I replaced a monthly client report with a 90-day review — and found two broken comparisons in my own numbers

The August report was finished. Fifteen pages, formatted, ready to send. I deleted the frame and rewrote it as "the first ninety days".

Not because the August work was weak. Because a monthly report was quietly making me argue the wrong case.

What a month-shaped report forces you to say

Reconstructing the case took an afternoon of pulling my own session notes out of my knowledge base with Claude — every decision, audit and handoff since April, in order. That is the part that made the reframe possible at all: I could see the shape of the three months, not just the last one.

I started this mandate at the beginning of June. Two of the three months since had no advertising spend at all — they were sortiment audits, keyword work, a bidding model, campaign build, a gated price list. Foundations.

In a report called August, all of that is backstory. It sits in a paragraph labelled "preparation" and the reader's eye slides past it to the part with the numbers. Worse, it puts you on the defensive about the two months where nothing measurable happened, which converts real work into something you have to justify.

In a report called the first ninety days, the same facts do the opposite job. The groundwork is the deliverable. The fact that money only started moving in month three is the plan working, not a delay to explain.

And crucially: the growth plan we agreed back in spring had already defined a first stage running June to August, with a second stage from September. The mandate started on the first of June and the reporting window closed on the last of August. Three months to the day. I did not invent a flattering frame — I noticed the frame was sitting there in the plan the whole time, and I had been ignoring it in favour of a calendar month because that is what reports are usually called.

As David McRaney puts it in You Are Not So Smart: you don't think in statistics, you think in examples, in stories. The numbers were identical in both documents. Only one of them made sense.

The part where I had to correct myself twice

Rewriting the frame made me re-derive the numbers, and re-deriving numbers is where you find out what you got away with.

First: my comparison windows were different lengths. The original draft compared the first seventeen days of August against the last fourteen. Seventeen against fourteen. I had not done it deliberately; I had cut the window at the date something went live and not gone back to check the other side. Every analytics tool that offers period comparison defaults to an equal-length preceding range for exactly this reason, and I had hand-rolled my way out of that protection.

Corrected to two clean fourteen-day windows, the picture actually got better — but "better after the correction" is not the point. The point is that a report can carry a number nobody will ever check, and the only defence is checking it yourself before someone else has to.

Second, and worse: my baseline and my target were not measuring the same thing. The starting figure was enquiries-per-day counted one specific way. The current figure included a second enquiry route that did not exist when the target was set. So a comparison that looked like progress was partly a comparison between two different definitions.

This is the exact failure I wrote about a couple of weeks ago when three separate ways in were being reported as one, and I still walked into a version of it. Adding a measurement route is a change to the ruler. If you change the ruler mid-window, the before and the after are not comparable, and saying so out loud is not a weakness in the report — it is the report doing its job.

Both figures are now carried separately, forward, from here. And a third correction, smaller but the kind that erodes trust fastest if a client catches it: I had described the baseline as "enquiries with no advertising at all". Not true. There were directory listings running the whole time. Untracked is not the same as absent.

The reframe that mattered most wasn't in the numbers

Late in the same week, a nightly sync flipped several hundred products to private. It looked like a fresh incident. Something broke in September.

Except the audit from spring had already recorded that the large majority of the catalogue had no long description. The sync had not broken anything. It had made an existing structural gap visible by acting on it. "Something is broken since September" is a fight with whoever runs the sync. "The audit we did in May told us this, and it has now surfaced" is a plan.

Same facts. One version makes an enemy, the other makes a roadmap. If you have done the diagnostic work up front, you get to choose which one you tell — and you get to choose it honestly, which is the only version worth having.

What I take from it

Reporting cadence is not an administrative choice. Monthly is a container someone else designed, and when your work doesn't fit it you end up narrating around the container instead of the work.

So before writing the next one, three questions. Does a period already exist in the agreed plan that this report should map to? Are the two windows I'm about to compare the same length? Has anything about how I count changed between the start of this period and the end of it?

The third one is the one that catches people. It caught me. And it is the reason I now put the exclusions and the counting rules in the document itself rather than in my head — the same instinct that makes me itemise a quote instead of quoting a round number. Show the working, and the number stops being something the client has to take on faith.

I threw away a finished report. It took an afternoon to rebuild and it is the first one I would defend line by line.

Sources & further reading

External
Kaplan & Norton — The Balanced Scorecard: Measures That Drive Performance (HBR)
Matomo — Compare date ranges (equal-length comparison periods)
Adobe Analytics — Date comparison in Analysis Workspace

Related posts
Three ways in, none of them a funnel
The files were never empty
I put every hour in the quote