Performance marketing is advertising bought against a measured outcome rather than against reach, and in ecommerce that outcome is normally a sale. The promise is appealing and the measurement is harder than the pitch admits: platforms report conversions they had a hand in, attribution windows differ between them, and the same order can appear in three dashboards at once. Buying well means agreeing what will be counted before anyone spends anything, and hiring an agency willing to be judged on the number you chose rather than the number that flatters it.
The metric argument you should have up front
Return on ad spend measured inside a platform is the friendliest possible number, because each platform claims credit generously within its own window. Blended return, meaning total revenue divided by total advertising cost across everything, is unflattering and honest. Incremental return, meaning what the advertising produced beyond what you would have sold anyway, is the truest and the hardest to obtain, usually requiring holdout tests. A serious agency will bring this distinction up on its own and propose which to manage against. If it reports platform-attributed return only, you will eventually find a quarter where every dashboard is green and the bank balance disagrees, and there will be no way to establish which of you was right.
Where the margin actually comes from
Performance work in ecommerce fails more often on economics than on media skill. If contribution margin after cost of goods, shipping, payment fees and returns leaves little room, no bidding strategy will rescue the account, and the honest advice is to fix pricing or product mix before spending more. Ask a prospective agency what it needs to know about your unit economics and how it would use it. One that asks for cost of goods, repeat purchase rate and return rate in the first conversation is planning to manage to profit. One that asks only for your current spend and platform access is planning to manage to a dashboard. The same applies to what the money is aimed at: acquiring a new customer at a loss is a defensible strategy only if you can show the second purchase actually happens.
The site side of performance, which agencies often ignore
Half of performance is what happens after the click, and it is the half most media agencies are not staffed to fix. Page speed on a phone directly affects how many paid visitors ever see the product, and the loading metrics Google publishes as Core Web Vitals are a reasonable proxy for what a shopper on a poor connection experiences. Beyond speed: how many steps stand between product and payment, whether guest checkout exists, whether shipping cost appears before the final screen, whether product pages answer sizing and returns questions that otherwise become abandoned carts. Ask any performance agency what it would change on the site, and whether it will do that work or only recommend it. The answer determines whether you need a second supplier.
How these engagements are priced
Percentage of spend is the most common model and the one with the clearest conflict, since the fee rises with the budget regardless of outcome. Flat retainers are neutral on spend and easier to compare between bidders. Commission on revenue aligns better in principle and needs careful definition of which revenue counts, because a fee on all store revenue rewards an agency for sales it never influenced. Hybrids with a modest base plus a performance element are common at larger budgets and are reasonable if the target is honest and the measurement was agreed in advance. Whatever the model, insist that ad accounts, pixels and audience data stay in your ownership, and that the invoice separates media from fee. Brands buying this as a standalone discipline generally shortlist performance marketing specialists rather than full service shops, since the day to day work is media buying, not brand.
Questions people ask about ecommerce performance marketing
What return on ad spend should we target?
There is no universal figure, because it depends entirely on your contribution margin. Work out what gross margin remains after product cost, shipping, fees and returns, and derive the break even multiple from that. A store with slim margins may need a high multiple to survive, while one with strong margins and repeat purchases can profitably run much lower. Any agency quoting a target before seeing your margins is guessing.
How much budget do we need to start?
Enough for the platform to gather conversion data at a usable rate. If your budget produces only a handful of purchases per week, automated bidding has too little to learn from and results will be erratic. Either concentrate the budget on fewer products and audiences, or optimise toward an earlier signal such as add to cart until volume supports the real one.
Should we run one agency across all channels?
One agency makes budget shifts and blended reporting simpler, which matters more than channel specialism for most stores. Splitting channels between specialists can raise craft but creates an attribution argument nobody owns, with each supplier claiming the same orders. If you do split, appoint someone to hold the blended number.
How long before we can judge performance?
A meaningful read usually takes a quarter: several weeks for tracking, creative and learning phases to settle, then enough weeks of stable spend to see a trend. Judge the first month on setup quality and hygiene rather than on return, and hold the agency to the metric you agreed at the start rather than the one that looks best in retrospect.