Direct answer
Write seven rules into the contract: outcomes named in business terms from your own records; your ownership of every account and login; a written baseline before work starts; fixed definitions of lead and conversion; a quarterly reconciliation against your CRM and invoices; a fee structure you have asked about, including what the supplier earns if you spend less; and a stop rule agreed in advance. The aim is not suspicion. It is that nobody needs to be trusted.
The usual advice about agencies is to trust them less. It is poor advice, because it gives you nothing to do on Monday. The better advice is to write the contract so that trust is not what holds it together.
Most agency contracts specify deliverables and fees in detail and measurement in a sentence: "monthly reporting". That sentence hands the supplier the right to decide what is measured, where it is measured and what counts. Almost every dispute about results, a year later, starts there.
A note on fairness. This is a structural problem, not an accusation. The difference between a good agency and a poor one is rarely competence. It is mandate, access and incentive. An agency briefed on execution, reporting to a marketing manager, paid on activity and graded in its own dashboard will behave like one, however capable its people are. If you already know your position and need execution capacity, an agency is the right purchase. These rules simply make it a safer one.
Seven rules, written as clauses
| Rule | What the clause says | What it prevents |
|---|---|---|
| 1. Outcomes in business terms | Success is defined in enquiries, qualified meetings, orders or revenue, taken from the client's CRM or invoices | Success defined as clicks, impressions or "engagement" |
| 2. The client owns the accounts | Ad accounts, analytics, tag containers, domain and CRM are in the client's name; the agency has user access and the client has full read access to raw data | Losing history, and the ability to check anything, on exit |
| 3. A written baseline | Before work starts, the last twelve months of enquiries, win rate and average order are recorded, with gaps stated | Comparing results with an unknown starting point |
| 4. Fixed definitions | What counts as a lead or conversion, and the counting window, are written down; any change needs the client's written sign-off | Totals that move because a setting moved |
| 5. Quarterly reconciliation | The agency's reported results are tied to the client's own count, and differences are explained in writing | Two sets of numbers nobody has compared |
| 6. Fee structure disclosed | The contract states how the agency is paid if the client spends less, and any commission or platform rebate | An incentive nobody has looked at |
| 7. A stop rule | The result expected by a stated date, who judges it, and what happens if it is missed are agreed in advance | Contracts that only end by dispute |
Why the definitions matter more than the dashboard
Rule 4 is the one most owners skip, because definitions look like technical detail. They are where the number comes from. Take the counting window. In Google Ads, according to Google Ads Help, the default click-through conversion window is 30 days unless the advertiser sets a different one. The advertiser can change it for each conversion action, at setup or at any time afterwards. Changing it changes what the platform reports. It does not change a single customer.
None of that is improper. It is an ordinary setting. But a total that depends on a setting somebody else controls is not a measurement you can hold anyone to. The same applies to what counts as a lead: a form submission, a call of any length, a qualified meeting. Write the definition down once, and every later report is judged against it.
A result you can't tie to your own records is an opinion with a chart.
The fee clause nobody reads
Rule 6 is uncomfortable, and it is the one that matters most. Many agencies are paid a percentage of the media budget they manage. That arrangement is not wrong, and plenty of honest agencies work this way. But it pays more when you spend more, whether or not the extra spend works. Ask how the agency would be paid if you halved the budget next quarter. If the answer is "less", you have found an incentive. Name it in the contract, and agree a cap or a review point so that more spend has to be justified by evidence you can see.
Asking costs nothing. A supplier who answers plainly has told you something useful. One who finds the question insulting has told you something too.
What this looks like in practice
Working pattern, from experience — not a published study: when I am shown a marketing report and asked whether it is any good, the first thing missing is almost always a baseline. The second is a written definition of the result being reported. Without those two, nobody can say whether the numbers are good, only that they are numbers. The reports I have seen are rarely wrong in any detail. They are unanswerable.
The same gap sits behind the strategy that was paid for and changed nothing on Monday morning: a decision nobody could check, reported in a format nobody could challenge. And the argument for governing marketing with the same rigour as any other function that costs money runs through Reports, Not Revenue.
One disclosure. I sell advisory work, not execution, and I have a commercial interest in owners who want help deciding rather than doing. Weigh these rules with that in mind. They cost nothing to apply, they work whichever supplier you hire, and you can check every one of them yourself.
- Before signing, confirm the accounts will be in your name and that you can see raw data, not only a report.
- Record the last twelve months of enquiries, win rate and average order from your own records.
- Write the definitions of lead and conversion, and the counting window, into the contract.
- Diarise a quarterly reconciliation against your CRM and invoices.
- Ask how the agency is paid if you spend less, and put the answer in writing.
- Agree the result, the date and the stop rule before the first invoice.
