Agency reporting template: section by section

Friday, 6pm. The analyst exports a 22-slide report, attaches it to an email and signs off with "let me know if you have questions." On Monday the client replies with one line: "so, is this going well or not?"

That is the real problem with agency reporting. Not a shortage of data, but a pile of data with no hierarchy. The document has ad impressions, CTR by placement and a landing page heatmap, and it does not have the sentence the client needed to read in the first paragraph.

What follows is the template I use to fix that: seven sections, ordered the way clients actually read, with what belongs in each field and the mistake that shows up in each one.

What an agency reporting template is

An agency reporting template is a fixed set of sections that repeats every month, with only the numbers and the interpretation changing. The version that works has seven blocks: cover and period, the period's answer, key numbers against target, channel performance, tests and learnings, next steps with owner and date, and appendices.

Why most templates fail

The failure is almost never about design. It is about scope: the agency reports what it controls (impressions, clicks, cost per lead) and the client asks about what it pays for (revenue, new customers, margin). Until both sides look at the same number, every report turns into a negotiation.

There is also a structural problem on the client side. According to RD Station's 2026 Marketing and Sales Panorama, which surveyed 2,790 Brazilian professionals, 62% of companies do not track sales funnel conversion rates and 54% operate without a CRM. Translated into agency reality: for more than half of your clients, the bottom-of-funnel number does not exist on their side. You deliver a lead, sales closes it or does not, and nobody records the outcome.

That changes what the template needs. An honest report declares where measurement ends instead of pretending media drove all the revenue. The key numbers section below has a specific line for that.

Reading order: who reads how far

Before the sections, the logic behind their order. A report is read by three different people with three different time budgets, and almost nobody reaches the end.

The template, section by section

1. Cover and period

Fields: client name, period covered with exact dates, issue date, account owner. One status line under the title, written by you: "September: revenue 8% above target, with rising CPL on Google."

Common mistake: writing "Monthly Report" and nothing else. The cover is the one page everyone sees. Use exact dates rather than "September," because half of all reporting arguments come from window mismatches: you closed on the 30th, the client checked the dashboard on the 2nd, and the numbers disagree.

2. The period's answer

Three short paragraphs at most, answering what happened, why it happened and what you will do about it. Write it in plain language, with zero acronyms. If the month was bad, say so in the first paragraph.

Common mistake: turning this into a generic executive summary ("we continue optimizing campaigns with a focus on performance"). In practice, that burns the only section everybody reads. Prefer: "Sales dropped 12% because the retargeting campaign was paused for 9 days after a card decline. We reactivated it on the 18th and volume returned to previous levels within 4 days."

3. Key numbers against target

One table and nothing else. Five to seven rows, four columns: metric, target, actual, change against the previous period. Plus a final row declaring the limit of your measurement.

Metric	Target	Actual	vs. last month
Media spend	$40,000	$42,000	+5%
Attributed revenue	$190,000	$210,000	+11%
Orders	640	700	+9%
Average order value	$297	$300	+1%
ROAS	4.8	5.0	+4%
Contribution after media and fee	$24,000	$29,100	+21%
Measurement coverage	n/a	87% of orders with identified source	flat

Common mistake: a table with no target column. A number alone means nothing: ROAS of 5.0 is excellent for one client and a loss for another, depending on margin. If the contract has no targets, set one in the first meeting and get the client to agree, or you will spend a year reporting numbers nobody can judge.

4. Channel performance

One block per channel, with the same metrics in each so they can be compared. Meta Ads, Google Ads, organic, email, direct. In each block: spend, result, unit cost and one sentence of interpretation.

The temptation here is to dump everything the platform exports. Resist it: pick four metrics per channel and keep the same four for twelve months. Comparability is worth more than completeness. If you need a starting point, this breakdown of digital marketing KPIs by funnel stage settles the choice in fifteen minutes.

Common mistake: adding up conversions across platforms. Meta and Google each claim the same sale, so the sum comes out larger than the client's actual revenue. Label the source of every row and use a single source (the store or the CRM) for totals.

5. Tests and learnings

What you tested in the period, with result and decision. Three columns: hypothesis, result, verdict (keep, kill, scale). Two or three tests are enough.

Common mistake: only reporting tests that worked. A test that died is a deliverable: it shows budget went to learning rather than waste. It also protects you in the meeting where the client asks why CPA spiked that week.

6. Next steps

Three to five actions, each with an owner and a date. Owners include the client side: "upload the new catalog photos, Fernanda, by Oct 10." With no owner and no date, this is a wish list.

Common mistake: repeating the same next steps for three months. When that happens the bottleneck sits with the client, and the report should show what it costs: "the new landing page test has been on hold for 60 days pending legal approval."

7. Appendices

Raw data, platform screenshots, the full campaign-level table, attribution methodology. It goes last because almost nobody opens it, and it exists because when someone does, the conversation is serious.

Common mistake: treating the appendix as the report. If your document runs 22 slides, you built the appendix and forgot the report. The distinction between a narrative document and a dashboard people check continuously is covered in dashboard vs. report: the practical difference.

How the number reaches the table

Key numbers is the section that takes the most work, because it means joining sources that do not talk to each other. Map the path before automating anything.

An example with the math closed out

Simulated figures, calculated for this exercise, not survey data. Fashion ecommerce client, closed month:

• Attributed revenue: $210,000
• Product contribution margin (45%): $94,500
• Shipping and packaging (700 orders × $22): $15,400
• Contribution before media: $79,100
• Media: $42,000
• Agency fee: $8,000
• Result: $29,100

A report that says only "ROAS 5.0" supports no decision. One that shows the $29,100 supports two. First: contribution before media is 37.7% of revenue, so break-even ROAS is 1 ÷ 0.377, or 2.65. Include the agency fee and it rises to roughly 2.95. With actual ROAS at 5.0, there is room to scale spend to around double before the month turns unprofitable.

Second: the metric that deserves the headline row is not ROAS, it is contribution in dollars. That is what grew 21%. To review the underlying math, see how to calculate ROAS, including break-even ROAS.

Weekly, monthly or quarterly?

Cadence	Sections included	Time to build	Good for
Weekly	2, 3 (trimmed) and 6	15 to 25 minutes	Operations: cutting what is burning budget
Monthly	All seven	1.5 to 3 hours	Accountability and budget decisions
Quarterly	2, 3, 5 plus a trend read	3 to 5 hours	Contract renewal and planning

The most expensive cadence mistake is a full weekly report. It eats the analyst's week, the client stops reading by week three, and the decision still gets made in the monthly meeting.

The year-over-year trap

One technical detail that breaks quarterly and annual reporting: on standard Google Analytics 4 properties, event data retention offers only two options, 2 or 14 months, and high-volume properties are capped at 2 months, per Google's GA4 data retention documentation.

In practice, the comparison "this September versus last September" may simply not exist in your client's explorations, and nobody discovers that until the day they need the number. Two items belong in your onboarding: check the retention setting on day one, and keep a monthly extract of key numbers somewhere you control.

Frequently asked questions

How long should an agency report be?

Four to eight pages of narrative content, plus unlimited appendices. Length is not the problem, order is: if the conclusion sits on page one, nobody complains about a fat appendix. Reports over 15 pages with no appendix usually mean nobody decided what mattered.

What is the difference between a report and a dashboard for clients?

A dashboard answers "where are we right now" and stays available all the time. A report answers "what happened, why, and what we will do," and it has an author, a date and a point of view. Neither replaces the other: the dashboard kills status questions, the report delivers interpretation.

Should media spend appear in the report?

Always, at the exact figure and separate from the agency fee. Hiding spend makes every efficiency metric impossible to audit, and it is the single biggest source of distrust at renewal time. If the budget is variable, show planned versus spent.

How do you report a bad month without losing the client?

Put the bad result in the first paragraph, with the cause identified and the fix already in motion. What damages the relationship is not a weak month, it is the client discovering something halfway through the report that the agency knew two weeks earlier. For a diagnostic script, see sales dropped? A six-step diagnosis with data.

Is it worth automating reports?

Automate the collection and the key numbers table. Do not automate the period's answer or the next steps: that is precisely the work the client pays for. An agency that automates interpretation becomes a dashboard exporter and starts competing on price.

What the template actually fixes

A reporting template does not exist to make the agency look polished. It exists to shorten the distance between data and decision, so the monthly meeting argues about what to do next instead of where a number came from. The hard part was never the writing: it is joining Meta, Google, GA4 and the sales database into a table that reconciles, every month, for every client.

That is where Sherlok fits: connect the ad accounts, GA4, the CRM and the spreadsheets, ask in plain language "what was net contribution by channel in September versus August," and get the analysis back with every number traceable to its source. The time you save goes into the part of the report no tool writes for you: the paragraph on page one.