In short
From May to July 2026, Meta reported $1,215 in purchase value for this campaign. The GoHighLevel funnel recorded $2,476 in sales over the same months. Meta was only receiving the front-end purchase and never saw the upsell. On real revenue, average revenue per buyer covered 97% of July's cost per buyer. The account was still not profitable, and I say so below.
The situation
The brand sold an online wellness program through a GoHighLevel funnel: sales page, order form, a one-time upsell, and a thank-you page. Traffic came from a single Meta campaign. Inside Ads Manager, it looked like a clear loss.
So I compared Meta's purchase data with the order records in the funnel itself.
What the numbers showed
| Month | Spend | Purchases | Cost per purchase | Meta ROAS |
|---|---|---|---|---|
| May | $119.93 | 2 | $59.97 | 0.45 |
| June | $2,216.80 | 16 | $138.55 | 0.19 |
| July | $1,532.78 | 27 | $56.77 | 0.48 |
| May to July | $3,869.51 | 45 | $85.99 | 0.31 |
Meta counted 45 purchases worth $1,215. The funnel counted the same 45 front-end buyers, and 13 of them also bought the upsell. Total funnel sales for May to July came to $2,476.
The purchase event was firing on the front-end order only. The upsell, the most valuable step in the funnel, never reached Meta. Every optimization decision the algorithm made was based on about half the real revenue.
The funnel itself
| Step | Unique visitors | Orders | Order rate |
|---|---|---|---|
| Sales page | 1,656 | 9 | 0.54% |
| Order page | 107 | 36 | 33.64% |
| Upsell | 58 | 13 | 22.41% |
Once people reached the order page, one in three bought, and more than one in five buyers who saw the upsell took it. The leak was higher up: most ad traffic landed on the sales page, and very few visitors moved on from there.
The July turnaround
Across May to July I tested 103 ad variations in named concept families: data-led ads, myth-busting ads, metaphor ads, comparison ads, and tutorials. July was the month the numbers turned.
- Cost per purchase fell from $138.55 in June to $56.77 in July, a 59% drop.
- Purchases rose from 16 to 27, up 69%, while spend fell 31%.
- The best ad, a myth-busting concept, bought customers at $26.99 each, 69% below the $86.59 blended cost at ad level.
That last point was also a warning. The myth-busting ad received $161.93 of budget. The top-volume data-led ad received $1,953.13 at $84.92 per purchase. Budget was following volume, and the cheaper winner stayed underfunded.
What real revenue changes
Average funnel revenue per buyer over May to July was $55.02 ($2,476 across 45 buyers). July's cost per buyer was $56.77. On Meta's view the account looked hopeless. At that average, each July buyer brought in about 97% of what it cost to acquire them.
Two fixes follow directly from this. First, send the upsell purchase back to Meta with its value, so the algorithm optimizes toward full revenue. Second, move budget toward the lowest-cost concepts instead of the highest-volume ones.
The honest result
The account did not reach profit. May to July spend was $3,869.51 against $2,476 in funnel sales, a true return of 0.64. I include this project for the diagnosis and the July turnaround, both of which a new client can check line by line in the tables above.
The client's name and offer are withheld. Figures come from Meta Ads Manager and GoHighLevel funnel stats, pulled September 2026. Monthly totals and the ad-level window (May 1 to July 30) differ slightly.
Published 2026-09-27 by Clint Balbontin.