METHOD
Test. Cut. Concentrate. Scale.
I don’t scale everything that moves. Every test dollar has to earn the right to be multiplied — on purchases, not clicks. Here is the decision logic, with the real numbers behind it.
- 01$5.76/d
Test
Small, recurring test budgets to compare the purchase efficiency of several ad sets in a controlled setting, before committing meaningful spend.
IN THE CLOTHING ACCOUNTMost test ad sets ran at $5.76/day.
- 025.0%
Stop waste early
A test that doesn’t buy stays small. I don’t let weak patterns average into the account: they are stopped before they absorb scale budget.
IN THE CLOTHING ACCOUNT14 zero-purchase tests: $36.99 in total, 5.0% of spend.
- 0379.7%
Concentrate
Budget goes where purchase economics justify it: a few ad sets producing most purchases at a lower CPA than the rest of the account.
IN THE CLOTHING ACCOUNTTop 3: 161 of 202 purchases (79.7%) on 67.1% of spend, $3.11 CPA vs $5.99.
- 043.8x
Scale validated winners
A validated winner gets a scale-stage budget well above the test budget — and CPA is checked as volume grows.
IN THE CLOTHING ACCOUNT67 purchases at a $2.88 CPA, $22.16/day budget, about 3.8x the test budget.
READ THE FULL FUNNEL
A click alone decides nothing.
I read CPM, CTR, landing arrival, conversion, CPA and ROAS together, in that order. Example from the supplement test: $164.76 of spend.
- 01CPM$0.86The cost of reaching the audience: it sets the room to test.
- 02Link CTR2.92%Evidence the creatives earn attention — not the final criterion.
- 03Click → landing81.9%Do paid clicks actually reach the page?
- 04Landing → purchase1.65%The conversion that closes the loop.
- 05CPA$2.17The cost of an order: the number that decides scale.
- 06ROAS10.80xAttributed value per dollar spent — read with care in COD.
CTR IS NOT ENOUGH
Same traffic, different outcomes.
Ad Set 3 had a lower CTR than Ad Set 2, but converted 29.5% better. That is the one I favor.
| METRIC | AD SET 2 | AD SET 3 | CHANGE |
|---|---|---|---|
| Spend | $137.63 | $170.74 | +24.1% |
| Link clicks | 1,716 | 1,715 | −0.1% |
| Purchases | 39 | 55 | +41.0% |
| CPA | $3.53 | $3.10 | −12.2% |
| ROAS | 8.16x | 9.37x | +14.8% |
| Purchase CVR | 2.44% | 3.16% | +29.5% |
| Link CTR | 5.41% | 4.40% | −18.7% |
| CPM | $4.34 | $4.38 | +0.9% |
WHAT THIS TELLS ME
- Nearly identical click volume1,716 vs 1,715 link clicks — traffic volume was essentially the same.
- More orders from the same trafficAd Set 3 produced 55 purchases vs 39: +41.0% purchase volume.
- Better unit economicsCPA improved 12.2% while purchase ROAS increased 14.8%.
- Post-click quality mattered moreLower CTR, but 29.5% higher purchase CVR. I favor the stronger business outcome over the vanity metric.
- More attributed valueAbout $1,600 vs $1,123 in Meta-attributed purchase value. In COD, this remains attributed order value, not collected cash.
COD REALITY
In COD, a Meta purchase is not cash collected.
In a cash-on-delivery business, a Meta “purchase” is an attributed website order, not necessarily a delivered, cash-collected sale. I separate media-buying efficiency from confirmation and delivery performance.
DECISION FRAMEWORK
Five principles, every time.
- 01
I protect test capital.
Small budgets to collect enough signal without letting weak ad sets consume meaningful spend.
- 02
I stop weak patterns instead of averaging them in.
Zero-purchase tests stay marginal; the account stays focused on validated demand.
- 03
I scale on purchase economics.
Volume is earned where CPA holds, not where clicks are plentiful.
- 04
I look beyond CTR.
The business outcome matters more than the surface engagement metric.
- 05
I understand the COD constraint.
A Meta purchase is only the acquisition layer: confirmation, delivery, refusals and collected margin complete the P&L.
Let’s talk about your next campaigns.
Startups, e-commerce and COD brands: I take over your Meta Ads account with a readable method and verifiable numbers.