Done For Now: Why Analytics Work Never Really Ends

The whole point of analytics is to enable meaningful decisions. But how do we make that happen on an ongoing basis? Calling an analytics project “done” depends on who’s asking. Every project has multiple stakeholders, from the data team to the business users, so there’s a “done” stage for each of them. For me, you can boil “done” down to two questions: is it accurate, and is it helpful?

Accuracy covers the technical side: are we pulling the right data? Is it aggregating it correctly? Do numbers match and split out as expected? Helpful is the business side: does the accuracy actually mean something? Can someone act on it? Does it shape or influence a decision?

That’s a pretty standard definition of done, but I think the step most people miss is what comes after done. How do we make sure it stays accurate? How do we make sure it stays helpful? Because the truth is, any good analysis is never truly finished. At best, it’s done for now.

Close enough isn’t always good enough

The risk of calling something done too quickly is that you stop before you’ve really met the need. A report might be useful in the short term, but could it be better for the long term? Could it serve more stakeholders? Does it spark the right follow-up questions? Will it adapt to new contexts? Every time we wrap too quickly, we take on the opportunity cost of not digging into the data enough or pushing that to-do item down the line when the stakeholder comes back with a follow up question. 

That’s why I don’t think of “done” as final. I think of it as a checkpoint. We’ve answered this question, supported this decision, and then what? What happened as a result of that analysis? Was it the right decision? Should something change?

When Taylor Swift breaks your model

You might be wondering, “So what, Katrina, you want me to have an ever growing to-do list where everything is in a constant state of V1, V2, and V3 development?” 

No, of course that would be unrealistic for how organizations actually function. This is where my favorite phrase comes in: rigid flexibility. That means balancing the process with the people.

Rigid means having a plan and alignment to company goals, which gives you the ability to say “no.” You can’t let every ad hoc request jump the queue, and you can’t give a half-baked artifact the same weight as a polished one. Without solid guardrails and processes, one-off requests pile up, users struggle to navigate the existing analytics, and the result is a self-fulfilling cycle of more ad hoc and repeat asks.  

Flexibility means making space for the unpredictable. You can’t only work on six-month projects when the world throws surprises every day. Imagine Taylor Swift is on tour in your city. How will that impact hotel prices and restaurant bookings? Six months earlier, no one knew that was coming. If your system is too rigid, you miss the chance to react to moments like that.

Rigid flexibility means holding on tightly to fundamentals while leaving room to adapt. It’s knowing your long-term goals and direction, but being willing to wiggle and pivot to answer urgent questions that matter today.

Accuracy plus helpful

So how do we make rigid flexibility a reality? By creating a clear framework for balancing accuracy plus helpful on an ongoing basis. 

Here are a few questions to gauge your approach:

• Self-service path: Can the requester answer this safely with today’s tools? If not, what single change would make that possible?

• Promote: When a quick analysis proves useful, what is the path to promote it into the governed layer?

• Purpose, owner, and timing: Who owns this workbook, what decision does it support, and when must that decision be made?

• Accuracy guardrails: Which metrics can be directional and which can’t? Is there a tolerance for “close enough for today and better tomorrow”? 

• Lifecycle benchmarks: What usage threshold triggers action? For example, can we hide this if there are no views in 60 to 90 days, archive at 120 to 180 days (unless tagged for quarterly or annual use)?

• Hide-and-listen flow: When something is hidden, who watches for “where did it go” pings? Could we move parts of the analysis into a better home?

Rigid flexibility is not meant to be more red tape. It keeps analytics accurate and helpful as the business shifts. A little upfront effort and regular pulse checks save time later and keep people confident in the numbers. Treat done as a starting point, usable today and ready to evolve as you learn.