Which channels produced in-store and online revenue?

A customer researches online, checks stock, and buys in a shop. The session ends without a transaction, the campaign is judged a failure, and the revenue lands in a store's till attributed to nobody.

If the till captures a loyalty identifier, an email or a phone number, the in-store transaction can be matched to the campaign that produced the customer. In most chains this moves a majority of paid-media value out of Direct / Unknown.

A flagship in a city center and a suburban store draw from different demand and respond to different channels, and a chain-level average conceals both.

A store column mapped as Location shows each store's revenue and how much of it traces to a channel, so local budgets can follow local evidence rather than a national average that describes no individual shop.

Retail revenue is provisional for as long as the returns window is open. A channel driving volume in a category with a high return rate books impressive gross sales one month and gives a large fraction of them back the next, and every report built on the gross figure ranks it near the top. The returns are recorded, usually against the original transaction, and almost never propagated into anything marketing reads. The effect is not uniform either: it concentrates where fit or expectation is hard to judge online, which is exactly where paid social tends to spend.

Exporting net of returns changes the ranking, sometimes sharply. It is more work to produce than a gross sales extract, because refunds may sit in a different table and be dated to the refund rather than the sale, but the decision only has to be made once. Whichever basis is chosen, use the same one across every period being compared or a trend appears that was created by the export rather than by the business. A middle path is a column flagging returned or exchanged transactions, mapped as Product / service, which shows them as their own line so the gross figure is honest about how much of it came back.

Exchanges deserve a decision of their own. A customer who returns one item and takes another has not stopped being revenue, and treating an exchange as a full reversal understates whichever channel produced them. Treating it as two sales overstates the same channel by roughly the same amount. Neither is right, and the practical answer is to export the net value of the visit and write the convention down where the next person will find it. This sounds pedantic until two years of reports rest on a convention nobody recorded and nobody now remembers choosing.

Chains spend heavily on media that produces no click at all. A door drop, a catalog, a regional radio buy, a poster campaign around a new store — all defended with reach estimates and a sense that footfall improved. Digital channels are measured to the transaction and everything else is measured on faith, so the budget drifts toward whatever can be counted regardless of what is working. That drift is not a decision anybody made and is very hard to argue against without a competing number, which exists more often than retailers expect.

Anything that produced a list of people with contact details can be uploaded and ranked beside the paid channels. A catalog mailing file is such a list. So is an event registration, a competition entry, a new-store opening sign-up, or a partner's shared audience where the sharing is properly permitted. Each one is ranked on the revenue that followed rather than on impressions delivered, which is usually the first honest comparison a marketing team has managed between a mailing and a media buy. Where no such list exists the activity stays unmeasured, and the report says so rather than estimating.

This is also the practical answer for store openings. A new location's early revenue is normally credited to whatever digital campaign ran nearby, because that is the only thing being counted at the time. With the opening's sign-up list added to the lead file, tagged with the opening, the opening itself appears as a channel with revenue attached and can be compared against the media that supported it. Chains opening several stores a year get a repeatable answer to what an opening is worth — a question usually settled by whoever presents most confidently, and one of the few in retail with a clean before and after.

Everything in a store depends on whether the transaction carries a contact detail. A loyalty scan, an emailed receipt, a delivery address, a warranty registration, a returns record — any of them is enough, and a till capturing none of them produces a sale that cannot be matched to anything at all. That capture rate, rather than the quality of the analysis, is the hard limit on how much in-store revenue can ever be traced. It varies enormously by chain, by category and often by shift, and knowing the number is the first useful output of the exercise.

What matters is that the limit is visible rather than hidden. Revenue from transactions with no identifier sits in Direct / Unknown and is never spread across the paid channels to improve a headline, which means the attributed share is a floor rather than an estimate. A chain capturing identifiers on two in five transactions is reading a report about two fifths of its store revenue, and it knows that while reading it. A tool reporting a complete picture from an incomplete till is guessing, and the guess is not marked anywhere on the page.

It also points at the highest-return operational change available. Raising the capture rate at the till by a few points does more for the quality of every future report than any refinement of the analysis, and it is a staffing and prompting question rather than a technology one. Most chains discover the rate varies more between stores than between channels, which is itself worth acting on. Nothing about the method changes as it improves; the same two exports simply cover more of the book each period, which makes the improvement measurable in the report it improves.