Review my startup: Similarity, online dating in the age of Facebook | Hacker News Reader

Review my startup: Similarity, online dating in the age of Facebook

10 pointsstickfigure7 comments

Here it is: http://www.similarity.com/

I'm trying to build an online dating site that actually makes good use of the social graph. It uses some elements from traditional algorithmic matching systems to make a game of sorts out of setting up your friends. Or you can just use it for yourself.

Since we're all geeks here:

* Automatically builds a profile for you out of your FB info when you arrive.

* Matching is based on Likes. You can like and dislike more stuff.

* You get to play with the algorithm and get different results.

* You can limit match results to the friends-of-friends circle.

* You can pick a friend to match as, and see their match results - even if they aren't a Similarity user.

* You can suggest dates on your friends' walls, which posts a cute blurb about the things they have in common.

There's just one developer right now, me. I built it in about four months using Appengine, GWT, and a half-dozen opensource projects - some of which I published myself (http://objectify-appengine.googlecode.com/, http://batchfb.googlecode.com/). I pulled off the "coming soon" page a week ago so my friends can try it out, and now it's ready for public consumption (I hope!)

I'd love to hear your thoughts.

Don't do the "friends" thing--reason: the friend zone. What would be interesting is if users could post content to a generic stream sorted by proximity, sex, and age. Users set thresholds for mutual likes before their information is made available to one or the other. The more you like another user's content the larger or more visible it becomes in the stream; the more another user likes your content the "hotter" (perhaps in color?) it becomes in the stream to you. The important thing here is to make sure there is a positive feedback loop between two users, ultimately creating content posts that are both "hot" and "large". If anyone is wondering, there are studies showing that compatible people like the same things, however obvious the notion may seem.

If you want to beat a dead horse: allow fb users to add people they like to your service whom aren't their friends using their URIs on fb. Use fb's ad api to allow users to pay for highly targeted ads to these people they like (<--monetize this). This will also help solve your chicken and egg problem. Where two anonymous users add each other's fb URI, there is a match and both are notified (you can crush crush notifier).

Bonus points if you can combine both these concepts into one service uniformly. This would be how I would make a dating service. Btw, nice domain name.

I've been planning something along the lines of your first suggestion - essentially a replacement for craigslist personals, with temporal data but sorted efficiently. I can do that pretty well with the dataset I have. The problem I see is that it has an even worse chicken-and-egg problem; it requires people to post a bit of content. Right now the steps to going in the match system involve: 1) Accept FB connect and 2) Click "I'm available". If folks had to write content too, there wouldn't be anything to start.

I figure that a feature like this makes sense to add once the kernel of a critical mass exists. It can happily coexist with the traditional "search everyone's profile" feature.

Which feature are you referring to as the "friends thing"? There are two features that leverage the social graph - you can limit matches to friends-of-friends (2°, not 1° - you already know those people) and you can suggest dates for your friends (I suspect your bro's/best gf's will be the prime user of this feature).

Thanks for the feedback.

Background color really hurt my eyes, also it asks for too much data access. Normally I'd never authorize an app to do that but I let it slip this time to give you a feedback.

At first,to test this app I basically switched to monkey mode and started clicking on links and buttons to list some chics' profiles. Nothing happened.

Then I switched to manual mode and clicked on 'matches' tab. I only had 2 parameters in my interests list, so no result was returned.

Then I clicked on 'you' tab, started rating the things that the app threw at me, after about 15 votes or so, I started to have some results in matches section.

If I were a regular user who was using the app for the first time you'd probably lose me after the 'clicking around and not getting any chic profiles' part. So directing newcomers using visible (shiny, bouncy) tooltips is crucial.

I was getting a limited number of profiles in results, so I tried to play with sliders. I was expecting to get different profiles but instead I just got the same profiles, but fewer of them. I don't know if this is a bug or a feature, but I'd love to see different people based on the search criteria. Do you keep showing the same results until a member pays you money?

Thanks, this is great feedback. I must work on the newuser process.

The chief problem is that there are only a few dozen "published" profiles in the system right now, so you're probably seeing all of the candidates that match to your likes. The sliders just aren't going to be much fun until there are at least a few hundred users.

I suspect that the chicken-and-egg problem is the biggest issue with online dating startups, but it's one that (I think) will go away after a thousand or so users sign up.

I concur on the orange background. For some reason its a big turn off.

The text on the home page is a bit difficult to read. I'd suggest a color with more contrast.

"Oh no, a Facebook app?!

    * We won't spam your wall (unless you ask)
    * Your profile starts out hidden"

Fixed. I was attempting to make that text unobtrusive, but apparently succeeded a little too well. The landing page is currently the "minimum viable experience" - I will redo it soon.

Reply on news.ycombinator.com