Found A$18,342 a month of spend that was buying nothing, inside an account reporting 4.5x.
The account was not underperforming. It was misreporting. Once the attribution window was corrected, more than half the budget turned out to be sitting below break even.
The headline export: Premium Supps Meta audit, 30 days to 6 April 2026.
Shown in full and uncropped when supplied.
The account was spending A$31,488 a month across six active Meta campaigns and reporting a blended return of roughly 4.5x. On paper it was one of the healthier supplement accounts I have looked at.
Shopify disagreed. The order count did not match, the gap had been widening for months, and nobody could explain it. The working theory inside the business was that Shopify was undercounting. So budget kept moving toward the campaigns Facebook said were winning, and one of them had been renamed Scaling Winners.
That name turned out to be the most expensive thing in the account.
Measurement. Every ad set was running Meta’s default seven day click plus one day view.
That setting gives the platform credit for a sale when somebody merely saw the ad and then bought anywhere within twenty four hours. Existing customers, email subscribers and direct traffic all get counted as paid acquisition.
Rebuilding the account on seven day click only changed the picture completely. Of 1,844 reported purchases, 840 were real. The other 1,004 were view-through credit for sales that were going to happen anyway. Fifty four per cent of the account’s reported performance did not exist.
That single change reordered the entire account. Three campaigns reporting 3.04x, 2.63x and 3.01x were actually returning 1.03x, 1.08x and 1.29x. Together they were consuming A$18,343 a month, fifty eight per cent of the budget, to roughly break even. Meanwhile one catalogue campaign reporting 18.06x was genuinely returning 7.27x, and it was the only thing holding the account’s blended number up.
The second finding sat underneath the first. The best ad set in the account, catalogue remarketing at 9.98x real return and A$11 per purchase, was running at a frequency of 8.31. It was the most valuable asset in the business and it was two to three weeks from exhausting its audience. Nothing in the reporting flagged it, because a campaign at 9.98x does not look like a campaign in trouble.
What made this hard was not finding it. It was what had to be true before anything could change.
The client's own theory pointed the other way
The business believed Shopify was undercounting, not that Meta was over-reporting. Budget had been moving toward the campaigns Facebook said were winning for months on that assumption. The first job was not analysis, it was arguing against a working theory the team already held — which only evidence wins, not opinion.
The worst campaign was called Scaling Winners
Recommending a pause on something the team had branded a winner needs numbers that leave no room for a second reading. Three campaigns reporting 3.04x, 2.63x and 3.01x were actually returning 1.03x, 1.08x and 1.29x, and I had to show the same figure two ways before the decision could be made.
Fixing the reporting makes performance look worse overnight
Switching to seven day click only meant every reported figure in the account would drop, immediately and visibly. Unless that is framed before the change rather than explained after it, a correction reads as a collapse and gets reversed within a fortnight.
The best asset could not absorb the freed budget
The obvious home for the released A$18,343 was the 9.98x catalogue remarketing ad set. But it was running at a frequency of 8.31 and two to three weeks from exhausting its audience, so pushing spend into it would have accelerated the one thing holding the account up.
Six domains, in order. Each one is a decision, not a task list.
Strategy
Reset the definition of a winning campaign from reported return to return on click, and agreed a 3.0x floor with the business before touching anything.
Acquisition
Paused the three campaigns returning between 1.03x and 1.29x, releasing A$18,343 a month. Held the creatine campaign at 2.0x rather than cutting it, because the product was clearly converting and the problem was the ads, not the offer.
Creative
Briefed a refresh for the burnt remarketing pool rather than the top of funnel, which is the opposite of the usual instinct. The top of funnel ad set at 3.86x did not need new creative. It needed budget.
CRO
Left the store alone. With attribution this distorted, no landing page test would have produced a readable result until the measurement was fixed.
Analytics
Switched every ad set from seven day click plus one day view to seven day click only, permanently aligning Meta reporting with Shopify. This is the change that makes every future decision in the account trustworthy.
Scale
Reallocated the released budget into the two ad sets that were genuinely working, and widened the remarketing audience before scaling it, so the spend increase did not accelerate the burnout.
monthly spend released from below break even
phantom purchases removed from reporting
real return, established against a reported 4.5x
All three are established findings, not projections. The reallocation target that follows them is labelled separately below.
| Before | After the reallocation | |
|---|---|---|
| Monthly spend | A$31,488 | A$31,488 |
| Real purchases | 840 | Measured against a 3.5x blended target |
| Real blended return | 2.1x | 3.5x target |
| Spend below break even | A$18,342 | Nil |
| Reporting accuracy | 54% of sales unverifiable | Meta and Shopify aligned |
The right hand column is the plan, not the result. This was a diagnostic engagement and the reallocation was handed to the client’s team to execute. The post-implementation figures go here once the next thirty day period closes.
Source: Premium Supps Meta audit, 30 days to 6 April 2026. Platform metrics sit here as support, not as the headline.
The account had been running on default attribution for months before anyone questioned it, and I found it on day one of an audit rather than in a routine check. That is the wrong way round.
I now read the attribution window in the first hour of any engagement, before I look at a single creative or campaign structure. Everything downstream of a broken measurement setting is guesswork, including my own judgement about which creative is working. Checking it takes ninety seconds. Not checking it cost this account A$18,000 a month for an unknown number of months.
of budget below break even once measured on click
return on click against reported return
reported purchases never recorded in Shopify
Measurement, with frequency as the second order problem
Meta, six active campaigns
Diagnostic and plan
A diagnosis, not a pitch.
Bring the account, the numbers and the problem as you understand it. You leave knowing where the broken link is, whether or not we work together.
A read on where your growth ceiling actually sits, across all six parts of the system.
Which of frequency, measurement or structure is costing you the most right now.
What fixing it would take. Scope, sequence and who does what.
A straight answer on whether I am the right person to own it.
Thirty minutes, held on Google Meet. The link is in the calendar invitation.