Every export is a conversation you didn't have
READ TIME - 4 minutes ⏳
Email formatting looks broken? click here to view in browser.
I'm in the early phase of a design system project with a client right now. Before you build a system, you audit what already exists, and the usual version of that audit looks at standardization: how consistent the layouts, the colors, the naming and the navigation are from one report to the next.
I suggested we add one thing these audits almost always skip. Before we judged how the reports looked, I wanted to know how they were actually being used. Specifically, I wanted the export-to-Excel numbers.
Here is why. When you are choosing where to start a design system, the temptation is to pick the report that looks the worst. That is the wrong filter. The report you want for a pilot is the one failing its users the hardest, and the export rate points straight at it. A report people constantly export is a report that is not answering them on the screen. That makes it your best candidate, because that is where a redesign has the most to prove.
The process
I started by asking the Power BI admin team for a usage extract over a six-month window, just for the sample of reports we had flagged as pilot candidates. I wanted a sense of how those specific reports were being consumed.
But looking at a sample raised an obvious problem. A handful of reports cannot tell you whether what you are seeing is normal or exceptional. You need the baseline. So I asked for the complete pull across the whole estate, and that is where the extract turned from useful into genuinely revealing.
A full usage log carries far more than view counts. It tells you how many reports exist and how many are actually used, how many workspaces and apps you are running, and how many datasets you have. It also shows how many of those datasets are shared across several reports, versus standalone models that should really be promoted to a proper semantic model. That alone is a map of the estate that most organizations have never actually seen.
And sitting inside all of it is the number I came for: for every report, how many of the views ended in an export to Excel.
What you do with it
Once you can rank reports by export rate, the work stops being about logs and starts being about people. You take the reports at the top of that list, you look at who is exporting them, and then you do the thing the whole exercise was really for. You go and talk to those users.
You ask them why they export. What they are doing in Excel that they could not do in the report. What is missing. What decision they are trying to reach that the report does not get them to.
Every one of those conversations is the conversation that should have happened before the report was ever built. So the export rate does two jobs at once. It finds your pilot, and it hands you the exact list of users to interview along with a concrete reason to interview them. That is a far easier door to knock on than a vague "can I get some feedback," because you arrive with evidence and a real question already in hand.
The lesson underneath
This whole move is a way to reopen a conversation after the fact. It works, and I would run it on any estate. But it is worth being honest about what it is: a repair. You are paying, six months later, to have a discussion that costs almost nothing at the start, when all it takes is asking who the report is for and what they need to decide before you place a single visual.
The export button is what that skipped conversation sounds like when it finally reaches you.
It is thousands of people quietly telling you the report did not answer them, in the only language the report left them.
One last thing worth doing
If your organization runs more than one BI tool, run the same export analysis on the others, then compare.
If the export behaviour is bad only in Power BI, you may genuinely have a Power BI problem, something about how those particular reports are built. But if you see the same pattern everywhere, across every tool you own, then it was never about the tool at all. It is a practice problem. It means reports get built the same way regardless of the technology, without the conversation at the start, and no new platform is going to fix that for you.
That is the more uncomfortable finding. It is usually the more valuable one too.
None of this needed new tooling or a budget. It needed one number the usual audit forgets to pull, and the willingness to treat that number as a question instead of a statistic. The export rate doesn't just tell you which report to fix first. It tells you the fix was never really about the report. It was about the conversation nobody had at the start.
So find the exports, find the people, and ask them why. Then, on the next report, ask before you build.
That is the whole thing.
See you in two weeks,
Julien
