The next generation of marketplaces may create implicit listings for things we didn't even know we wanted to sell
Traditionally marketplaces have been set up to connect sellers with buyers. But there are lots of people who own things — collectibles, homes, skills — that they may not even know they could sell.
Lately I've seen a handful of startups that are addressing pre-markets. They use statistical analysis to determine who is likely to start selling something soon and then build a marketplace or network around that.
TripleMint does this for real estate sales by using publicly available data to identify who is likely to sell their home sometime soon, and then they pro-actively reach out to see if the homeowner is willing to listen to offers, thereby augmenting the sizes of the real estate market. You can imagine how different data points about apartments and owners should be good predictors of whether someone is likely to sell their home sometime soon.
DocDelta does this for talent amongst doctors, enabling healthcare institutions to both recruit more efficiently and increase retention by pro-actively reaching out to doctors who, the statistics say, might be inclined to leave soon.
There are going to be lots more opportunities like this. For example, if Matterport can do a 3D scan of a room, why not also take inventory of what's in the room and list it somewhere for you?
We'll have to grapple with with this trend. For one, the point of not putting yourself or your possessions onto a marketplace is to not have to deal with offers — be they from recruiters or buyers. Likewise, a common trope within workplaces is that when someone updates their LinkedIn, it probably means they're looking around; it's therefore difficult for a person who updates their LinkedIn to keep their private desires discreet.
For another, it's uncomfortable to think that statistical analyses could predict our future behavior without our adding explicit signals — after all, DocDelta isn't relying on doctors updating their LinkedIn to try to figure out who's likely to move — but that algorithms can actually predict our next moves before we're even aware of them. People don't like being reduced to statistics, to the extent that if a service could tell me I'm 75% likely to leave my job in the next year, I might even be more likely to stick around to prove that I couldn't be reduced to statistics.
In a similar vein, there's something icky when you are able to internalize an algorithm (or at least think that you can). When I listen to Wilco on Pandora, Spotify, or Soundcloud, they're all pretty likely to start making recommendations within the alt-country genre, and I know why they're doing that even though it's wrong. I don't like any other alt-country band besides Wilco; I love them because of how they capture the grandiosity and stupidity of youth, which is why I also love Frank Ocean. I've never seen an algorithm that will recommend Frank Ocean based upon me listening to Wilco, nor can I really conceive how that's possible. Similarly, when I watch an episode of Friday Night Lights, Netflix will start recommending football documentaries. I know why they're doing that, and in this case it isn't even inaccurate — I do love football — but it feels kinda gross to be able to internalize the algorithm like that.
Recommendation algorithms are no doubt going to get much smarter in the future, and in the next 10 years I suspect we'll see a bunch more "pre-market" products and service spring up. There are strange and perhaps unpredictable second order effects that will come about, but they'll probably make for interesting fodder for the products and services that come about 20 years from now.
This blog post sprung, in part, from conversations with Josh Wolfe and Joshua Landy.