Three accounts, measured
What it cost.What it returned.
Three accounts we run, across the three engines. Every figure below is pulled from the ad platform over our own window on that account, not the account's full history. Each one ends with what it does not prove, because a number without its limits is just a screenshot.
Engine three, tickets
Two clubs, one season,2,123 seats at $1.97.
The situation
Two clubs running full season schedules over the same months. Neither had run paid acquisition against a fixture list before. A season is not one campaign, it is dozens of separate sell-through problems, and the dates that need help are never the ones that sell themselves.
What we ran
- Weighted spend to the soft midweek dates instead of spreading it evenly across the schedule
- Worked the existing season ticket base separately from cold prospects, so the two were never bidding against each other
- Wired ticketing platform tracking so a sale could be attributed back to the click that caused it
- Reconciled every figure below to the ticketing platform, not to platform-reported conversions
| Club A | Club B | Combined | |
|---|---|---|---|
| Spend | $2,361 | $1,821 | $4,182 |
| Tickets sold | 1,358 | 765 | 2,123 |
| Cost per ticket | $1.74 | $2.38 | $1.97 |
| Tracked revenue | $83,738 | $41,077 | $124,815 |
| Return | 35.5x | 22.6x | 29.9x |
| Month | Spend | Tickets | Revenue |
|---|---|---|---|
| Month 1 | $1,326 | 599 | $36,325 |
| Month 2 | $2,449 | 1,244 | $72,651 |
| Month 3 | $407 | 280 | $15,840 |
Spend falls away in month three as the schedule closes out. The tickets do not fall at the same rate, because by then the campaigns are selling the last soft dates rather than the whole season.
What this does not prove
- This is one season. It is not evidence that the result repeats across a second season, and season two is where a ticketing account is actually tested.
- Both clubs sit in similar markets and ran the same months, so this is one set of conditions, not a general claim about live events.
- A 29.9x return is unusually high and is a function of ticket price against a low cost per acquisition. Do not expect this shape in a category with a longer consideration cycle.
Engine one, lead generation
Thirteen months,1,373 leads at $14.
The situation
A specialty retailer in a small market a long way from our home base. High consideration purchases, an installation step that caps how much work can be taken on, and a customer base drawn from a wide radius of surrounding towns. They were already running search advertising and it worked. The question was whether paid social could add volume without wrecking the cost per lead.
What we ran
- Started on a deliberately small monthly cap and did not raise it until the cost per lead held for a full quarter
- Put a tracked number on every ad and landing page, so phone calls attributed instead of disappearing
- Built the CRM pipeline through to quoted and sold, so a lead could be followed to the thing that pays
- Rebuilt creative against the live offer each season rather than running one evergreen set
| Launch quarter | Latest full quarter | Change | |
|---|---|---|---|
| Spend per month | $1,098 | $1,718 | +56% |
| Leads per month | 92 | 131 | +42% |
| Cost per lead | $11.89 | $13.15 | +11% |
| Quarter | Spend | Leads | Cost per lead |
|---|---|---|---|
| Launch | $3,293 | 277 | $11.89 |
| Deep winter | $4,570 | 230 | $19.87 |
| Spring | $5,150 | 364 | $14.15 |
| Latest full | $5,153 | 392 | $13.15 |
Read the middle row, not the headline. Cost per lead rose 67% in the deep winter quarter and we held spend through it rather than pulling back. It recovered over the following two quarters without a rebuild. Over the full run, spend per month is up 56% and cost per lead is up 11%, which is the trade we were making on purpose: more volume at a slightly higher unit cost, in a market with a hard ceiling on how many people are in it.
What this does not prove
- Cost per lead is not cost per job. This account tracks 36 attributed purchases worth $80,188, which is real but is a fraction of what the business actually closed from these leads. We are reporting the input, not the outcome.
- Search advertising was already running and working throughout. We did not run a holdout, so some share of this volume would likely have arrived anyway.
- One account in one category. A thirteen month run is long enough to show seasonality, not long enough to claim a repeatable benchmark. Your market will not behave like this one.
Engine two, ecommerce
Spend tripled.So did the return.
The situation
A direct to consumer brand in a category the platforms restrict, where much of the standard playbook is unavailable and a large share of creative never clears review. The account was small, inconsistent, and had never been scaled. The usual outcome when you push spend in a category like this is that the return collapses.
What we ran
- Built creative specifically to pass review in a restricted category, rather than writing ads and appealing rejections
- Sent conversions server side, so the platform was optimising on the fuller picture rather than what survived ad blockers
- Raised spend in steps, holding each level until the return held, instead of stepping up on a schedule
- Worked average order value alongside acquisition cost rather than treating them separately
| Month | Spend | Orders | Cost per order | Return |
|---|---|---|---|---|
| Month 1 | $736 | 16 | $46.00 | 2.04x |
| Month 2 | $1,632 | 57 | $28.63 | 4.17x |
| Month 3 | $2,150 | 78 | $27.56 | 4.27x |
| Month 4 | $2,306 | 121 | $19.06 | 6.76x |
| Across the four months | Start | End | Change |
|---|---|---|---|
| Monthly spend | $736 | $2,306 | +213% |
| Cost per order | $46.00 | $19.06 | -59% |
| Average order value | $93.94 | $128.78 | +37% |
| Return | 2.04x | 6.76x | +231% |
The average order value moving is the part that matters. Acquisition cost falling on its own usually means you found the cheap buyers. Acquisition cost falling while order value rises means the offer is working, not just the targeting.
What this does not prove
- Four months is a ramp, not a track record. The hardest test in ecommerce is holding a return through a full year including the post-holiday trough, and this account has not seen one.
- This ad account was later actioned at the platform level and the campaign was rebuilt on a new one. That is a standing risk in a restricted category and it is planned for, but it means the run above is not continuous to today.
- Figures are platform attributed with server side conversions. The brand's own store reporting will not match exactly, and where the two disagree we treat the store as the truth.