A sofa is browsed online and bought on the floor.

Furniture is the textbook considered purchase. People measure the room, argue about it for three weeks, come in twice, and buy on the second visit. Every one of those steps happens on a different device and most of them happen off your website entirely.

The result is a channel report where paid search looks mediocre and 'direct' looks like your best salesperson. Neither is true. What is true is that the sale closed somewhere the pixel does not go.

Your POS or order system already records a phone number and an email against every delivered order, because you need them to schedule the delivery. That is the whole input.

Match it against your ad platform lead exports or your call tracker and the showroom orders line up against the campaigns that produced them. Orders with no usable contact detail stay in Direct / Unknown, where they belong, rather than being spread across paid channels to make a report look tidier.

The largest orders in a furniture business are rarely written across a counter. They come out of a design consultation, in the store or in the customer's home, where somebody measures, draws a plan, specifies fabrics and puts a whole room together. That appointment is the real conversion event, and it is the thing a furniture retailer's marketing is genuinely buying. It is also, almost universally, recorded as a form fill and then forgotten the moment the designer takes over the relationship.

Putting consultation bookings in the lead file changes what the marketing conversation is about. Instead of cost per appointment, which is easy to lower and tells you nothing worth knowing, the report shows what an appointment from each channel became in order value. Two channels producing the same number of consultations at the same cost can differ by a factor of three once the rooms they sold are counted, and nothing in a booking report will ever say so, because a booking report has no money in it.

The gap usually traces back to intent rather than to the designer. A channel advertising a promotion books people who want the promotion; a channel advertising a room books people planning a room. Both fill the diary at the same rate. Only one of them fills it with projects. Knowing which is which lets the store decide how many appointment slots to give each source, which is a more useful lever than the bid adjustments a budget conversation usually ends at, and one the floor can act on the same week.

Furniture is sold long before it is delivered, and a proportion of what gets written never arrives at all. The special order is discontinued, the fabric is unavailable, the lead time moves from eight weeks to twenty and the customer cancels, or the sofa will not go up the stairs. Every one of those is a written order sitting in the system and none of them is revenue. A channel report built on written orders is therefore ranking sources partly on things that fell through.

Export delivered and invoiced orders where the system allows it, or leave cancelled orders out of the export, because every row in the sales file is counted as a sale at the amount it carries and there is no status filter to remove them later. A cancellation rate by source is not something this report calculates, and including cancelled orders would credit channels for revenue that never arrived. A channel selling heavily from a clearance floor cancels quite differently from one selling special order, and the honest way to keep that difference from distorting the ranking is to rank only the orders that stood.

The date question follows from the same place. An order written in March and delivered in June can be booked on either, and the two conventions produce months that look nothing alike. Many retailers use the written date for the channel report, because that is when the customer decided and therefore when the advertising did its work, and keep the delivery date for operations. Whichever is chosen, applying it consistently is what makes a comparison against last year mean anything at all.

The first report tends to produce three reactions. The unattributed share is smaller than the sales floor expects, because the customer recorded as a walk-in usually rang about stock or delivery first and left a phone number while doing it. Email ranks far higher on revenue than it ever does on clicks, because a list of past customers produces large second orders. And the channel with the cheapest leads sits somewhere in the middle of the table rather than at the top of it.

A second reading is about the mix. Once a department or category column is in the file, the ranking splits: one source sells upholstery, another sells bedroom, another sells the accessories and rugs that carry a better margin than either of them. A single blended ranking hides all of that and pushes a store toward funding whichever source happens to sell the highest-ticket category, regardless of what it earns along the way. Splitting it turns one budget decision into several better-informed ones.

None of this required a change of process. A furniture retailer already records a phone number against every order, because a delivery cannot be scheduled without one, and the advertising exports already exist inside the ad accounts. That is why a completed year can be analyzed on the first afternoon instead of accumulated over the next one, and why the sensible first move is to run last year rather than wait for this one to finish. The answer either changes a decision or it does not.