Treating listing content as a per-deal cost versus fixed overhead

Overhead, and the per listing version is a category error. Anything whose effect shows up before a deal exists cannot be underwritten inside that deal, because at underwriting time the content that sourced it has already been paid for. Charging the deal for a cost that produced it double counts against the same decision rule it is meant to inform. The right approach is to hold it as a fixed monthly line and test it annually against deal count rather than per deal. If deal count and pipeline hold flat while spend rises, the line is wrong and should be cut. That is a slow test, and it is the only one that will not lie. The two sellers who mentioned the posts are the strongest data point in this and deserve more weight than they are getting. Unprompted recall from a counterparty is about as good as evidence gets here. It is not attribution, it is proof the spend bought recognition, which is what it was meant to buy.

I've got no deals closed yet, so take this from the outside, but the failure mode you named for model two is the one I'd be scared of. "Cancelling feels like going dark" is how subscriptions survive forever. At least the per listing version has a natural stopping point.

I split it and it isn't a fudge. Production costs that exist only because a specific property exists, the photos of that property, the walkthrough of that unit, go in the deal. Anything that would exist whether or not I owned anything, the account, the email list, the person keeping it fed, is overhead. The test for each half is different. The deal half has to beat the alternative of a phone camera. The overhead half gets a hard annual review where I ask whether I'd start it today at today's price.

That second question is the one that kills lazy retainers. It's much harder to answer yes to than "should I cancel."

Canton's split is the one I'd copy. The part I'd add is that the two halves are moving in opposite directions right now. The property specific half is getting cheaper fast because drafting and editing tools got good, so if your per listing number hasn't come down over two years someone is keeping the savings. The overhead half is getting more expensive in real terms, because everyone's producing more and being noticed costs more attention than it did. If you model them as one line you'll miss both moves.

1,900 across four transactions works out to 475 a deal, and that is nothing, it does not need a model. The 4,400 figure is the line worth arguing about, and the honest reading of that kind of data is usually two soft mentions and no numbers. That does not mean cut it. It means an option is being funded, and the model should say so rather than pretending there is a measurable return in it. The question worth asking before voting on a case like this is what the 4,400 buys per month and whether that number would drop if the scope were rewritten around the two or three things sellers actually see. Most presence retainers end up paying for volume nobody looks at, and a single afternoon spent reviewing what actually got engagement usually settles it.