The short answer
A monthly report has four jobs. It has to let the firm reproduce each figure in its own accounts, compare it with last month on the same definition, tell what the agency did apart from what followed, and record the date the numbers were read. A UK professional services firm should require those four in writing.
Key takeaways
- The useful question is not which metrics the report carries. It is which of the four jobs the report format allows at all.
- The National Audit Office and the Treasury set criteria for a performance measure. Verifiable is one of them, and a screenshot is not verifiable.
- Google Ads counts a conversion on the date of the click, so last month's figure rises for weeks after the report was sent.
- Analytics withholds rows to protect identity and folds rows past a limit into one called "(other)". Neither shows in a picture of a chart.
- The cheapest repair is a clause naming the column a figure came from and the date it was read.
The criteria for a performance report were settled in the UK years ago, by auditors. Choosing the right FABRIC, a joint publication of the National Audit Office, HM Treasury, the Cabinet Office, the Audit Commission and the Office for National Statistics, sets out what a single measure has to be before anybody acts on it. Among its criteria: relevant, attributable, comparable with earlier periods, and verifiable, with documentation behind it so the measure can be validated.
Hold a marketing agency's monthly report against that. Most fail on verifiable, because a figure inside an image cannot be traced back to the account that produced it. The framework binds public bodies rather than suppliers, and no agency is in breach of it. It is useful here because it is the nearest thing to a specification of a report not written by somebody selling marketing.
The format decides what the report can do, before the metrics do
Three formats cover almost every monthly report a UK professional services firm receives, and the most common one answers none of the four jobs. That is a property of the format, which is why changing the metrics inside it does not help.
| What the report must let you do | Deck of screenshots | Platform export | Live dashboard link |
|---|---|---|---|
| Reproduce the figure yourself | No. An image names neither the account nor the column it came from. | Partly. The date range travels with it. The conversion action does not. | Yes, where it reads your account and the filters are visible. |
| Compare it with last month | No. Nothing records whether the definition changed. | Only where the same export ran both months. | Partly. A live view restates history, so last month moves. |
| Tell activity from outcome | No. The work sits in the commentary, the result in the picture. | No. A platform reports its own surface and credits it. | No, for the same reason. |
| Know when the numbers were read | Rarely. The timestamp is cropped away with the screen. | Yes. An export carries the date it ran. | It is always now, so not the figure you were shown. |
| Verdict | No format answers all four. The export and the link each answer one cleanly and the deck answers none, so what is worth requiring is a link for checking and a dated note for the record. | ||
Last month's figure is still moving
Google Ads calculates its primary conversion columns on the time of the click, not the time of the conversion. Google's documentation on understanding your conversion tracking data gives the example: an ad clicked last week whose traffic converts this week has both reported back to last week. The reason is sound, because the cost sits on the click date too.
The consequence is the part nobody puts in the report. A September figure read on 1 October is not the September figure: it keeps rising while late conversions land against September clicks, and the version in the deck is whichever reading the agency happened to take. Google publishes separate columns reporting by conversion time, so the choice exists and somebody made it, which is also why advertising figures and analytics figures disagree by design.
Two Analytics behaviours remove rows before the report is made
Analytics withholds data from a report when showing it could let somebody infer who an individual user is. Google describes these as data thresholds. They are system defined and cannot be adjusted, and they apply where a report includes demographic information or audiences built from it. When they bite, the interface says so at the top of the report. A screenshot of the chart underneath does not.
The second behaviour is cardinality. A dimension with more than 500 unique values in a day is high cardinality, and where a report's dimensions exceed its row limit, everything past the limit is condensed into one row labelled "(other)". A report carrying a secondary dimension or a filter is the likeliest to hit it.
Neither behaviour is an agency hiding anything, and both mean a total and the rows beneath it can disagree.
Before the next report lands
Not sure whether the reporting you get could be checked?
Free to your firm. The agency pays us, and only if the relationship works, which is why we would rather ask how a supplier's numbers are produced than place one that has never been asked.
Get a recommendationThe reporting clause, in five lines
Reporting is cheaper to fix in the engagement letter than in a review meeting nine months later. None of the five below asks for extra work. Each asks for something the agency already has.
Put these in writing before the first report
- Name the accounts. Which advertising and analytics accounts the report was read from, with the firm holding owner access to each.
- Name the column. Not "conversions" but the conversion action, and whether it counts on the click date or the conversion date.
- Name both dates. The period covered and the date the numbers were read. The first does not fix the second.
- Split the two halves. What was done, with dates. What followed, with the source of each number beside it.
- State what a bad month looks like. Agreed in advance, in the firm's units rather than the platform's: enquiries, or fee income where it is tracked.
The fifth line is the one that gets resisted and the one worth holding, because a target set after the fact is not a target. What the report has to show is decided by what the spend was meant to buy, a question settled when the budget was set rather than when the invoice arrives.
What belongs to the firm, and what no report can reach
The outcome column is the firm's to supply, and it is the only column answering the question the partners are actually asking. No advertising platform knows whether an enquiry became a client. That lives in the practice, and handing an agency a login to find it is a separate decision, worked through in whether an agency should have access to your practice systems. A dated export of the enquiry date and what happened to it closes the gap without naming a client, and settles the attribution argument before it starts.
When to require less, and when not to hire at all
A monthly report is the wrong cadence for most of this market.
A practice taking a handful of enquiries a month cannot read a change in a month of data, and asking for a monthly document produces what a monthly document is for: commentary. A quarterly note against the firm's own figures, with the account open in between, is better evidence and less work.
There is also a case for hiring nobody yet. Where nobody at a UK firm can say what a good month looks like in enquiries or fee income, no report will settle it, and the agency gets marked against whatever it chooses to measure. That is worth an hour of partner time before a supplier is appointed.
Saying so costs us the introduction fee, because we are paid by the agency and only when a relationship works. What we examine before recommending anybody is set out for accountancy practices, recruitment agencies, tax advisory and financial advice firms.
Frequently asked questions
What partners ask when the monthly report arrives.
What should a marketing agency report every month?
Four things, in a form the firm can check: which accounts the figures were read from, what each figure is counting, what work was done and when, and what would count as a bad month. Anything else is commentary, and commentary is not the report.
Why does last month's conversion number keep changing?
Because Google Ads counts a conversion on the date of the click rather than the date it happened. A click in September that becomes an enquiry in October is reported back to September, so a September figure read on 1 October is lower than the same figure a fortnight later.
Is a live dashboard link better than a monthly PDF?
For reproducing a figure, yes, because it reads the account rather than a picture of it. For comparing months it is worse, because a live view restates history as definitions change. A UK firm wants both: the link, and a dated note for the record.
Our agency says the platform numbers are the only numbers. Is that right?
No. A platform reports what happened on its own surface and cannot tell anybody whether an enquiry became a client. That comes from the practice, which is why the outcome column belongs to the firm.
Can we ask for the raw account access instead of a report?
Ask for both. Access lets a partner reproduce a figure, and the report records what the agency says it means, which is the thing to hold them to later. Access to an analytics account is a different request from access to the systems holding client work.
Sources and useful reading
- National Audit Office, Choosing the right FABRIC, with HM Treasury, the Cabinet Office, the Audit Commission and the Office for National Statistics, and the framework itself, for the criteria a measure has to meet.
- Google Ads Help, understand your conversion tracking data, for the primary conversion columns being calculated on the click time, and for the by-conversion-time columns.
- Google Ads Help, data discrepancies, for why advertising and analytics figures differ.
- Google Analytics Help, about data thresholds, for thresholds being system defined and applied where a report includes demographic information.
- Google Analytics Help, cardinality, for the 500 values a day definition and for rows past the limit becoming "(other)".
This article is commercial decision support and is not legal advice. The National Audit Office framework applies to public bodies and is used here as a specification of what a performance measure has to be, not as a duty on any supplier or firm. Each platform behaviour is documented by the platform that operates it and is changed without notice, so read the current page. All sources were located on 5 October 2026. Nothing here describes any named agency or engagement.