Most monthly marketing reports are not dishonest. They are worse than dishonest, in a specific and useful way: they are true and irrelevant. Impressions really did rise eighteen percent. Engagement really is up. The keyword really did move from position fourteen to position nine. Every number in the deck can be correct and the document can still fail to answer the only question the person paying for it has, which is whether more work came in and whether the money that produced it was well spent. Reading a report properly is mostly a matter of sorting its numbers into two piles and noticing which pile the summary slide was built from.
Two piles: activity and outcome
An activity metric describes something that was done or delivered. An outcome metric describes something that happened to your business. Activity metrics are not worthless — they are exactly how you diagnose a problem once you know one exists, and an agency that cannot show you clicks and impressions cannot troubleshoot. The danger is structural rather than moral: activity is easy to move, easy to present, and improves under effort even when the effort is aimed at the wrong thing. A report led by activity can look better every month for a year while the phone rings the same number of times.
The ladder below is worth keeping next to the report. Each row proves something real and fails to prove something specific, and the failure is what the row above it is for.
| Metric | What it proves | What it does not |
|---|---|---|
| Impressions | Your ads were served. Delivery is happening. | That anyone looked, or that the people served were plausible buyers. |
| Reach and frequency | How widely and how often the message landed. | That the message was right, or that reach was the constraint. |
| Clicks and click-through rate | The ad interested somebody enough to act on it. | That the interest survived the landing page. |
| Engagement, likes, shares | The creative provoked a reaction on the platform. | Almost anything commercial. This is the most decorative row here. |
| Sessions and pageviews | Traffic arrived and browsed. | That the traffic was human, in market, or in your service area. |
| Form submissions | Somebody filled something in. | That it was a real person, a qualified one, or one you could serve. |
| Calls over a duration threshold | A conversation happened, long enough to be a conversation. | That it was an enquiry rather than a supplier, a wrong number or a recruiter. |
| Qualified leads in the CRM | A named opportunity exists that your team accepted. | That it will close, or what it will be worth. |
| Booked jobs and revenue | Money changed hands. This is the row the business runs on. | Nothing you should be arguing with, once the source is recorded. |
The test is not whether the low rungs appear. It is where they appear. Outcomes belong on page one and activity belongs in the appendix, because that ordering forces the agency to lead with the number it has least control over, which is precisely the discipline the arrangement is supposed to buy.
The blank source field
Open your CRM, take the last hundred inbound leads, and look at the column that says where each one came from. In a great many businesses that column is empty, or it says "website" for everything, or it holds whatever the receptionist typed when they remembered to type something. That single blank column silently invalidates most of what the report says, because every channel claim downstream of it is now unfalsifiable. The agency says paid search produced forty leads. Your CRM has a hundred leads and no idea which forty.
The blank field is not usually laziness. It is the visible end of a chain that broke somewhere between the ad and the record, and the mechanics of that break are worth understanding in their own right — click identifiers stripped by a redirect, a form handler that never passes the hidden fields, a call that arrives on a number nobody is tracking. That chain is the subject of a companion piece on why leads lose their source. For the purposes of reading a report, the only thing you need is the question: what percentage of the leads in my own system carry a recorded source, and what does the agency do with the ones that do not?
A channel number that cannot be checked against your own records is not a measurement. It is a claim.
Platform conversions against CRM reality
Put the platform-reported conversions for a month next to the leads your team actually recorded for the same month, and they will disagree. That is normal, and an agency that pretends otherwise is either inexperienced or managing you. The disagreement has at least six legitimate causes, and knowing them lets you tell a legitimate gap from a broken one.
- Different clocks. Ad platforms credit a conversion to the date of the ad interaction, not the date the sale happened. A sale in June can appear in the May column.
- Different models. Each platform assigns credit its own way, and each is counting from inside its own walls without any view of the others.
- Double counting across platforms. Two platforms that both touched the same buyer will both claim them, so adding the platforms together overstates the total, sometimes badly.
- Modelling. Part of the platform figure is an estimate of conversions it could not observe directly, and the proportion moves month to month.
- Identity. One person on a phone and a laptop may be two users to a platform and one record in your CRM, or the reverse.
- Qualification. A conversion is a form submission. A lead in your system is a form submission that someone accepted. Those are different populations by definition.
So the question is never "why do they not match". It is "how far apart are they, is the gap stable, and can you explain its direction". A gap that holds steady at fifteen percent and points the same way every month is a measurement characteristic you can work around. A gap that was ten percent in March, sixty percent in April and inverted in May is a defect. The common practitioner threshold is that divergence beyond roughly twenty percent is worth treating as a fault rather than as modelling noise.
20 percent threshold: common industry rule of thumb, not a CurrentAds benchmark
View-through, modelled, and the windows around both
Three accounting rules quietly decide how large the headline number is, and none of them is visible on the slide unless somebody chooses to show it.
A view-through conversion is credited to an ad that was served and never clicked. The logic is not absurd; people are influenced by things they do not click. But it is the weakest evidence in the document, because it will just as happily credit an impression served to somebody who was already walking through your door. The rule is simple: view-through belongs on its own line, never blended into the same total as click-through, and never used on its own to justify more budget. If your report does not separate them, that is the first thing to ask for, and the answer tells you a lot about the report.
A modelled conversion is a statistical estimate of a conversion the platform believes occurred but could not observe, typically because of consent choices or browser restrictions on cross-site measurement. It is disclosed, legitimate, and now standard. What matters for your purposes is that part of the number is a count and part of it is a forecast, and the ratio is not stable. That is the strongest argument there is for reconciling against your own system, which is the one ledger in the stack that does not estimate anything.
An attribution window is how long after an interaction a conversion still counts. A thirty day click window credits a sale on day twenty-nine; a seven day window does not, and the business did not change between those two reports — only the accounting rule did. Windows are configurable, they differ between platforms by default, and a month-on-month comparison across a window change is not a comparison at all. Ask for the window in writing, and ask to be told inside the report on any month it is altered. An agency that changes the window and reports the improvement without mentioning it has not lied, and has still misled you.
Nine questions to ask your current agency
Send these as a list. The tone of the reply is nearly as informative as the content: a good operator will enjoy most of them, because they are the questions they wish more clients asked.
- 01 Of the leads in our CRM last month, what percentage carried a recorded source, and where did the rest go?
- 02 Show me the platform-reported conversions next to our own lead records for the same month, and name the gap.
- 03 Which of the conversions in this report are view-through, and which are modelled?
- 04 What attribution window is each platform set to, and has any of them changed in the last six months?
- 05 Which conversion action is the bidding actually optimising toward, and when did we last confirm it still fires?
- 06 How many of the leads you reported were marked unqualified by our team, and does that flow back into the bidding?
- 07 If we added the platforms together, how much of the total is the same person counted twice?
- 08 What did you change in the account last month, and what did you expect it to do?
- 09 What is currently not working, and what would you stop doing if the budget fell by a third?
The last one is the tell. Every account has something in it that is not working. An agency that cannot name theirs is either not looking or not saying, and both produce the same report.
What a reconciled number looks like
A reconciled number has four properties. It comes from your system rather than from an ad platform. Every record behind it carries a source. It has been placed next to the platform figure with the difference named out loud. And the definition behind it — what counts as a lead, what counts as qualified — is written down somewhere both sides can point at. That is the whole standard, and most reporting fails the first two.
The single published case study on this site is a measurement engagement of exactly that shape, and it needs its disclosure stated before its numbers, not after. It is an engagement with a company under common ownership with CurrentAds, published anonymised for that reason. No revenue increase is attributable to the work and none is claimed. CurrentAds has no paying clients yet. Read it as a first-party build rather than an arm's-length client result.
With that on the table: before the work, no inbound lead in that business carried a recorded source at all. After it, all 48 inbound leads in the measured period carried one. That is not a marketing result and it is not presented as one — nothing about it says the advertising made money, and the disclosure above is the reason it can be published honestly at all. What it is, is the difference between a company that can answer question one on the list above and a company that cannot. In the same engagement the paid social channel was measured properly for the first time and returned nothing on the money that had gone into it, which is why that spend was stopped instead of scaled. A number that lets you stop something is worth more than a number that lets you celebrate something. The full figures and their sources are on the results page, with the same disclosure attached.
The four-part report
If you are rewriting the brief for whoever reports to you next, the whole specification fits in four headings. Outcomes for the month, from your records. Reconciliation, with the gap to the platforms named and explained. What changed in the account and why, in sentences rather than screenshots. What happens next, and which decision each number on the previous three pages is supposed to inform. Everything else is appendix, and the appendix is allowed to be long.
Getting to that report is not a reporting project, which is the part most people find annoying. It is a measurement project — sources captured at the ad, carried through the form, landed in the CRM, and fed back to the platforms as qualified outcomes rather than raw submissions. Until that exists, every claim about paid media performance is a claim about a platform's opinion of itself. The order to build it in is set out in the measurement floor, and the commitments we attach to it are on the guarantee page.