or starting another analysis, think carefully about what you’d actually do with the results. Sometimes interesting insights grow into actionable change — but if you’ve already started trying to solve the problem, make sure the analysis will genuinely help you solve it.
A quick pre-analysis checklist
Before starting a new data project, it’s worth pausing to ask:
• What decision could this data influence?
If no decision changes, the analysis probably won’t matter.
• What action would I take if the result confirms my current belief?
And just as importantly…
• What action would I take if the result contradicts my current belief?
If the answer is “nothing,” that’s a red flag.
• What’s different now compared to the last time this question was asked?
New data, new constraints, new stakeholders, or new levers to pull?
• Who is the audience for this analysis, and how will they use it?
Interest alone is rarely enough.
• Do we have the ability to influence the outcome we’re measuring?
Insight without agency often leads to frustration.
• Is this the best use of time right now?
What are we not doing if we do this?
This blog post was written on a snow day while negotiating with a toddler nap. She eventually slept. I’m calling that a win and choosing not to collect any further data about how shockingly long it still takes to get her to nap.
By Meriel O’Conor, Director of Data Consulting and Training