A client once told us their previous agency's reports had shown improvement every month for two years. Revenue over the same period was flat. Both things were true, which tells you something about what was being reported.
Reporting is where the agency relationship is either honest or not, and it is unusually easy to evaluate once you know what to look for.
Metrics designed to always improve
Some numbers rise almost regardless of performance. Cumulative keyword counts. Impressions. Social reach. Total backlinks. Emails sent. Any metric where more activity mechanically produces a larger number is a poor measure of whether the activity worked.
- 'We now rank for 4,200 keywords' — for what, and does anyone search them with intent to buy?
- 'Impressions up 60%' — did anyone click, and did clicking lead anywhere?
- 'Reach grew 3x' — reach among whom, and what did they do next?
- 'We built 40 links' — from where, and did anything move?
What a report should contain
- Performance against the goal set at the start of the period, including when it was missed.
- The business outcome — pipeline, revenue, qualified leads, cost per acquisition — not only channel activity.
- What was actually done during the period, in enough detail to evaluate.
- What was learned, particularly what did not work and what that implies.
- What changes next period and why.
- Anything the agency needs from the client to make progress.
That last point matters more than it seems. Programs stall on client-side blockers — approvals, access, subject-matter time — as often as on agency performance. A report that never mentions them is either not tracking dependencies or avoiding an awkward conversation.
Attribution honesty
Every channel platform reports its own contribution generously. If you add up what Google, Meta and your email platform each claim, you will exceed your actual revenue — sometimes substantially. An agency reporting platform-attributed numbers without acknowledging this is either unaware or hoping you are.
Good reporting acknowledges attribution uncertainty explicitly, uses blended metrics alongside channel-level ones, and periodically validates with holdout tests. It is less flattering and considerably more useful.
Questions worth asking your agency
- What did we try this quarter that did not work?
- If our budget was cut 30%, what would you stop doing first and why?
- What is the biggest risk to this program that is not on the report?
- Which of our competitors is doing this better than us, and what are they doing?
- What would you do differently if you were starting this engagement today?
The answers are informative regardless of content. An agency that cannot name a failed experiment either is not running experiments or is not telling you about them.
The most useful monthly report we send is usually the one explaining why something did not work.
That report is uncomfortable to write and it is the reason clients stay. Nobody believes a program that only ever succeeds, and reporting that never surfaces a problem removes the client's ability to help solve one.