Write findings a busy manager will act on

Here are two reports on the same analysis. Both are accurate.

Report A

"Analysis of Bisi Provisions sales data, January to June 2026. The dataset contained 2,010 rows after cleaning. Revenue by branch was calculated using SUMIFS and cross-checked against a pivot table. Yaba recorded ₦11,915,250, Ikeja ₦8,664,700 and Surulere ₦4,930,400. Growth was analysed on a month-on-month basis…"

Report B

"Your group revenue grew 31% last quarter, but about a fifth of that was the April price rise rather than extra sales. Yaba and Ikeja are roughly flat once you allow for it. The real growth is Surulere. I'd give it more stock before December."

Report A describes what the analyst did. Report B tells Bisi what to do on Monday.

A took longer to write. B is the one that changes anything.

Most analysts write Report A, because it is the honest record of their effort, and because putting the conclusion first feels like showing off. This lesson is about writing Report B without losing anything that made A trustworthy.

Words in this lesson

Put the answer first

Your reader is busy, will read the first two sentences properly, and may stop there. So the first two sentences must carry the answer.

This feels wrong the first few times. You did the work in order — clean, analyse, conclude — so you want to write it in that order. Your reader isn't reproducing your work; they're deciding something. Give them the decision first and the working underneath.

The four-part finding

Every finding worth reporting has four parts. Miss one and the reader has to guess.

The one people drop is so what. "Grains is 50% of revenue" is a fact. "Half the business depends on one category's supply, and it has the thinnest margin, so a rice shortage is a business problem not a category problem" is a finding.

Test every sentence in your report with "so what?" If there's no answer, it's background, and background belongs in an appendix.

Say how confident you are

Not everything you find deserves the same weight, and a report that presents everything at one confidence level is misleading even when every number is right.

Analysts lose credibility by being certain about the wrong things, not by admitting uncertainty. Say "I don't know, and here's what would tell us" and you'll be trusted more, not less.

Where the caveats go

Not in a footnote nobody reads, and not scattered through the text where they interrupt.

Put the important ones in the sentence they qualify:

"Surulere took 23% of revenue over February and March — the months all three branches traded."

Put the rest in a short, clearly-labelled block near the end:

What this covers, and what it doesn't Bisi Provisions sales, January to June 2026, all three branches, 2,010 lines, reconciled to ₦25,510,350. Surulere opened on 2 February and has no January sales. Its branch record gives an opening date of 2025-11-01 and four of its five regular customers have earlier first-purchase dates, so roughly three months of its sales appear to be missing from the source system. Every Surulere figure here is therefore a floor, not a total. April to June includes a 7–8% price rise on four products. The products table still lists pre-April prices; cost prices carry no date and may also be stale, so Q2 margins are provisional.

That block is the lesson 4.4 coverage statement, and it is the most professional thing in the document.

Recommend something

A finding with no recommendation makes the reader do the last and hardest step alone.

A good recommendation is specific, achievable by this reader, and tied to the finding:

And where the data can't support a recommendation, say that instead of inventing one:

"I can't tell you whether the price rise cost you customers. Volume grew 17% in April, which suggests not, but that's one month and other things changed. Two more months would settle it."

That sentence is more valuable than a confident guess, and a reader who has been burned by confident guesses will notice.

Structure that works

For a one-page finding:

1. The answer — two or three sentences
2. What I'd do — two or three bullets
3. The evidence — a short section per finding, with a chart each
4. What this covers and what it doesn't — the caveat block
5. What I'd check next — the open questions

For a single email, just parts 1 and 2, with the rest attached.

Write for the reader you actually have

Bisi runs three provision shops. She is sharp about her business and does not know what a percentile is.

Nobody has ever been annoyed by a report being too easy to read. Plain language is not dumbing down; it is the whole job. If your reader has to decode you, you have moved work from yourself onto them.

Your method belongs in an appendix, or in the workbook, where the person checking your work can find it. It does not belong in the first paragraph.

The test before sending

Read your report and ask:

1. If they read only the first two sentences, do they know the answer?
2. Can they name one thing to do differently on Monday?
3. Is every number in it reconciled?
4. Does every uncertain claim sound uncertain?
5. Would they be surprised later by something I knew and didn't say?

Question 5 is the one that matters. A report that surprises the reader later — "you never told me Surulere's data was incomplete" — destroys trust permanently, even if every sentence in it was true.

Common mistakes

The short version

• Answer first. The reader may only read two sentences, so put the conclusion there.
• Every finding has four parts: what, so what, how sure, what now. The missing one is usually so what.
• Match your language to your confidence. Being wrong while certain costs far more than admitting uncertainty.
• Important caveats go in the sentence; the rest go in a labelled coverage block.
• Recommend something specific, or explain why the data can't support one.
• Write for the reader you have. Plain language is the job, not a compromise.
• Before sending, check they can't be surprised later by something you knew.

Try it yourself

1. Write Bisi's answer in two sentences. No method, no jargon.
2. Take three findings from Modules 4–6 and write all four parts for each.
3. Find a fact in your earlier work with no "so what". Either give it one or delete it.
4. Write the full coverage block for the half-year data, including Surulere and the price rise.
5. Write three recommendations, each specific enough that Bisi could act on Monday.
6. Write one paragraph about something the data cannot tell her, and what would settle it.
7. Give the whole thing to someone outside this course. Ask them what Bisi should do. If they can't say, rewrite.
8. Re-read it against the five-question test. Note what you changed.

You'll know you've got it when your reader replies about the decision rather than asking what the numbers mean.