Three vendors, one sale, and everybody has a screenshot

The meeting goes like this. The search agency reports four hundred conversions. The social agency reports three hundred and ten. The affiliate platform claims ninety. Email claims a hundred and forty. The business closed five hundred and twenty deals. Everybody has a screenshot, everybody is confident, and somebody senior is about to conclude that at least one of these vendors is inflating their numbers.

When your vendors' conversion claims sum to twice your actual sales, that is not evidence of dishonesty. It is evidence that you asked four parties to count the same thing independently.

This is the most predictable dispute in marketing and one of the most damaging, because it discredits measurement generally and it usually ends with the budget going to whoever presents most confidently. It is entirely resolvable, and the resolution is procedural rather than technical.

Why the overlap is structural

Every platform answers the question did I influence this sale, within its own window, using only its own data. A customer touched by four channels generates four affirmative answers, and no participant is in a position to know about the other three.

The important word is influence. Platforms are not claiming sole causation; they are reporting that a person they served an advertisement to subsequently bought something. That is a true statement, and it is true simultaneously for every platform that reached the same person.

The windows compound it. A view-through window of a week or a month means a platform can claim a conversion from an impression the customer never consciously registered. Different vendors use different windows, so the same journey produces different claims depending on who is counting.

So the sum exceeding reality is not a bug to be fixed at the vendor level. It is the necessary consequence of independent counting, and expecting the vendors to solve it among themselves is expecting competitors to agree on how to divide credit for revenue they are each being paid to generate.

The size of the problem

In a business running several paid channels, combined vendor claims commonly reach one and a half to three times actual sales. The ratio grows with the number of channels and the length of the attribution windows in use.

The four hundred and twenty phantom conversions are not fabricated. They are real sales counted more than once. What makes this dangerous is that each vendor's report is internally consistent and independently defensible, so the error only becomes visible when somebody adds them together — which, remarkably often, nobody does.

It becomes actively expensive when budget is allocated on those claims, because the channels with the longest windows and the broadest view-through counting will claim the most, regardless of what they contributed. The reporting method rather than the performance determines the allocation.

The one number that can adjudicate

Your own sales system counts each sale exactly once. Matching those sales to your own source records produces a channel table that sums to reality by construction, and it is the only such table in the room.

This is the entire resolution and it is unglamorous. There is no clever reconciliation algorithm, no way to fairly deduplicate four vendors' claims against each other, and no neutral arbiter among the platforms. There is only the business's own record of what it actually sold, joined to its own record of where enquiries came from.

The table it produces will disappoint everybody slightly, which is a reasonable sign that it is not favouring anyone. Every vendor will find their contribution smaller than they reported. That is arithmetically inevitable when a total of one hundred and eighty-one percent is compressed into one hundred.

This is the specific job CloseRev does: reconcile closed sales against your own lead and call records, so the channel table sums to the revenue in your accounts rather than to the sum of your vendors' opinions.

How to run the reconciliation meeting

Circulate the method before the numbers, present your own table first, and ask each vendor to explain the gap rather than to defend their figure. The framing determines whether this becomes a diagnosis or a fight.

Sending the method in advance matters more than anything else in this list. A vendor who learns in the room that their reported number is about to be contradicted will respond defensively, and reasonably so. A vendor who knew a week earlier what would be compared and how usually arrives with a useful explanation.

The question to ask is not why is your number wrong. It is what does your number include that mine does not. That is answerable, it is often illuminating, and it keeps the conversation on definitions where the real disagreement lives.

It is also worth saying explicitly at the start that you do not think anybody is inflating their figures. It costs nothing, it is almost always true, and it prevents the defensive crouch that otherwise consumes the first twenty minutes.

What the gaps usually turn out to be

Most of the difference resolves into four causes: view-through counting, window length, conversions that were not sales, and sales your matching could not trace. Only the last is a problem with your own measurement.

• View-through counting. The platform counted an impression nobody clicked. Legitimate to report, rarely appropriate to pay for at full value.
• Window length. A thirty-day window catches journeys a seven-day window does not. Neither is wrong; they are different questions.
• Non-sale conversions. Newsletter signups, downloads and chat sessions counted as conversions. Often a configuration choice nobody revisited.
• Unmatched revenue. Sales your own matching could not trace to any source, which the vendor may well have genuinely generated.

That fourth line is the one to handle with care, because it is where a vendor has a legitimate case. If a third of your revenue is unmatched, a vendor claiming conversions you cannot see may be right, and dismissing them on the strength of your own incomplete data is not sound.

Fix it in the contract, not in the meeting

Specify in the contract which measure the vendor reports, computed from which data source, over which window, and on what cadence. Most reporting disputes are definitional and disappear when the definition is contractual.

The clause does not need to be elaborate. Name the metric, name the source of truth, name the window, and state that performance will be assessed against the client's own matched revenue table. Vendors who are comfortable with that are usually the ones worth keeping.

Expect some resistance, and distinguish between two kinds. Objecting that your matching is incomplete is a fair technical point that deserves an answer. Objecting to being measured on your data at all is a commercial position, and it tells you what the relationship is going to be like.

The affiliate and last-touch problem

Channels that intercept demand at the final step — affiliates, coupon sites, branded search — will always claim well in a last-touch world, whether or not they created the demand they are capturing.

This is the most politically difficult version of the problem, because the claim is technically correct. The customer did click that link before buying. What is unresolved is whether the customer would have bought anyway, which no amount of tracking can answer.

The only reliable tool here is incrementality testing: switch the activity off in a comparable market or period and observe what happens to revenue. It is disruptive, it costs money, and it is the sole method that distinguishes creating demand from intercepting it.

Businesses that run this test are frequently surprised, in both directions. The important discipline is deciding in advance what result would change the spend, because the temptation to explain away an inconvenient outcome is considerable.

What to do if you have no matched data yet

Even without a full match, dividing actual sales by claimed conversions gives every vendor a haircut factor that makes the reports comparable. It is crude, it is better than nothing, and it takes an hour.

If vendors claim nine hundred and forty conversions and you closed five hundred and twenty, every vendor's claim is running at about one point eight times reality on average. Applying that uniformly is certainly wrong in detail — the overlap is not evenly distributed — and it is a great deal less wrong than treating the raw claims as additive.

Treat it as a stopgap with a deadline rather than a solution. It stops the worst allocation errors while you build the real reconciliation, and it demonstrates the size of the problem persuasively enough to get that work funded.

When the vendors are all inside one company

Platform families that own several ad products deduplicate internally and present a single figure, which removes the visible overlap without removing the underlying question of whether the claim is fair.

This is an improvement in tidiness and not necessarily in truth. A single consolidated number from a large platform is still that platform's estimate of its own contribution, computed under its own rules, and the fact that it no longer visibly collides with a sibling product does not make it independent.

It is also harder to interrogate. When two vendors disagree you at least have a discrepancy to investigate; when one vendor reports one number you have nothing to compare it against except your own data, which is exactly why having your own data matters more as the market consolidates.

The practical guidance does not change. Use the consolidated figure to optimise inside the platform, where it is genuinely the best available signal, and use your own matched revenue to decide how much that platform should get relative to everything else.

The internal version of the same fight

The same double counting happens between internal teams. Brand, demand generation, field marketing and sales development will each claim the same closed deal, using whichever internal report supports the claim.

It is less visible than the vendor version because it rarely gets added up, and it is more corrosive because the participants have to keep working together afterwards. The pattern is familiar: two teams present at a quarterly review, both showing healthy contribution, and the totals quietly exceed what the business actually sold.

The fix is the same and easier to implement, since you control both sides. One matched revenue table, produced centrally, used by every team, with contribution reported against it rather than against each team's own instrumentation. Teams may still argue about the allocation, but they will be arguing about the division of a real number rather than about which of two inflated numbers is correct.

Where teams have genuinely different roles in the same journey, resist the urge to split credit precisely. Reporting that a deal was influenced by brand and closed through demand generation is more useful and more honest than assigning sixty and forty percent on a rule nobody can defend.

What a good vendor report looks like

A useful vendor report states delivery and cost precisely, states its own attribution claim with the window and method named, and does not present that claim as revenue. Vendors who volunteer the method are usually the ones whose numbers survive scrutiny.

It is worth noticing which vendors do this unprompted. A report that says conversions, seven-day click and one-day view, platform-attributed is telling you exactly what it is and inviting comparison. A report that says revenue generated with no qualification is presenting an estimate as a fact, and the omission is a choice.

The second useful signal is how a vendor responds when their number is smaller in your reconciliation than in their report. The good response is curiosity about the gap and a specific hypothesis. The poor one is a general argument that attribution is complicated, which is true and is not an answer.

None of this requires an adversarial posture. Most agencies would prefer to be measured on something defensible, because it protects them too — when the client's own data shows the contribution, the vendor no longer has to be believed on trust every quarter.

The cultural change that matters most

Stop asking vendors to report on revenue and start asking them to report on what they control. Impressions, clicks, cost, reach and delivery are theirs. Revenue attribution belongs to you.

This division is cleaner than it sounds and removes most of the conflict at the root. A vendor reporting delivery metrics is reporting facts about their own execution, which they are well placed to know and have no structural reason to distort. A vendor reporting revenue is estimating their own value, which is a different activity entirely.

It also improves the relationship. Vendors generally dislike the current arrangement too — being asked to prove their own worth using tools that everyone knows are self-interested is an uncomfortable position, and most agency people would rather be judged on a number the client produced.

Ask your vendors what they did. Work out for yourself what it was worth. Almost every reporting dispute in marketing comes from asking one party to do both.

Questions people actually ask

Why do my ad platforms report more conversions than I actually had?

Because each platform counts any conversion it believes it influenced within its own attribution window, and none of them can see the others. A customer who saw a social ad, clicked a search ad and opened an email will be counted once by each. The totals overlapping is arithmetic, not misreporting.

How much overlap should I expect between platform reports?

It varies with channel mix and window length, but a combined platform total of one and a half to three times actual sales is common in businesses running several paid channels. If your platforms sum to roughly your real revenue, that is worth investigating too — it usually means something is not tracking.

Which vendor's number should I believe?

None of them, for reporting purposes. Each is a reasonable answer to the question did I influence this, asked by a party with an interest in the answer. Use them to optimise inside each platform and use your own closed-sales data to decide budgets between them.

How do I reconcile agency reports against actual revenue?

Match your closed sales to your own source records, produce one channel table, and compare each vendor's claim against it. Do it before the meeting and share the method in advance, so the discussion is about the difference rather than about whose spreadsheet is right.

Should I make agencies report on a single attribution model?

Yes, and specify it in the contract along with the data source. Most reporting disputes are definitional rather than factual, and they disappear the moment everybody is required to report the same measure computed the same way.

Does deduplicating conversions across platforms solve this?

It helps within a platform family but not across independent vendors, because deduplication requires a shared identifier and a shared view of the journey. The reliable deduplication point is your own sales record, which by definition counts each sale once.

Found this useful?

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