Nobody argues for long reports. Everyone produces them anyway, because each individual inclusion is defensible and the total is not.
So rather than argue the principle, here is the editing done. A composite weekly summary of the sort that gets sent to restaurant owners, taken down in four passes.
The original, abridged
Weekly Review Summary — w/c 17 August
This week we received 34 new reviews across all platforms, comprising 19 five-star, 4 four-star, 2 three-star, 3 two-star and 6 one-star reviews. Our average rating for the week was 3.94, compared to an all-time average of 4.12. Total review volume was up 12% week on week.
Sentiment analysis: Positive sentiment appeared in 68% of reviews, with the most frequently mentioned positive themes being staff friendliness (11 mentions), food quality (9 mentions) and atmosphere (6 mentions). Negative sentiment appeared in 32% of reviews...
Themes: Wait times were mentioned in 7 reviews this week. Cleanliness was mentioned in 2 reviews. Value was mentioned in 4 reviews. Portion size was mentioned in 2 reviews. Staff attitude was mentioned in 3 reviews...
Response performance: 28 of 34 reviews received a response, a response rate of 82%, with an average response time of 31 hours...
Notable reviews: [six quoted reviews]
Recommendations: Consider reviewing Friday evening staffing levels. Consider a team briefing on greeting standards. Consider auditing bathroom cleaning schedules...
About 600 words in full. Everything in it is true. Nobody reads past the first paragraph.
Pass one: remove anything without a comparison
A number with no prior value cannot tell anyone whether to act. "Cleanliness was mentioned in 2 reviews" is unreadable — is that normal? Good? A crisis?
Out goes the star distribution, the sentiment percentages, and every theme count that appears once with no history. What survives is anything the report can say changed.
Roughly 60 per cent of the document goes in this pass, and this is the pass that does the most work.
Pass two: remove the positives
Contentious, so the reasoning. The owner's brief exists to allocate attention. Praise does not allocate attention — nobody reads "staff friendliness, 11 mentions" and does anything differently.
Positive material has a real use, but its audience is the team and its venue is pre-shift, where reading a review out loud by name changes behaviour. In the owner's brief it functions as reassurance, and reassurance is what makes a report skimmable-but-unread.
Keep exactly one thing: a pointer that there is something worth reading out.
Pass three: remove the recommendations
The hardest cut, because recommendations feel like the value added.
Three problems with them. They are written by whoever assembled the report, who usually knows less about the operation than the person receiving it. They are hedged into uselessness — "consider reviewing Friday evening staffing levels" is not an action. And crucially, a document containing proposals becomes a document requiring a decision, and documents requiring decisions get postponed to a quieter moment that does not arrive.
State the drift. Let the operator decide the response. The report's job is to be correct about what is happening.
Pass four: remove the response statistics
Response rate genuinely matters — it is the number most worth moving. But weekly it bounces around for reasons with no operational content: a quiet week, a holiday, one review that took two days to check a fact on.
It is a monthly number. Weekly, it is noise wearing a metric's clothing.
What survived
Negatives 11, up from 6. Rating 3.9 against 4.1 all-time.
Wait-time mentions up sharply: 7 this week, 2–3 in each of the previous four. Friday and Saturday evenings.
One needs you — Saturday two-star alleging a reaction to a dish, escalated to Priya, not yet replied.
Two one-stars are about parking. No action. One good one worth reading at pre-shift, names Maria.
Eighty-one words from six hundred. The structure is four lines and four questions, and every number in it carries a comparison.
The thing that makes this hard
Not the editing. The eighty-one word version requires more underlying work than the six hundred word one, because "2–3 in each of the previous four weeks" means four weeks of theme counts had to exist and be consistent.
The long report is easy precisely because it is a dump. It restates what arrived. The short one requires a baseline, which requires somebody to have been counting in February, and that is the part that fails — which is why most weekly review reports are long. Length is a symptom of not having history, not of trying to be thorough.
If you are building one of these yourself, start by keeping a single small table of theme counts by week. The report writes itself afterwards, in about four minutes. Without it, no amount of formatting produces a brief anyone acts on.
Ours is roughly the eighty-one-word version, sent weekly; the sample is a real one.