Account audit template for Ecommerce

The account audit with the checks a store needs: the catalogue feed and what it actually contains, which parts of the range are covered and which are not advertised at all, audience overlap between prospecting and retargeting, and whether the reported revenue matches the store's own.

Everything else is the general audit, unchanged. One row per finding, evidence attached, honest severity, one action, one owner, one date.

Run it quarterly and again after any season where the range changed substantially. A store's account accumulates campaigns for products that no longer exist faster than any other kind, because the range turns over and the campaigns do not.

Run it also before a peak, because the peak is when every unnoticed inefficiency is spending at four times its usual rate.

A store's catalogue turns over constantly. Its account does not, because campaigns are built during launches and rarely revisited, and nobody is responsible for noticing that a campaign is still pointed at a line that was discontinued two seasons ago.

That is the first thing this audit looks for, and it is the check most likely to produce something actionable in an hour: join the spend export to the current feed. Anything receiving spend and missing from the feed is discontinued, out of stock or mis-tagged. Anything in the feed with good margin and no spend is an opportunity that exists because of history rather than because of a decision.

Most stores find both lists longer than they expected, and neither is visible from inside the ad platform's own reporting.

Feed checks in an audit should look at the contents rather than at the error count.

A feed can report as healthy while carrying wrong prices on a subset, missing images on another, the wrong variant as the default on a third, and stale availability on a fourth. None of that shows up as an error, and all of it degrades performance in ways that look like creative or bidding problems.

So the check is concrete: sample the feed against the site, product by product, for a set that includes the highest-spending lines. Record what differs, with the date. Where the feed is generated by a third-party integration, record which one, because that is usually where the fix lives.

The overlap check is the one that most often changes how a store spends.

In a lot of accounts the prospecting audience and the retargeting audience contain a substantial number of the same people, because prospecting is broad and the platform will happily serve it to somebody who visited yesterday. The account then bids against itself, and the retargeting campaign is credited with conversions the prospecting campaign had already produced.

Measure it rather than assuming it. Where the overlap is large, the recommendation is usually exclusions rather than budget changes, and the effect on the reported return of each campaign will be dramatic even though the underlying business result barely moves. Say that in the impact column, or the fix will be read as a failure.

Platform-reported revenue and the store's own revenue rarely agree, and the size of the gap decides how much any reported return on ad spend can be trusted.

Compare a full month, in both systems, on the same definition. Then record the difference and the likely causes: attribution windows, returns not deducted, currency handling, orders from channels being double counted, tracking that fires on a page rather than on a confirmed order.

The point is not to make the numbers match. It is to know the size of the gap, so that a decision made on a platform-reported figure is made with the gap in mind. An account run on an unreconciled number is an account run on faith, and it will eventually be defended with a figure nobody can reproduce.

A store's account tends to accumulate campaign structures that were created for a specific promotion and never removed.

List every campaign that has spent in the last quarter and the promotion or season it was built for. The ones built for an event that has passed are either doing a different job than their name suggests, or they are spending on a message that no longer applies. Both are findings.

The recommendation here is usually consolidation rather than deletion, and the effort column should say so honestly: restructuring an account is a project, not an afternoon, and an audit that recommends it without saying so produces a row that stays open for a year.

A specific check worth carrying: spend in the last month against products that were out of stock for part of it.

Most platforms will keep spending on an unavailable product until the feed tells them otherwise, and the feed is often slower than the warehouse. The finding is the amount, and the recommendation is usually a feed rule rather than a campaign change.

Run this one before a peak as well as quarterly, because during a peak the same leak drains four times faster.

One more check earns its place in a store: how long each piece of live creative has been running, and how much of the spend is going to assets older than a season.

Stores rely on creative volume more than most advertisers, and the failure mode is quiet. A set that worked is left running because it is still the best performer in the account, which is true and is also a statement about the alternatives rather than about the set. Frequency climbs, the same people see it repeatedly, and the cost per order drifts upward slowly enough that no week looks like a problem.

The finding is a list of live assets with their first run date and their share of spend. The recommendation is usually a production schedule rather than a campaign change, which makes it a planning row rather than a fix.

Seven checks. Every finding separates what was found from what should be done, with evidence and a date range. The feed has been compared with the site, not only with its own error report. Coverage across the range has been measured, in both directions. Prospecting and retargeting overlap has been quantified. Reported revenue has been reconciled with the store's own, and the gap recorded. Every row has an owner and a date agreed with them. And findings somebody has decided to live with are closed as accepted.