Most client report questions do not start with a mistake in the data. They start with a number that is perfectly correct and a sentence next to it that does not say anything the client could not already see. The report gets read, the tile gets noticed, and the thread starts anyway.
The tell is who asks. If the question comes from your day-to-day contact, it is often a conversation you would have had regardless. If it comes from someone forwarded into the thread four days later, a finance lead or a brand director who was not on the original send, the commentary failed a specific test: it did not survive leaving the room it was written in.
Three mistakes cause most of that. They are repeatable, which is the good news, because a repeatable mistake has a repeatable fix.
Accuracy stops audit questions, not client report questions
Reporting accuracy is table stakes. Spend reconciles with the media plan, the impression counts match the platform, the pacing line lands where it should. None of that stops a client from writing back, because the client is not auditing your arithmetic. They are trying to decide whether anything in the report changes what they do next week.
That is the gap the commentary is supposed to close, and it is usually written last, at the end of a reporting day, by someone who has been staring at the same figures for two hours and can no longer tell which parts are obvious. The result reads fine to the author and thin to everyone else.
Mistake one: restating the tile instead of naming the cause
The most common line in agency report commentary is a prose version of the number directly above it. “CPM increased to €8.40 this month.” “CTR came in at 0.31%.” The client can read. What they cannot read off the tile is why.
This happens because naming a cause requires committing to an explanation, and explanations can be wrong. Restating is safe. It is also the reason the reply arrives asking exactly the thing you avoided answering.
The fix: every movement sentence gets a “because” clause, or it does not get written. “CPM rose 14% because we shifted 30% of budget into BVOD inventory in week three, where the floor price is structurally higher than display.” If you genuinely do not know the cause yet, say that in those words and say when you will know. “CPV climbed in the last ten days and we have not isolated it yet. We are pulling placement-level data on Thursday.” A stated unknown with a date attached almost never generates a follow-up. A restated number does.
There is a version of this mistake that hides in the vagueness of the cause. Writing “seasonality” or “market conditions” next to a 40% swing is restating with extra syllables. If the explanation would fit any client in any month, it is not an explanation.
Mistake two: leaving out what the number is actually being compared to
“Conversions were up 22%.” Against what? Last month, which had a public holiday in it. The same month last year, when the campaign was running on half the budget. The forecast in the media plan. The client’s own internal target, which you may not have seen.
The account manager knows which baseline they used, because they picked it. Nobody else does, and the assumption most clients default to is the one that flatters them least or most, depending on the week. When a CFO reads “up 22%” and mentally compares it to the annual plan while you compared it to August, you now have a disagreement that has nothing to do with performance.
The fix: name the baseline in the same sentence as the movement. Not in a footnote, not in the period selector at the top of the dashboard. “Conversions were up 22% against August, which ran four fewer days of paid search. Against the quarterly plan we are tracking 6% behind.”
Two baselines in one sentence is not clutter. It is the difference between a number and a judgement. If you are unsure which metrics deserve this treatment, the twelve metrics most client reports should carry is a reasonable starting list, and each of them has an obvious natural comparison point built into how it is read.
Mistake three: describing the movement but never saying whether anyone should act
This is the one that generates the politest and most persistent emails. The commentary explains the swing, names the cause, cites the baseline, and then stops. The client finishes reading with a clear picture of what happened and no idea whether they are supposed to do something about it.
So they ask. Usually in the form of “thanks, this is helpful, do we need to be worried about the CPA?” which is a question your commentary should have pre-empted in six words.
The fix: end each note with one of three positions, and be explicit about which one you are taking. No action needed and here is why. Action taken already and here is what. Decision needed from you by this date. “CPA rose to €41 from €34. This is within the range we expected when we opened the prospecting audiences, and we are leaving it for another two weeks before judging. No action needed from your side.” That paragraph ends a thread. The same paragraph without the last sentence starts one.
Agency owners sometimes resist this because stating “no action needed” feels like inviting scrutiny if things later go wrong. In practice the opposite happens. A documented position with a review date is what protects you in the quarterly review. Silence reads as an oversight.
The change you made and never mentioned is the question you will get anyway
Separate from the three mistakes, there is a reliable source of follow-up that has nothing to do with writing quality: the optimisation you made mid-flight and did not put in the report.
You paused two underperforming placements on the 12th. You moved budget between line items. You swapped in the second creative set a week early. All defensible calls, all inside your remit, none of them visible in a table of totals. The client sees a discontinuity in the trend line and asks what happened, and now you are explaining a decision retroactively, which always sounds more defensive than explaining it in advance.
A short “changes we made this period” block, three or four bullets, dated, costs five minutes and removes an entire category of question. It also builds a record that becomes very useful when someone asks in January what you did in September.
Deciding which movements earn a sentence
Not every metric that moved needs commentary, and writing a note for all of them is how the useful sentences get buried. A rough filter that holds up: a movement earns a sentence if it crosses a threshold the client cares about, if it changes what you would recommend, if it will look alarming without context, or if the client asked about that metric last time.
Everything else can sit on the dashboard uncommented. A 3% wobble in impressions does not need a paragraph. Writing one trains the client to read every number as significant, which is precisely the habit that produces reply-all threads about noise.
Commentary that survives being forwarded
Write every note assuming the reader was not on the call, does not know what you agreed in August, and will read it three weeks from now. That means spelling out acronyms the first time, naming the campaign or channel rather than saying “this line”, and never referring to “the conversation we had” without summarising it in a clause.
It also means keeping the commentary attached to the number rather than in the covering email. An email note is invisible the moment the report is forwarded, which is exactly when the third person on the thread starts asking. In Screebl, a report template can carry text rows alongside the KPI tiles and charts, so the movement, the cause and the position sit under the chart they refer to and travel with the link, the PDF and the client portal view. If you have written a campaign brief, the report assistant on the shared report can also answer the smaller factual questions directly from that report, which takes the “what does VTR mean again” traffic off your inbox without touching the judgement calls.
The bigger questions are still yours, and they are better handled live. Walking a client through a swing on a shared screen takes four minutes and replaces a week of asynchronous back-and-forth. Written commentary is not a substitute for that conversation. It is what stops the conversation from being needed for things that were never worth a meeting.
If you want reporting where the commentary lives on the report instead of in a thread only you can answer, the plan comparison is the place to start.
