I’ve seen some writing on what makes marketplaces resilient to “supply side defection” as you call it, but I’d love to hear your take. Specifically I’d love to hear more about structuring to leverage the relationship, or anything else you think is important.

What specific structural things are important?

Generally speaking: your participants will use you when they need you and cut you out when they can do so cheaply. This isn't an indictment of them: it's completely rational for them to do this. But that means it's up to you to add value to the transaction, until the perceived cost of defection is higher than the platform margin.

Airbnb did this beautifully: they set up insurance for all their hosts, and so were able to frame "stay on our platform" as a safety directive.

You can also go the opposite way, but that usually leads to user-hostile antipatterns. E.g., Freelancer tries to detect all the ways folks might send each other phone numbers on their messaging service. This is both annoying and easy to circumvent.

If your rate of defection is high enough, no amount of tactical tricks will save you. Incentive and ingenuity together are too much.

When you're small and still have structural degrees of freedom, target the incentive by adding value to your offering if you can. If the problem only appears when you're big, you may have the resources to play whack-a-mole.

If you're small, have high defection rates, and also can't economically add value to your offering - it's time to move into another line of business.

Yes, you even see this on marketplaces like Ebay, etc. Once you've established a relationship with some buyer / seller, it doesn't take long before said marketplace gets cut out from the equation.

I sell a lot of expensive stuff on the web, and I'd say 80% of my buyers are repeat buyers that started contacting me through mail after a couple of deals.

Upwork (formerly elance) is a good example of a platform which could leverage the relationship by providing better intermediary services (like dispute resolution). In practice, they don’t, and there is no real reason not to defect.

From guest point of view, Airbnb defection defense is pretty annoying, as it doesn't let you post URLs, addresses, or phone numbers prior to booking. Also pretty easy to circumvent.

How does supply side defection work for airbnb except in very edge cases, such as a long term rental. Even then it turns into something like apartments.com

Well it's pretty much driven by customers but same diff really. People who book a night or two then realize they want to extend and ask to pay cash, mostly. Someone who's already stayed with you just texting the next time, etc.

It works since Airbnb is mostly there to hook you up and protect from crazy people's wanton destruction, so once you know you're dealing with safe people and you've gotten the review, eh.

Wouldn't really be worth the penny-pinching, except people tend to forget about the long-term booking discounts and actually end up paying more than they would've just using airbnb.

That's exactly how it works. The fact that it's an edge case on Airbnb is what makes the resulting churn survivable.

I wonder if it would help to rank a provider based on retention and CSAT and give more incoming leads to providers with more in-app retention?

How would you rank UpWork in this respect?

Upwork offers payment platform/record keeping and will dole out harsh punishments if caught leaving.

That's the value they provide and the stick. It works because trusting third world developers is hard. It fails for more local developers.