Case study
Every Time They Raised the Budget, Returns Collapsed. Then They Raised It 7.5X, and Returns Went Up.
Scaling is supposed to cost efficiency. On this account it didn't. The collapse was never coming from the market. It was coming from what the algorithm couldn't see.
You have the capacity. The team could handle twice the clients tomorrow. So you buy the traffic to fill it, and every time you do, the math falls apart. CPA climbs. ROAS sags. You pull the budget back down, and the numbers recover, and you're back under the same ceiling, a little more tired than last time.
You've been told why. Audience saturation. Creative fatigue. Auction pressure. Diminishing returns, the law of paid traffic.
Here's what nobody told you.
When you scale, your first dollars spend against the platform's clearest signal. The incremental dollars, the ones you added to grow, spend against its guesses. If the platform can only see a fraction of your buyers, then more budget doesn't buy more buyers. It buys worse guesses, at a higher price, and your blended numbers report that as "scaling doesn't work."
The ceiling was never the market. It was the sample size.
This is the account that proved it. A high-ticket education business, over $100,000 a month, application funnel, setters and closers, stuck under the same monthly ceiling for a year. We rebuilt the conversion tracking on Google and Meta at the same time, so both platforms were finally being taught what the CRM already knew: which clicks became applications, which applications showed up, which calls became money.
Then the client did the thing that's supposed to be impossible. They put 7.5 times the spend into Meta.
Returns went up.
You are probably here because one of these is true of your account.
- Your ad platforms report fewer sales than your CRM books, and the gap does not close.
- Your spend has been flat for months, because nobody can prove the next increment would pay.
- You have tried to scale, and efficiency collapsed every time you did.
That is the account below, before the rebuild. Every figure on this page is from its own exports.
You have heard the tracking pitch before. So here is what this page will not do: claim a number we can't scope, count our best month, or hide what happened after the good part.
In fact, our best month on this account printed 19X. We threw it out of every average on this page. The reason is below, and you should read it before you believe anything else here.
Results are from one client's own ad-platform data and are not a guarantee of future performance; individual results vary.
Find out what your platforms can't see. That is the size of your ceiling.
Send us your busiest landing page for each traffic source you run.
We walk your funnels within one business day and tell you everything we submitted.
The external report follows within two business days, answer or no answer.
One question from us; your answer turns the report into a line-by-line trace.
No account access. No cost. No call unless you ask for one.
And if nothing is visible from outside, the report says exactly that. We would rather send you that report than invent a problem.
For accounts spending $100K a month and more, where even a small signal gap distorts a meaningful amount of spend.
Find what your platforms can't see01 THE RESULT
What the numbers did
A spreadsheet can show you where to spend. Only the signal can teach Meta who to find.
That is why this was not a channel switch anybody could have made from a spreadsheet. A monthly report of sales by source would have shown that YouTube was not converting. It would not have taught Meta which people to go and find. Feeding the platforms the CRM's truth changes what the bidding does with every dollar once it gets there. That is what made the move survivable: YouTube's spend came down by about 88 percent, the money crossed to Meta, totals stayed flat, and the channel absorbing 7.5x the spend got more efficient, not less.
Some of that lift is the platform finally being told about sales the business was already making. That is not the small print, it is the point. An ad platform bids with what it can see, and this one had been bidding on a fraction of the business for a year. Give it the whole picture and it spends your money on the people who actually buy.
What all of that was worth is the client's to say, not ours. In their own words, further down this page: their blended return across every channel went from 2-3X to 4-5X on the same money.
The client's own assessment of their blended results, from their own books. Not independently verified; not a guarantee of future performance.
02 THE PROOF
Check these numbers yourself
If you do not believe the numbers above, good. Here is what to pull on first. This is the same work we would do on your own account for nothing, so you may as well see what it looks like on somebody else's.
The month we threw out
One of the two changeover months printed 19X. That was months of purchases reaching the platform at once, the moment it could finally see them, and it is catch-up rather than a result. Neither changeover month is counted on either side of the comparison. Had we kept that month, every figure on this page would be larger and none of them would be true.
What happened after the four months
These are the four months the rebuild bought. Across the seven that followed, the account kept scaling on the same tracking and the averages came off it: cost per lead about 8 percent above where it started, ROAS about 15 percent below, at 7.5x the Meta spend. What we are accountable for is that every number they scaled on was true.
Part of that later cost is arithmetic: volume costs more per unit at the top than at the bottom. Part of it was deliberate. Once the platforms were optimizing toward the reply signal, a form submission became a more expensive thing to buy, because it was increasingly a person who would answer the phone. The client paid more per form and less per closed sale, which is the trade you run this whole system to get.
Two different numbers, and which one you are reading
The blended figure the client gives is their own, across every channel they run, out of their own books. The ROAS in the list above is one platform reporting on itself. The two do not match and they are not meant to: one counts the business, the other counts what a platform could see. Honest measurement means telling you which one you are reading.
Did they just catch a good few months?
Fair question, and it is the one we would ask. The four months are measured against the four immediately before them, not against last year, which is why the dates here are relative months. So a good season would have had to start in the same month as the rebuild, survive 7.5x the spend, and still be running ten months later.
The sheet somebody actually looked at
03 THE MECHANISM
What was rebuilt
Traffic was never the problem. The count of what that traffic turned into was wrong, and the ad platforms were bidding on the count they had.
First, the size of the gap
Nothing was changed until the gap had a number on it: how much of the business each platform could actually see, counted against what the CRM had booked. An ad account is the wrong place to measure its own blind spot, so the count started outside it.
The largest hole had no size, because nothing was leaking. The sales themselves, the calls a setter closed and the payments taken weeks after the click, reached the ad platforms not at all. There was no route for them to travel. So the platforms were bidding on a thin count of the cheapest events, none of the expensive ones, and guesswork for the rest. That is the difference between the algorithm guessing and knowing, and it decided the order of everything below.
Once the new tracking was live, that gap was counted every morning against the client's own backend, day by day, until it closed.
Then, what the platforms were told
The rebuild started with the definition of a lead. The platforms were being told about form submissions, and a form submission and a person the sales team would want to call are two different populations. Every route into the business had to report back, and report back on the same terms, before anything downstream was worth tuning. Getting every route reporting the same thing is The Complete Count, and the rest of the work sits on it.
Then we widened what the platforms hear about. The account started reporting the moments either side of the sale as well: a call attended, an email opened. A buyer takes weeks to travel from a click to money, and a platform told about the money and nothing else gets one lesson per buyer, late. Feeding it the earlier moments gives it something to learn from while the buyer is still moving.
We do not guess at those earlier moments. On this account we asked the sales team one question: which behaviour tells you, before the call ever happens, that a lead is real? They did not need to think about it. Leads who replied yes to the call confirmation text showed up 60 to 70 percent of the time. Leads who never replied almost never showed. So that reply became one of the events the platforms learn from, ahead of the sale itself. A form submission isn't a lead. A reply is a person.
The rebuild itself went in without losing a lead. Every enquiry that arrived during the changeover was still there afterwards.
Then, checking it daily
For the first month after go-live the numbers were reconciled every morning. A break would have reached somebody who could act on it the same day, while there was still only a day of spend behind it.
Last came a count against a count: what the platforms reported for a period, held up against what the CRM had booked for the same period. Keeping those two agreed, and keeping them agreed afterwards, is Parity Lock.
The first step above needs access. What we will do on your account for nothing is the outside-in version: the walk described at the top of this page, measured from where your buyers stand. This is for accounts spending $100K a month and more.
Find what your platforms can't see04 THE WITNESS
What the client said
Our tracking now is absolutely on point. We went from 2-3X ROAS to 4-5X ROAS after fixing the pixel issue.
The client calls it a pixel issue. Underneath it was the conversion-data problem across the funnel.
Results are from one client's own ad-platform data and are not a guarantee of future performance; individual results vary.
05 THE LIMIT
Where our work ends
We rebuilt the conversion tracking. The 90 days after that are what it bought: both platforms reporting what the CRM had already booked, and an account that could be scaled on numbers somebody had checked daily. What the ROAS did a year later belongs to whoever was buying the media. We do not claim it, and you should not accept it as proof from anyone who does.
That is also the honest shape of the offer. We are accountable for whether your data is true, on a timescale where that can be proven: 30 days of daily reconciliation you watch happen. Nobody can be accountable for a year of somebody else's bidding.
What that is, in order. First the audit, and it is deliberately quick: we count what each platform actually sees against what your CRM booked, network by network, and we stop counting the moment we know what to fix. You are bleeding budget while we measure, so the audit is there to aim the work, not to produce a document. Then the rebuild, per network, deployed against live traffic without losing a lead. Then 30 days of daily reconciliation you watch happen. After that it is monitoring, because tracking is not something you install, it is something you keep true.
06 YOUR ACCOUNT
What we can see without touching your accounts
Quite a lot, as it turns out. Whether the plumbing exists at all: whether a server-side endpoint is there, how thin your event coverage is, whether your forms fire anything, whether any Conversions API signal leaves at all, and whether the thing you call a lead event is really just a page view wearing its name.
What we cannot do from outside is size it. Following a handful of test leads through your funnel shows you the breaks; putting a number on how much of your business each platform is missing, across every campaign you run, needs to be inside both. That is the paid work, and it is worth talking about only after a free report has shown you something real.
And if nothing is visible from outside, we tell you that too. Your gap, if you have one, is internal. We would rather send that report than invent a problem.
07 THE SOURCE
How these figures were produced
Figures are from the client's own Meta and Google Ads exports, converted to USD from the client's local currency at a constant rate representative of the period, with dates shown as relative months rather than calendar dates.
Not ready to send anything yet? The Verified Truth Build is the same five steps this account went through, written out in full, so you can judge the work before you speak to anyone.
Find out what your platforms can't see. That is the size of your ceiling.
Send us your busiest landing page for each traffic source you run.
We walk your funnels within one business day and tell you everything we submitted.
The external report follows within two business days, answer or no answer.
One question from us; your answer turns the report into a line-by-line trace.
No account access. No cost. No call unless you ask for one.
And if nothing is visible from outside, the report says exactly that. We would rather send you that report than invent a problem.
For accounts spending $100K a month and more, where even a small signal gap distorts a meaningful amount of spend.
Find what your platforms can't see