Proving ROI to a client who has decided not to believe you

There is a meeting every agency knows. The reporting is thorough, the dashboards are current, cost per lead is down twenty percent, and the client's sales director says, without much interest in the slides, that the leads are rubbish. The room agrees with whoever is more senior. Nothing in the agency's reporting can settle it, because everything in the agency's reporting is measured on the agency's side of the funnel.

You cannot win an argument about lead quality using data that stops at the lead. The evidence you need is on the other side of a boundary most agencies never cross.

This article is about crossing it: how to get the revenue data, what to do with it, how to handle the case where it makes you look worse than your reports did, and why doing this is better for the agency even when the first answer is unflattering.

Why dashboards do not persuade

Platform dashboards report the agency's own activity measured by the agency's own suppliers. However accurate, they are structurally the account of an interested party, and a sceptical client is entitled to discount them.

This is not a criticism of agency reporting, most of which is honest and competent. It is a statement about where the evidence comes from. When the number originates in a system the agency operates, using conversions the agency configured, attributed by a platform that also sells the media, no amount of detail converts it into independent verification.

Adding more detail actively hurts at this point. A client who does not believe the headline will not be persuaded by twelve additional charts, and the volume of material tends to read as advocacy rather than analysis. Longer reports are a common agency response to declining trust and one of the least effective.

What changes the dynamic is a number the client's own finance function can verify against its own accounts. That is a different category of evidence, and it is available to any agency willing to ask for two exports.

The reason agencies do not ask

Most agencies never ask for closed-sales data because the answer might be worse than what they have been reporting. That is a rational fear and it is also the thing keeping the relationship fragile.

It is worth being honest about this, because the reluctance is rarely stated aloud. If your reports claim four hundred conversions and the client's revenue data shows that ninety of them became customers, the conversation that follows is uncomfortable. Avoiding it is understandable.

The trouble is that the discomfort does not disappear; it accumulates. Every quarter of reporting on unverified metrics increases the gap between what the client believes and what is true, and the eventual reckoning is worse than the one available today. Accounts are usually lost to that accumulated gap rather than to a bad quarter.

There is also a competitive angle. The agency that proposes revenue-based measurement is making a claim about its own confidence that competitors pitching on lead volume cannot match. In pitches this reads as unusually secure, because it is.

How to ask, so that you get it

Ask for a narrow, defined export rather than system access: a row reference, a phone or email, an amount, and a close date. Specificity is what gets this approved, because it is a request the client's operations team can fulfil in an hour.

Requests for CRM access fail far more often, and they should. They are large, vague, involve a security review, and require somebody to take responsibility for what an external party might see. A four-column export over a defined date range is a completely different ask and often does not need to leave the operations team at all.

1. Name the four fields and the date range in writing, and say explicitly what you do not need.
2. Explain what the client gets: a revenue figure they can verify, not another agency chart.
3. Offer to sign whatever data agreement they require before receiving anything.
4. Commit in advance to report whatever the data shows, including if it is worse than current reporting.
5. Propose a first run on historical data, so nothing about live performance is at stake in the first attempt.

That last point removes most of the remaining resistance. A backward-looking analysis of last year cannot embarrass anybody about this quarter, and it produces the maturity curve and match rate you need before the arrangement becomes routine.

What to do with the first result

Expect the matched revenue to be lower than platform-reported conversions implied, expect a meaningful unattributable share, and present both without hedging. The credibility comes from the candour more than from the number.

The first run almost always shows a smaller contribution than the platform reports suggested, because platform reporting counts influence and the match counts traceable outcomes. Explaining that difference clearly, before the client asks, converts what could be an accusation into a methodological point.

It is worth preparing the client for the shape of the answer beforehand. Saying in advance that the matched figure will be lower than the platform figure, and why, means the result confirms your explanation rather than contradicting your previous reporting.

CloseRev's agency plan exists for this: run the match on the client's own export, and produce a read-only report the client can forward to their finance director with your mark on it rather than ours.

When the data says you did badly

If the matched revenue is genuinely poor, bring it to the client first, with a diagnosis and a plan. An agency that discovers and reports its own underperformance is in a far stronger position than one whose client discovers it.

This is counterintuitive and it is repeatedly borne out. Clients rarely terminate over a bad quarter honestly reported with a credible remedy. They terminate over the sense that they were being managed, and nothing produces that sense faster than a problem surfacing from their side of the table.

The diagnosis matters as much as the disclosure. Poor matched revenue has a small number of common causes — targeting too broad, offer attracting the wrong segment, leads not followed up, routing sending enquiries nowhere — and at least two of those are frequently on the client's side. Presenting the analysis neutrally lets that emerge without an accusation.

It is also worth agreeing in advance what you will do if the answer is bad. An agency that has already said it will bring unflattering results and then does exactly that has demonstrated something about itself that no case study can.

The lead follow-up problem, handled carefully

Matched revenue frequently reveals that a large share of leads were never contacted, which is a client-side failure that has been damaging the agency's apparent performance. Raising it is necessary and requires care.

The finding is common enough to expect. Enquiries that arrive outside working hours, land in a shared inbox, or route to a branch that is short-staffed simply never get worked, and they appear in every agency report as leads delivered and in every client conversation as leads that did not convert.

The way to raise it is with the number and without the inference. Present the uncontacted count and the close rate of contacted versus uncontacted enquiries, and let the client draw the conclusion. An agency that arrives with data saying your sales team is the problem will be resisted regardless of whether it is right.

Structuring the reporting so it survives a new stakeholder

Report matched revenue, coverage, and the method in the same format every month, so the reporting survives a change of marketing director. Most agency relationships end shortly after the client-side sponsor changes.

A new stakeholder arrives with no history, an instinct to review incumbent suppliers, and no reason to extend credit. What protects the relationship at that moment is a reporting record that is legible without context: consistent format, verifiable numbers, stated method, visible trend.

What does not protect it is a strong personal relationship with the departing sponsor, or a body of reporting that requires explanation to interpret. Both are common and neither survives the handover.

The data protection question, answered properly

Receiving a client's customer identifiers makes the agency a processor acting on the client's instructions, which requires a written agreement, defined retention, and a deletion commitment. Handling this well is itself a selling point.

Many agencies are nervous about this and the nervousness is disproportionate. The obligations are well defined and unremarkable: process only on instruction, keep it secure, do not transfer it onward without permission, delete it when asked or when the relationship ends, and assist the client if a data subject exercises a right. None of that requires a legal department to implement.

What it does require is that you have thought about it before the conversation. A client's operations team asked to send customer phone numbers to an external agency will escalate, and the escalation is answered in ten minutes if you already have a processing agreement, a retention period and an answer about where the data will live.

The narrow export helps here too. An agency asking for four columns over a defined window, with a stated deletion date, presents a much smaller decision than one asking for CRM access, and it is a decision the client's own privacy position is improved by.

What this changes about the pitch

Proposing measurement on the prospect's own revenue data differentiates a pitch more sharply than creative or case studies, because almost no competitor will offer it and the prospect immediately understands why.

The reason it lands is that every prospect has been burned by exactly the dynamic this article opens with. They have sat in the meeting where the agency showed green dashboards and the sales director said the leads were poor, and they have no way to adjudicate it. An agency that opens by proposing a method for settling that argument is speaking directly to a known scar.

It also filters prospects usefully. A prospect who declines revenue-based measurement has told you something important about how the relationship will be evaluated, and you can price and staff accordingly rather than discovering it in month nine.

Be prepared for the obvious challenge, which is whether you will still propose this when the numbers are unflattering. The answer that works is a commitment in the contract to report matched revenue whatever it shows, which costs nothing if you meant it.

The first ninety days of doing this

Run the historical match first, agree the method and the coverage figure with the client, then move to a monthly cadence. Three months is enough to make it routine and to have the first difficult conversation on your own terms.

Month one is the backward-looking analysis on a year of history, which establishes the match rate, the maturity curve and the composition of the unattributed bucket without anybody's current performance being at stake. It is also where the data quality problems surface, and they always surface.

Month two is agreement: the method written down, the coverage figure accepted, the format of the monthly report fixed, and the treatment of unattributed revenue settled. Getting this in writing while the mood is collaborative is considerably easier than after a disputed quarter.

Month three is the first live report in the agreed format, and from there it is routine. The whole exercise costs a few days of work spread across a quarter, and it changes the basis of the relationship from assertion to evidence permanently.

Performance pricing, if you must

Pricing on attributed revenue is workable only with a jointly agreed method, a stated coverage figure, an agreed treatment of the unattributed bucket, and a dispute process written before the first invoice.

The failure mode is predictable: the agency's compensation depends on a number, the number has judgement in it, and the incentive to interpret the judgement favourably exists on one side. Even scrupulous agencies end up in disputes under these arrangements, because the structure invites them.

If you do it, the essential clause is who runs the measurement and how disagreements are resolved. Measurement run by the party being paid on the result is a conflict; measurement run by the client with a defined method the agency has agreed to is workable.

One further protection is worth writing in: agree that the measurement continues unchanged for a defined notice period after either party gives notice. Disputes about performance cluster around terminations, and a method that can be renegotiated at exactly the moment it matters most is not a method. Fixing it in advance costs nothing while relations are good and is impossible to agree once they are not.

The agencies that survive client-side leadership changes are the ones whose reporting the client could verify without them. Build the reporting you would want to inherit.

Questions people actually ask

How can an agency prove ROI to a sceptical client?

By reporting on the client's own closed-sales data rather than on platform-reported conversions. Matching the client's closed revenue against the leads you generated produces a number the client's own finance function can verify, which is the only kind that ends the argument.

How do you ask a client for their sales data?

Directly, early, and with a specific field list: an identifier, an amount, a close date, and a row reference. Asking for a defined narrow export is far more likely to succeed than asking for CRM access, and it is also better practice on the client's side.

What if the client refuses to share revenue data?

Then they have chosen to be reported to on platform metrics, and it is worth saying so plainly and without resentment. Some clients will change their mind once the limitation is stated as their decision rather than your failure.

What if matched revenue shows worse results than we reported?

Present it anyway, first, before anybody else finds it. An agency that brings unflattering data voluntarily is treated very differently from one whose client discovers it, and the difference is usually the account.

Does reporting on revenue put an agency at more risk?

It transfers risk from a vague argument you cannot win to a specific number you can influence. Agencies judged on lead volume are always one sales director's opinion away from losing the account; agencies judged on matched revenue have evidence.

Should agencies charge on attributed revenue?

Only with a jointly agreed measurement method, a stated coverage figure and a defined dispute process. Performance pricing on a number one party controls is a dispute waiting to happen, whichever party that is.

Found this useful?

• Share on X
• Share on LinkedIn
• Share on Reddit
• Share on Email