Which channels produced members who actually stayed?

Co-living sells a bedroom with shared living space on a flexible term, often to somebody new to the city. The price point is lower than a whole apartment, the decision is faster, and the turnover is much higher.

That combination makes cost per booking a poor guide. A channel that fills rooms with three-month stays and one that fills them with fourteen-month stays produce identical booking counts and very different revenue, and only one of them is worth scaling.

Members extend, move between rooms and sometimes between buildings in the same portfolio. Treated as new bookings, each of those looks like an acquisition the marketing produced twice, which double-counts the channel.

Matching on the member rather than the booking keeps one person as one acquisition, with all their later revenue attached to the channel that found them. It reduces the headline number and makes it correct, which is the trade this product exists to make.

The product is other people. A member who likes the house tells a colleague, brings a friend into the spare room, and posts about the Sunday dinner, and the next enquiry arrives already half-convinced. Operators know this is happening and most of them run a referral credit of some kind to encourage it. What almost none of them do is treat it as a channel with a cost and a return, because there is no platform reporting on it and no invoice arriving at the end of the month to prompt the question.

That omission distorts everything around it. Paid search is judged against a field that excludes the largest producer of move-ins in the building, which makes every media decision look more consequential than it is. It also means the referral credit itself is never priced: nobody can say whether doubling it would fill more rooms than doubling the advertising, and the argument is settled by whoever feels strongest about it. The building's best marketing asset is therefore the only one with no owner, no budget line and no report. It is also the one most likely to be damaged by a bad month of operations.

Any list of referring members with contact details uploads as a source file and is ranked on nights billed alongside the paid channels. The referral credits are a real cost and can be set against a real revenue figure. For most operators the first version of this report is the first time community-building has appeared on the same page as media spend with a number next to it. Whatever the answer turns out to be, it is the first time the question has been asked with revenue on both sides of it.

Rates in co-living move constantly. A room is priced against how many are empty this week, how long the building has been open, and what the operator down the road is asking. A member arriving in a soft February pays materially less than one arriving in September, and both of them sign for the same bedroom with the same furniture. The booking count treats them as identical, and the rent roll does not. An occupancy report shows the difference and a channel report does not, so the two documents describe the same month in incompatible terms.

This is not a pricing error; it is how the business works. It does become a measurement problem the moment channels are compared on bookings, because a source that performs during quiet weeks is systematically credited with lower-value members and a source that performs in peak season looks better than it is. Nothing in a booking count can separate a channel's quality from the calendar it happened to work in. The distortion is systematic rather than random, which means it does not wash out over a year. It compounds, quietly, in favour of whichever source happens to work when rates are strongest.

Ranking on rent actually billed removes the distortion without anybody adjusting anything. A member who booked at a discount and stayed fourteen months outranks one who booked at full rate and left in ten weeks, which is the correct ordering and the opposite of what a cost per booking implies. Exporting the move-in month alongside the amount lets the same question be asked season by season. That is the version of the number an owner can set against the rent roll without anybody having to explain a methodology first.

Between a confirmed booking and a member actually living in the room sit a deposit, a reference check, a visa or a job that may not materialise, and a change of mind. Short-notice bookings cancel at meaningful rates in every operator's data, and the room that was held for three weeks earned nothing at all. Some channels produce this far more than others, particularly those reaching people planning a move from another country months ahead. A room held against a booking that evaporates is worse than an empty room, because it was not being sold to anybody else while it waited.

Counting bookings therefore rewards sources that generate intent rather than occupancy, and the gap is invisible until somebody puts the two files side by side. It is the same failure that makes a full pipeline feel like a full building, right up to the point where the operations report arrives and it is not. The pipeline number is not dishonest. It is simply measuring a different thing from the one the business is paid on, and nobody notices while occupancy holds up. Once it slips, the two reports diverge in a way that is hard to explain to an owner.

Because the outcome being ranked is billed revenue, a cancelled booking simply produces nothing against that source. No adjustment is applied and nothing is estimated. Uploading the booking file as the enquiry side and the billing file as the revenue side makes the shortfall explicit per channel, which is usually a more actionable number than the overall cancellation rate everybody already quotes. Cancellation is a property of the source as much as of the member, and treating it that way is what turns it into something a media budget can respond to.