Reputation vs Curation

A couple of weeks ago I had a great conversation with John Breinlinger, a VC from Silicon Valley. We spoke about the future of reputation in the Sharing economy.

I was arguing how my company was trying to build the best reputation systems, so that people could better choose who to transact with. His take was that it would not be necessary to choose, because users would be curated.

What is the difference between reputation and curation?

• With reputation, you have access to the previous history of transactions, and some metric that puts those transactions together, in many cases just an average rating from 1 to 5.
• With curation, the company decides who is allowed in the game. This means that you arguably don’t need to worry about whether people are safe or not. Everyone in the system is assumed to be “5 stars”, and you simply cannot see the bad people.

Examples:

Today we have all sorts of platforms and services doing different things:

• Pure reputation: eBay. People make their own decisions about whether to transact with somebody or not. I personally don’t consider people under 97% positive rating. I guess that many people who go below this level and see their sales-flow decrease will probably start a new account from zero.
• Pure curation: You have probably heard of this new Tinder for the elite of Stanford. I would say that’s pure curation. It’s not clear how they will manage people who misbehave later. They probably need to wait for another user to trigger some alarm, and then they would re-curate.
• Reputation + Curation: Companies like Uber or Hailo show the star rating of their drivers, and if that metric falls below certain value, the driver will be taken out of the system. This ensures there are sufficient incentives for drivers to behave at a very high standard.

What is best:

The argument for curation is that the use of the service should be invisible to the user. “I don’t want to have to look at the Uber rating of my driver. I just want to go from A to B.” Why should I have to worry about whether my driver has 4.5 or 4.9? I want Uber to worry about it for me. It seems to make sense for binary services where value-add of the person involved is limited.

Things are not always black or white. In a system like oDesk you will find hundreds of people offering services for “SEO”. People set a pay rate, and you can read their previous ratings from other customers. Here it’s not a matter or being or not being part of the system. It’s a matter of value. I might prefer 4.9 stars at $20/hour over 4.3 stars at $15/hour, but I might also prefer the same 4.3 stars person if the price was $10/hour. That “price elasticity of stars” makes crowdsourced reputation valuable.

It could be argued that oDesk could curate their users based on skill, giving their own shades of gray, but it’s difficult to imagine a centralized skills marketplace assessing all the possible skills their users could have and their qualities. Users are a lot more likely to provide more thorough feedback.

Uber´s mix of reputation and curation is interesting. Why, if they already curate out the bad people? I have three possible explanations:

• If you don’t see the rating of people, you will feel no visual incentive to give a rating to your driver, unless it was a negative one. The rating would be too invisible. Uber needs those ratings in order to perform the curation.
• It creates some expectation management in the user. If the user gets a 4.3 driver (on the verge of being curated out), their expectations for a great car might be lower while the car arrives. If the ride goes well, satisfaction will be perceived as higher.
• People want to make their own judgements. It’s great that Uber curates for me, but I still want to be able to make a decision about who I do business with. If I don’t see their rating, I might feel “in the darkness” and stressed. This might happen because it is wired in our brain (we want to work with people we have been introduced to, or at least, we know something about their past) or simply because the Internet was born with eBay and we need a few years to get used to pure curation.

My take is that the last one is by far the most important one, and I would say it is because it is wired in our brain that we want to know something about other people’s previous history. When you take some knee Surgery you want to know your surgeon. Of course the surgeon has already been curated by the hospital, but it makes you feel comfortable to know that she has performed 100s of similar operations successfully.

The additional problem of curation is that, at some point, even if today curation is done by humans, it will be performed by algorithms. And it would be harsh to lose your job because an algorithm curated you out. It would also be hard that some algorithm decides for me that I cannot buy a product from a 96% eBay user, even if I usually prefer 97%+. Making that decision ultimately belongs to me. Maybe I have read the negative reviews and I believe that the reviewer was unfair. Or it’s a very rare item and I am willing to take risk. Or maybe I am getting a lower price precisely because of the higher risk of dealing with that person, and it is up to me to decide whether I want to take 1% more risk in exchange for $10 difference in the product.

Conclusion

I am glad I had that conversation. My brain was challenged to think about something I had just assumed. I believe that reputation and curation will have to co-exist in almost every application, and each application will simply put some more weight on one over the other depending on the value-add of the individual.