up from 2.81x
down from $70
against week 1
The brand
Tinge makes fabric storage furniture. Soft stain resistant drawers, light frames, 6 colours, built for people who move house more often than they buy furniture.
Naima is the premium line. 5, 6 and 10 drawer dressers, and the reason the average basket sits around two hundred dollars.
The challenge
A brand new ad account. Zero pixel data. No purchase history for Meta to learn from.
Most accounts die here. You point a purchase campaign at a cold pixel, the algorithm has no idea who buys, and you burn a month of budget teaching it the expensive way. The risk was not the spend. It was spending it on the wrong lesson.
The approach
Two phases, and we refused to skip the first one.
Phase 1, May to June. No purchase campaigns at all. We ran Add to Cart and Initiate Checkout objectives purely to feed the pixel enough real signal to work out who buys a Naima dresser. That budget bought data, not orders, and none of it is counted in the returns below.
Phase 2, June onward. Once the pixel had a profile to optimise against, we switched to purchase led campaigns. The algorithm started warm instead of guessing.
Creative strategy
Twelve plus variants into the account, tested against sales and nothing else. Not clicks, not saves, not what looked nicest in the review.
Two clear winners came out. Everything that could not carry a sale was paused on a schedule rather than left running out of politeness, which is how most accounts quietly leak half their budget.
The budget then went where the evidence pointed, and kept going there.
The results
Week one cost seventy dollars a sale. A fine start and nothing to write home about.
By week 2 the pixel training showed up. Orders more than doubled and cost per purchase halved to $36. Week 3 wobbled, as they do, when a creative fatigued and we rotated it. Week 4 the cost dropped again to $29.
By week 5 it was $23 a sale at 9.08x. 3 times the orders of week 1, on a smaller budget than week 3.
| Week | Cost per purchase | ROAS | Orders vs week 1 |
|---|---|---|---|
| Jul 3 | $70 | 2.81x | base |
| Jul 10 | $36 | 6.22x | +120% |
| Jul 17 | $42 | 3.85x | +120% |
| Jul 24 | $29 | 6.42x | +160% |
| Jul 31 | $23 | 9.08x | +220% |
Across the full purchase campaign the account returned 5.11x. Counting the pixel training phase that came before it, the all in return is 3.4x. We show you both, because one of those is the number that actually left the bank.
Spend and revenue figures belong to the client, so we do not publish them. Cost per purchase and return on ad spend are ours to show, and they are the numbers that tell you whether we are any good.
What's next
The budget roughly doubles from here. Two new creatives are queued, a carousel format and a university audience that fits a light, portable dresser better than anything we have run so far.
Doubling a budget usually costs you some efficiency. The target is to hold cost per purchase under $30 while it happens. The curve will not stay this steep, and the plan does not assume it will. The account is still climbing.