How to Read an Agency Performance Report
A framework for reading agency performance reports critically. What the numbers actually mean, what gets glossed over, and the five questions to ask every month.
By The Spend Report Editorial Team. Published June 7, 2026. · 6 min read
On this page
Most operators read agency performance reports the same way they read airline safety cards. The numbers move past, the cover sheet looks fine, and the meeting moves on. The result is that bad reports go unchallenged for months, and the agency learns that the bar for reporting is low.
This is the framework we use to read a paid ads agency report critically. What the headline numbers actually mean, what the report tends to skip, and the five questions to ask every month so the agency knows the standard is high.
The report has three jobs
A useful agency report does three things, in order:
- Tells you the state of the account in plain language.
- Explains what happened since the last report and why.
- States what the agency is doing next, and what evidence supports that decision.
That's it. Reports that fail one of those jobs tend to fail the other two.
The most common failure mode is reports that are mostly visualization. Charts everywhere, screenshots of platform UIs, occasional emoji. No narrative. The agency is showing you what they did. They are not telling you what it means.
The headline numbers, decoded
Every report leads with a few summary metrics. Read them in this order, and treat them as the start of a conversation, not the conclusion.
Spend. Up or down month over month. If spend changed materially, the agency owes you a one-sentence reason. "We scaled what was working" is not a reason. "We scaled the campaign targeting cold lookalikes from the 3 percent audience, because conversion rate held at $X CPA over a $Y window" is a reason.
ROAS or CAC. Pick the one your business runs on. The one the agency leads with is usually the one that flatters them. If they always lead with ROAS, ask for CAC. If they always lead with CAC, ask for blended ROAS. The agency that responds with both numbers and is fluent in why each tells a different story is the agency you can trust on numbers.
Revenue. Attributed by the agency, sometimes inflated by their attribution model. Compare to your Shopify or back-end revenue. If the agency-attributed revenue is consistently 20 percent higher than your back-end shows, the attribution model is over-counting and the conversation needs to happen.
Conversion rate. A composite metric. Could be moving because of creative, audience, landing page, offer, seasonality, or inventory mix. The agency should know which of those moved.
AOV. Often ignored. Movement in AOV usually means a product mix shift or a promo, and it changes the math on every other metric. The agency that does not mention an AOV move that is happening is not reading their own report carefully.
What the report tends to skip
These are the omissions that recur most often. Notice when you do not see them. Ask about each.
Wasted spend. The campaigns that did not work and were cut. The report should name them with the dollar amount and what was learned. Reports that only show wins are showing the highlight reel.
Comparable-period analysis. A 30-day window matters more in context. Show me the same 30-day window from last year, last quarter, and 60 days ago. Reports that only compare to "last period" are gaming the comparison.
Statistical significance. The agency reports a 12 percent CPA improvement. Cool. Is that real, or did 18 conversions happen to land favorably? Reports that treat every metric movement as causal are reports written without statistical care.
Creative performance breakdown. Which specific ads worked? Which formats? Which messages? Reports that lump creative into "we tested new ads" are hiding either weak testing structure or weak results.
Channel cannibalization. When paid spend goes up and organic revenue goes down by the same amount, that is not growth, it is reshuffling. Reports that count incremental paid revenue without checking for organic decline are over-counting their contribution.
Account-level health. Things like account quality score, learning phase status, policy violations, and platform-side warnings. These are not glamorous, but they predict next month's performance.
The five questions to ask every month
Read the report, take notes, then run these five questions through it. The agency that answers all five well is doing real work. The agency that struggles with two or more is producing a sales document, not a performance report.
1. What is the one thing this month that you want me to focus on?
Forces narrative over volume. If the agency cannot pick the one most important thing, they are not reading their own report.
2. What happened that you did not expect?
Surprises are where learning lives. An agency that has no surprises is an agency that has not actually engaged with the data. Their answer should be a specific surface, a specific number, and a hypothesis about why.
3. What did you stop doing this month, and why?
The default question to ask. Reports skew toward what was added. The decision to stop something usually matters more than the decision to start something, and a healthy report covers both.
4. If you had 50 percent more budget next month, what would you spend it on, and what evidence supports that?
Tests whether the agency has a real next-move thesis. The answer "we would scale what is working" is the wrong answer. The right answer names a specific audience, a specific creative angle, or a specific channel, and points to a number that justifies the bet.
5. If you had 30 percent less budget next month, what would you cut first?
The reverse pressure test. The agency that cannot answer this is an agency that has not done the prioritization work. Their answer reveals their actual ranking of what matters in the account.
How to give feedback on reports
When a report falls short, do not just sit through the meeting and follow up later. Push back live. The conversation should look like this:
- Name a specific section that did not work.
- Explain what would have been useful instead.
- Ask the agency to update the template before next month.
Most agencies will adapt. The good ones welcome the feedback because most clients never give it. The bad ones get defensive, which is information.
If after two months of explicit feedback the report is still missing the same elements, the issue is not the template. It is what the agency can produce. Read the red flags article for the patterns that say it is time to start interviewing replacements.
The reporting baseline contract clause
The best version of this is solved at contract signing. Specify in the contract:
- Weekly written summary by Monday at 10am ET.
- Monthly performance review on the first business day of the month.
- A standard template that includes spend, CAC, ROAS, revenue (agency-attributed and platform-reported), conversion rate, AOV, top campaigns, paused campaigns, creative performance breakdown, and one written narrative section per channel.
- A 90-day report at the close of each quarter that covers strategic direction, not just numbers.
Agencies that resist documenting reporting in the contract are agencies whose reporting will drift over time. The clause is the operational expression of what professionalism looks like.
For the rest of the operating relationship, the pillar guide covers the full arc. For the first quarter specifically, first 90 days with a new paid ads agency is the companion piece.