For agencies of 2 to 15 people with monthly retainers. Why most report pages go unread, what the report has to do to keep a client, and a format to test on one client.
A twelve-page report that nobody answers is not a sign that the client is happy. Clients who end an agency relationship cite delivery, and the monthly report is the one piece of delivery they see every month. Before changing tools, it is worth changing what the report says first: what changed, why, what happens next, and one thing the client needs to decide.
Data shapes fewer decisions than it seems
In surveys of marketers, only 37% to 53% of decisions use or are influenced by the analytics available, and those are people whose job is marketing.
37% to 53%of marketing projects or decisions use or are influenced by the analytics available, in two surveys of marketers ten years apart.
Source: The CMO Survey, February 2012; Gartner, 377 users of marketing analytics, May to June 2022 (marketers in general, not agency clients). See all benchmarks
Decisions that use the analytics available
The CMO Survey, 2012
37%
Gartner, 2022
53%
Share of marketing decisions or projects. Marketers in general, not agency clients. Higher means more use. See all benchmarks
A client who skims the report is not unusual, and the agency is not failing alone. In Gartner's survey, the top reasons analytics went unused were data that is inconsistent across sources and data that is hard to access. A report built for the client to read in one sitting removes the second reason.
What the report has to do
The report has two jobs: show the client the work, and give the client an easy place to say what is not working. Clients leave over delivery, and they rarely say so first.
Between 48% and 61% of clients who ended an agency relationship cited delivery in Setup's 2024 and 2025 Marketing Relationship Surveys. In 2025, 18% of agencies named it. An older UK study by AAR found the same gap: 83% of clients named declining output or service as the early warning sign, against 41% of agencies.
Source: Setup, Marketing Relationship Survey, 2024 (sample not stated by Setup) and 2025 (about 100 responses, 57% client-side); AAR, reported by Marketing Week, In-depth interviews: 26 client companies, 20 agencies, 77 failed relationships. See all benchmarks
The same AAR study found that 62% of clients and 70% of agencies thought too few relationships had an effective way to resolve problems, and 15% of agencies said they had no warning before the account went to review. A monthly report that ends with a question is a light version of that missing channel.
Source: AAR, reported by Marketing Week, In-depth interviews: 26 client companies, 20 agencies, 77 failed relationships. See all benchmarks
A doctoral thesis on client tolerance in agency relationships at the University of Leeds (Davies, 2002) points the same way: consistent work processes, proactivity and stable account management were among the predictors of how much a client forgives. A report that arrives the same way every month, and raises issues before the client does, is all three.
Keeping the client is worth more than the monthly fee suggests. In a study that valued the customer bases of publicly traded companies, Gupta, Lehmann and Stuart (Journal of Marketing Research, 2004) found that a 1% improvement in retention is likely to raise customer and firm value by 3% to 7%. The companies were consumer businesses, not agencies, so the size of the effect does not transfer; the direction does.
Shorter, without hiding the work
A shorter report decides more, as long as it still shows what was done. Cutting charts is the point; cutting the record of work is a mistake.
Research on information overload, reviewed across organization science, accounting, marketing and information systems by Eppler and Mengis (2004), and by a 2023 systematic review of accounting research in Management Review Quarterly, finds that decision accuracy tends to fall once the amount of information passes a threshold, while decision time and the feeling of overload rise. The research says where the curve bends depends on the person and the task. It does not say five numbers, and neither does this page; five is a ceiling chosen for a first test.
The other half matters as much. In experiments by Buell and Norton (Management Science, 2011), people valued a service more when they could see the effort behind it, even when the result was identical. The experiments were about websites, not agencies, so applying them here is a working assumption. The practical version: one line per piece of work done this month keeps the effort visible without twelve pages of charts.
The one-page format
One page, five blocks, in this order: three sentences, up to five numbers, the work done, one decision for the client, and a link to everything else. It is a format proposed by Agency Arithmetic, not a research finding, and it is meant to be tested on one client.
The five blocks of a one-page client report, in order
1Three sentences
What changed, why it changed, and what happens next. Written by a person, read first.
2Up to five numbers
Only numbers that would change what the agency does next month, each against last month and the target.
3Work done
One line per piece of work delivered this month.
4One decision
A question the client needs to answer, with a date.
5Everything else
A link to the full dashboard. Linked, not attached.
The three sentences go first because they are the part the client reads. Tools can collect and lay out the numbers; nobody else can write these.
Example, made-up client. Numbers are illustrative.
What changed. Leads from paid search rose from 41 to 52 in May, while cost per lead fell from $39 to $38.
Why. The new landing page converts better on phones, where most of the clicks come from.
What happens next. In June, the same page goes live for the two remaining service lines.
Numbers (May / April / target)
Metric
May
April
Target
Leads from paid search
52
41
50
Cost per lead
$38
$39
$40
Booked consultations
19
15
20
Phone calls from ads
33
30
35
Ad spend
$1,976
$1,599
$2,000
Work done. New landing page for the main service line. Two ad groups paused for low quality score. Call tracking fixed on the contact page.
Decision needed by June 12. Keep the Saturday budget, or move it to weekdays, where the cost per lead is lower?
Everything else: full dashboard (link)
Six steps, one client
The format is tested on one client for one month, next to the usual report, before anything else changes. Each step fits in an afternoon.
1
Pick the client who replies least. If the format works there, it works anywhere.
2
Write the three sentences before building anything. If they are hard to write, the report has no story yet, and more charts will not give it one.
3
Choose up to five numbers. For each candidate, ask whether a change in it would change next month's work. If not, it goes to the dashboard.
4
List the work in one line per item. Deliverables, fixes, tests. No adjectives.
5
End with one decision and a date. A real choice the client can answer in a sentence.
6
Send it with the usual report for one month. Count replies and questions. Change the other clients only if this one answers more.
Where a tool helps, and where it doesn't
A reporting tool can fill blocks 2 and 5 every month without anyone pulling numbers. It cannot write block 1 or choose the decision in block 4.
If the hours spent collecting and assembling numbers are the bottleneck, the reporting time calculator shows whether a per-client tool pays for itself. If the hours go into writing, no tool changes that, and an AI summary still needs a person to read it before it goes out: people checking automated output tend to approve it without catching its errors, experts included (Parasuraman and Manzey, Human Factors, 2010).
When to keep the current format
The current format stays when the client already engages with it, or when the contract fixes it.
Keep if
If the client replied to at least three of the last four reports, the format is working. Spend the afternoon on another client.
If the contract specifies the report's contents, keep them and add the three sentences as a cover page.
Next step
Rewrite next month's report for the client who replies least, in the one-page format, and send it alongside the usual one.