[photon_off]

Firstly, congrats on the launch. Designing and implementing a new concept from scratch, by yourself, is definitely fraught with difficulties. It's hard to know if what you're designing is "right" and if you're implementing "the right way". Sticking with it for an entire year and actually launching is definitely something you should feel good about.

I think this is a great idea, and for a first version the execution is actually quite good. Once I figured it out, I enjoyed using it and flew through all of the cards. I looked at a lot of matches and even sent a few messages. Overall, I can see myself coming back for random chats from time to time.

Here is my feedback, in order of what I perceive to be most important.

1. It was not at all obvious to me how to swipe. I tried swiping the card but it didn't move. So I clicked the arrows. Since I "agreed" with the first 4 or so cards, I did not realize the arrows were to go back and forth. (In fact, I would not have ever expected back and forward to be an option). Then I realized that the "help button" thing was actually how to slide, and I got it from there.

I suggest you make the slider thing look "slideable" and obvious that left is disagree (red), right is agree (green). Right now, it definitely is easy to see it as a "help" button.

I also suggest you make the card navigation arrows not ambiguous -- I think it is fair to say people will assume they mean "agree" and "disagree" rather than "next" and "prev". Either provide context so that it's obvious what they do, or use words "next", "prev".

Here is a crappy mock-up that hopefully shows what I mean: https://jsbin.com/wekuzebasu/ - In this shows progress and is obvious what next/prev do, and (for me at least) it is much more obvious what the call to action is, and how to operate it. (You could replace "|||" with "<->", or show arrows next to the slider circle for even more emphasis).

2. I really love the "percent" agree/disagree. I think that's way better than just "yes/no" and it greatly increases my confidence that I'll get better results.

3. I think your initial card set is actually quite good. There are a few where it is ambiguous of "left" and "right" mean. Agree / Disagree?Like / Dislike? Support / Don't support? Care / Don't care?

4. If I were you, it would be very easy for me to worry about spam/abuse/gaming of this. But, honestly, your biggest problem will be overcoming the network effect to get a critical mass of users to make this better. If for an average user there are more similar people, and they are closer, it becomes much more compelling for more users to join -- leading a virtuous cycle. If you want this to catch on, your #1 concern should be getting the most users in the least amount of time.

Handling spam/abuse is a problem you want to have, and I'd put it off until it's necessary and/or until it's clear it is slowing user growth.

5. I've read all of your comments, and I think you're making some very good decisions and are putting a lot of thought into this. I hope you stick with it.

6. The clustering is good and does provide some level of "privacy" -- though for my "similar" people I often do not see any clusters at all. Anyway, this seems like the type of app where I want people to know my opinion on stuff, and if there's a question for which I want my opinion to be unknown, I'll just skip it.

-----

Now a question regarding the matching. Is there a difference between "unanswered", "skipped", "0% (as opposed to -X% or +X%)" ? Consider these cases:

1) I don't care about a card. This is distinct from being "0%" -- I don't want this card counted towards producing my matches. If someone is 100% or -100% on it, don't penalize the similarity.

2) I am neutral about a card, but I do care if somebody is +/- on it. Do consider this in similarity.

3) I have not answered a card. This should not count towards similarity.

4) I previously had an opinion on a card, but now I don't care. How do I "reset" it?

Depending on how this is handled, you'll need to change the UI to make it more clear. Eg, you may need a "skip" button.

[pitherandd]

Wow, this is an incredible reply. Thanks so much for such an in-depth analysis and compliments!

I actually really like your mock-up and really appreciate that you took the time to make it. Hopefully you don't mind if I take quite a bit of inspiration from it, because I definitely think it looks much better than my current one.

Yeah, I agree that the network effect is one of the biggest problems. Retention seems very tricky with a concept like this because while posting it to reddit or HN can result in many registrations, few people stick around because there isn't much to do on the site because there's so few users at the moment. It honestly does make motivation pretty difficult, but I am indeed going to stick with it regardless.

With regard to your last 4 numbered points, currently 0% == not answered == neutral, and this is mostly due to technical limitations. Your default similarity cube is just (0.0, 0.0, 0.0, 0.0, 0.0, ..., 0.0) (50 zeroes), and voting on cards adjusts this accordingly. Such that if you answer 100% to every card in the first cluster and answer no other cards then your CUBE would look like (1.0, 0.0, 0.0, 0.0, 0.0, ..., 0.0)

I was thinking about a potential way to solve this, but as of right now due to the constraints of CUBE similarity matching, this is the current solution.

Thanks again for such an insightful reply!

[photon_off]

Feel free to use it however you want :)

Regarding the 0.0/default/skipped -- see my other response about "cardsets", which the more I think about it, the more I am convinced will solve nearly all of your issues:

  - you don't have to worry about the skipped vs neutral problem (as it is fair to expect a user to answer all cards in a cardset).
  - you don't have to worry about clustering. a cardset is a "cluster".
  - you can work around technical limitations by having each cardset be its own cube
  - you boost retention by being able to have a library of dozens of cardsets (rather than 250 cards).  limitless swiping, and swiping on things that your users will find interesting because they are selecting cardsets that are important to them.
  - you can make your site viral by allowing users to send a link where the receiver can (anonymously) swipe all the cards of a cardset and see their "similarity" to the link sender.  (Then you can ask them to register to be able to send links themselves).
  - futures paths for revenue are endless