How We Took Cost Per Sale From $14 to $5.50 in 90 Days

I want to write this one down carefully, because "we cut CPS by 60%" is the kind of line that appears on a lot of slides and means nothing without the mechanics. This is the 90-day sequence at a multi-country DTC brand where I owned marketing and, soon after, the whole revenue stack. Cost per sale went from $14 to $5.50. Average order value went from $19 to $34. The company reached net profit for the first time.

Nothing in it is a trick. Most of it is arithmetic that nobody had been allowed to do.

Day 0: the numbers nobody agreed on

Marketing reported cost per sale from the ad platforms. Finance reported it from the bank. The gap between them was large enough that every budget conversation turned into an argument about measurement, and the argument always ended with "let's keep spending and see".

The first two weeks went into one thing: a definition of a sale that finance would sign, and a report that produced it automatically from the shop, the payment provider and the ad accounts. That report is the reason everything else was possible, and it is the step most teams skip because it is not exciting.

Weeks 3–4: stop paying for orders that lose money

With honest attribution, we could finally see cost per sale by product and by country. The picture was uncomfortable. Some products could not be profitable at any acquisition cost once shipping and returns were counted. Some countries had delivery costs that ate the entire margin.

We cut them. Fewer SKUs, fewer countries. Revenue dipped for three weeks and margin rose immediately, because we had stopped buying customers we could not afford. Cost per sale fell from $14 to roughly $9 on this step alone, before touching a single ad.

Weeks 5–8: rebuild the offer before rebuilding the ads

An average order of $19 cannot carry international shipping, payment fees and paid acquisition. Rather than push discounts, we rebuilt what we were selling:

• Bundles designed around fulfilment. Products that shipped together, in one box, at one weight tier. The customer saw a better offer; we saw a cheaper order to serve.
• Price points set from margin backwards. Each bundle had a floor price that cleared fulfilment and acquisition with room to spare. No bundle went live below its floor.
• Marketplaces stopped undercutting the store. Our own marketplace listings had been competing with our site on price. Aligning them removed a leak we had been funding ourselves.

Average order value went from $19 to $34 over these weeks. Because a higher order value makes the same acquisition cost cheaper per dollar of revenue, cost per sale kept falling even where bids did not change.

Weeks 9–12: now the marketing work

Only now did the performance work start, and it started with a real target: every campaign had a maximum cost per sale derived from the margin of what it sold. Creative and bidding were rebuilt around the bundles rather than single products, and the channels that only looked good in their own dashboards were cut. This took cost per sale from around $9 to $5.50.

The order of these steps matters more than any individual step. Done in reverse — ads first, offer second, measurement last — the same team had spent a year and a lot of money going nowhere.

The weekly rhythm that kept it there

Every Monday, one report, five numbers: cost per sale, contribution margin per order, average order value, fulfilment lead time, returns. One owner per number. The meeting was thirty minutes because the report was automated and nobody had to defend their spreadsheet. When a number moved, someone already knew why, because they had been the one to move it.

What I would tell a founder starting this today

1. Get one definition of a sale and one automated report before anything else. If marketing and finance disagree, you have no numbers.
2. Cut before you optimise. Unprofitable products and regions do not become profitable with better ads.
3. Fix the order before the funnel. AOV is a pricing and packaging decision; the checkout upsell is the last five percent.
4. Give every campaign a cost ceiling derived from margin. Then let the performance team be as clever as they like inside it.
5. Review weekly, thirty minutes, five numbers. Consistency beats intensity here.

Three mistakes that would have undone it

Every one of these came up, and each would have quietly reversed the gains.

• Reporting ROAS to the board again. Once cost per sale was under control, there was pressure to go back to the flattering metric. We kept contribution margin per order as the headline number; ROAS stayed as a diagnostic for the performance team only.
• Re-adding the cut products "because customers asked". Some did ask. We re-added two, at new price points that cleared the floor, and left the rest out. A request is not a business case.
• Letting the weekly report drift back to manual. A new marketplace, a new payment provider, and suddenly someone is exporting CSVs again. Every new system got wired into the report before it got budget.

What this cost and what it returned

The integration work behind the automated report took a few weeks of one engineer's time plus my own. The cuts cost three weeks of lower revenue. Everything else was decisions, meetings and a rebuilt price list. Against that, cost per sale fell by roughly 60%, order value nearly doubled, and the company moved from losing money on growth to net profit — within a quarter, on the same product catalogue and largely the same team.

That ratio is typical of operations work in e-commerce. The expensive part is rarely the fix. It is the months spent arguing about whose numbers are right while the loss compounds.

The operating-model side of this same engagement is in Fractional COO for e-commerce brands: where the margin actually hides.