How to Judge Ads by Deals Not Clicks: Two Ledgers
How to judge ads by deals not clicks: keep your own ledger of real people, follow each lead into the pipeline and divide spend by closed deals, not conversions.
To judge ads by deals not clicks, keep a second ledger of your own: count the distinct people who arrived from your ad links, follow each lead into the pipeline, and divide spend by closed deals rather than by the conversions the platform reports. Wait for the sales cycle before you judge on revenue.
This is where most small company advertising goes wrong, even when everything before it was done well. The campaign runs, the platform reports results and those results are taken at face value. The platform's numbers are not lies. They are its own account of its own performance, measured its own way.
How to judge ads by deals not clicks: the two ledgers
The platform counts conversions. You should count people. The two will disagree, and neither is wrong, because they answer different questions. The platform's question is how many events it can connect to an ad. Yours is how many distinct people arrived and what they became.
| Ledger | What it counts | Who it serves |
|---|---|---|
| The platform's | Every conversion event it can connect to an ad, which may include one person who filled in two forms, or someone who only saw the ad and converted later by another route | The platform's own picture of its performance |
| Yours | The distinct people who arrived from the campaign's ad links and became a contact, counted once each | The decision about the budget |
Divide spend by real people, not by the platform's count, and keep the platform's figure beside it for reference. Yours is the one that pays the bills.
Why a cheap lead can be the most expensive one
Two campaigns can produce leads at a similar cost and wildly different value. One fills the pipeline with people who book calls and buy. The other fills it with people who download the free thing and vanish. A cheap lead that never becomes a conversation is not cheap.
The ad manager cannot show you this difference, because it stops looking at the form. Only following each lead to a qualified opportunity, a closed deal and a deal value can. This is also why the offer matters so much: it decides who responds, as discussed in how to write a paid ads offer.
The four questions to ask of every campaign
Judge a channel on how many, at what rate and at what value. For ads, add a fourth question, about cost.
| Question | What it asks |
|---|---|
| How many? | How many real people did the campaign produce? |
| At what rate? | What share of them became conversations and then closed deals? |
| At what value? | What were the deals worth? |
| At what cost per closed deal? | How does the spend per closed deal compare with the ceiling you set before launch? |
The fourth question is the one the platform can never answer for you. It needs your spend, your pipeline and your ceiling in the same place. The ceiling is worked out in how to set an ad cost ceiling.
What the second ledger looks like
You do not need special software to start. A spreadsheet and a careful person who reconciles it every week is enough. The point is to hold the platform's figures and your own side by side, one row per campaign.
| Column | Where it comes from |
|---|---|
| Spend | The platform |
| The platform's conversions | The platform |
| Real people | Your own contact records, counted once each, tied to the campaign by its identifiers |
| Cost per real lead | Spend divided by real people |
| Where each lead is now | Your pipeline: contacted, qualified, won or lost |
| Deal value | Your pipeline |
| Cost per closed deal | Spend divided by closed deals |
The left side of the table is what the platform told you. The right side is what actually happened. Reading across a row, you can see exactly where the story changes. A campaign with many conversions and few real people has been counting generously. A campaign with many real people and few closed deals has been attracting the wrong ones.
Leads that you cannot tie to any campaign should be listed as unmatched, not given to the nearest likely campaign. A guess dressed as a figure is worse than a gap, because it looks like knowledge.
Respect the lag
A lead from an ad today may close weeks or months from now. Judge a campaign on revenue too early and every ad looks like a failure. Wait indefinitely and you never kill anything.
So find your typical time from first contact to closed deal, and judge ad cohorts on revenue only once they have had that long. Decide the wait in advance, so it does not become a reason to postpone every decision.
Until the cohort is old enough, steer by the furthest down milestone that reliably predicts revenue, and call it a proxy. A booked call might be that milestone for one company and a qualified opportunity for another. Say which it is, and say it is a stand in.
Be suspicious of ads that take credit for demand they did not create
Some formats are excellent at appearing in front of people who were going to buy anyway. Examples are retargeting visitors to your pricing page, bidding on your own brand name and advertising to your existing list. They report attractive results because they are standing at the till when the purchase happens.
Some of that is real. A well timed reminder does close deals. Some of it is credit borrowed from whatever made those people interested in the first place.
The honest test is to pause the campaign occasionally and see how much of the result disappears with it. If very little does, the ad was a toll booth on a road people were already driving. Retargeting depends on the warm audience described in how to build ad audiences from your customers, so the same care applies there.
Compare paid with everything else
A paid lead should not be judged in isolation. Put the deal value produced by paid sources beside the deal value from your other sources. Ads that look good against their own ceiling can still lose to a quieter channel that costs far less, and you can only see that by comparing them.
Ask three comparative questions each time you review: how much deal value did paid sources produce, how does that compare with other sources, and is any campaign mostly reaching people who were already buying? The answers decide where the next unit of budget goes.
Scale only what passed, and in steps
When a campaign clears its threshold on deals, the temptation is to multiply the budget. A bigger budget reaches people further from your ideal customer, so the cost of each result tends to climb as spend does. Raise the budget one step at a time, give every step time to be read, and keep the stopping rule in force throughout.
Each increase is itself a test that can fail, so treat growth as a ladder of tests rather than one leap. The method for running each step is in how to run a small ad test.
Write the decision down
Every review should end with one paragraph: what you decided and why. Keeping notes on what you tried, what it cost and what came out of it builds the one advertising asset the platform can never see.
- Spend, the platform's conversions and your real people, side by side.
- Cost per real lead, against the ceiling.
- Where each lead from last month is now: contacted, qualified, won or lost.
- Whether older cohorts have had long enough to judge on revenue.
- The decision, and the reason for it.
What this changes about the budget
The platform's dashboard tells you what the platform did. Your pipeline tells you what the platform was worth. Only one of them is allowed to decide the budget.
That is a practical rule as much as a principle. When the two disagree, go with the pipeline. When the pipeline is incomplete, say so and use the best proxy you have, labelled as one. The overview in what is paid advertising for founders explains why the platform is an amplifier and not a judge.
Frequently asked questions
Why does the platform count more conversions than I have contacts?
It may count one person twice, for two forms, or credit someone who only saw the ad. Your own count of distinct people is narrower and more useful.
How long should I wait before judging on revenue?
As long as your typical time from first contact to closed deal. Until then, use the furthest down milestone that reliably predicts revenue and label it a proxy.
What if an ad looks great but a pause changes nothing?
Then it was probably a toll booth on a road people were already driving. Keep any real value, but do not scale it as if it created demand.
Do clicks tell me anything at all?
They tell you the ad was noticed. They say little about whether anyone became a customer.
The full method, with the checklists, is in The Paid Ads Playbook, a free 31-page playbook.