Inside Match.com: It's All About the Algorithm(slate.com)
slate.com
Inside Match.com: It's All About the Algorithm
http://www.slate.com/id/2300430/
18 comments
Two interesting things the articles blows right past:
- That they compare their algo to the ones used by Amazon and Netflix to recommend products, while ignoring that products have much better defined boundaries than people do. (And ignoring how both Amazon and Netflix suffered through some horribly bad recommendations in both their systems, early on).
- That they think Facebook (or any big social network) complements them, instead of rendering them irrelevant.
A dating service built on a matching you to real world friends-of-friends-of-friends has a much better chance at success than Match.com's "People who dated Alice also dated Mary, Jill, and Bertha. Sign up now for 1-Click Dating!"
- That they compare their algo to the ones used by Amazon and Netflix to recommend products, while ignoring that products have much better defined boundaries than people do. (And ignoring how both Amazon and Netflix suffered through some horribly bad recommendations in both their systems, early on).
- That they think Facebook (or any big social network) complements them, instead of rendering them irrelevant.
A dating service built on a matching you to real world friends-of-friends-of-friends has a much better chance at success than Match.com's "People who dated Alice also dated Mary, Jill, and Bertha. Sign up now for 1-Click Dating!"
It's not like people haven't tried to build the later, but none of them have succeeded, which suggests that you're mistaken...
I'm guessing because people actually meet their real life friends-of-friends at parties, etc. anyway and haven't met anyone from that pool of people who they want to date.
I'm guessing because people actually meet their real life friends-of-friends at parties, etc. anyway and haven't met anyone from that pool of people who they want to date.
Bravo! Fantastic illustration of how dating sites are broken.
Let's fix them?
Competitive. Yes. But, their may be a model for a small annual subscription instead of the quite high current monthly subscription model.
How about $20 per year? £10 I'm thinking - I live in the UK.
okcupid/pof are great, but well... not that great. Paid means you just have the people taking it seriously, and you can validate who people are. But no need to take all there money that could be spent on going out on dates ...
My 2p.
Any other rails coders fancy helping out?
Competitive. Yes. But, their may be a model for a small annual subscription instead of the quite high current monthly subscription model.
How about $20 per year? £10 I'm thinking - I live in the UK.
okcupid/pof are great, but well... not that great. Paid means you just have the people taking it seriously, and you can validate who people are. But no need to take all there money that could be spent on going out on dates ...
My 2p.
Any other rails coders fancy helping out?
Monthly is better than annual for one reason: People over-estimate that they'll get a date so they'll avoid annual unless you offer something very compelling beyond trying to get a relationship on the website. eg person will hang around to chat to friends or do quizes.
If you really want to create a dating website, you have these classic problems:
1) Abundant men. eg spam 100 women a day over several accounts until they get a response.
2) High turnover. Got a date? Great, they're not coming back. Your email gets quickly marked as spam. Among other issues.
3) Over-estimation. People don't say their true height, weight or even show the correct pictures.
4) Chicken and egg problem. The biggest hurdle to overcome first. How you overcome it will affect the above three.
Have fun!
If you really want to create a dating website, you have these classic problems:
1) Abundant men. eg spam 100 women a day over several accounts until they get a response.
2) High turnover. Got a date? Great, they're not coming back. Your email gets quickly marked as spam. Among other issues.
3) Over-estimation. People don't say their true height, weight or even show the correct pictures.
4) Chicken and egg problem. The biggest hurdle to overcome first. How you overcome it will affect the above three.
Have fun!
1) Ok, so let's say, you get ratio the number of men to no more than 20% of the women (overall/in certain geographic areas), and remove inactive profiles from search
2) That's fine - let's go for the low cost end. A fee that's half to 1/3 of the major sites, but for a whole year. No hassle of remembering to stop the billing, or checking your bills. Yes, this will mean after a year, you're likely to loose a high % due to not rebilling...but hey maybe this is a great product?
3) Fuzzy matching? i.e. I would like to meet someone <around> my height, or meet someone <slightly shorter> than my height. Perhaps we could have a feedback mechanism for those that actually meet. Perhaps we could just remove those people that never meet up with anyone? (Shocking approach I know - but the quality of the user experience would drastically improve!!)
4) Build it and they will come. (errm word play. bad)
But no seriously, whilst it's a competitive market, if you match up someone, and don't rip them off in the mean time, why on earth wouldn't they tell their single mates about the site? It's all about good user experience, which is sadly lacking in the majorly exploitative dating site game.
2) That's fine - let's go for the low cost end. A fee that's half to 1/3 of the major sites, but for a whole year. No hassle of remembering to stop the billing, or checking your bills. Yes, this will mean after a year, you're likely to loose a high % due to not rebilling...but hey maybe this is a great product?
3) Fuzzy matching? i.e. I would like to meet someone <around> my height, or meet someone <slightly shorter> than my height. Perhaps we could have a feedback mechanism for those that actually meet. Perhaps we could just remove those people that never meet up with anyone? (Shocking approach I know - but the quality of the user experience would drastically improve!!)
4) Build it and they will come. (errm word play. bad)
But no seriously, whilst it's a competitive market, if you match up someone, and don't rip them off in the mean time, why on earth wouldn't they tell their single mates about the site? It's all about good user experience, which is sadly lacking in the majorly exploitative dating site game.
1) Men are very aggressive, I would assume it needs to be something like 1:1.5 men:women ratio. But I do not have statistics.
2) Try it out with some market research!
3) Any sort of matching is going to fail if the data is not reliable. If I truthfully put 1.8m. Someone puts 1.9m when they're 1.7m. How do you determine that my input is true? Especially that since due to OkCupid, everyone is now automatically assuming the person is actually ~0.2m shorter. So if you put me and the other person together. The 1.9m person wins?
4) I can't tell if you have a plan for this. But a website that depends on a feedback loop seriously needs to overcome it somehow. Even PlentyOfFish for all the errors when I registered, I still tried to use it.
2) Try it out with some market research!
3) Any sort of matching is going to fail if the data is not reliable. If I truthfully put 1.8m. Someone puts 1.9m when they're 1.7m. How do you determine that my input is true? Especially that since due to OkCupid, everyone is now automatically assuming the person is actually ~0.2m shorter. So if you put me and the other person together. The 1.9m person wins?
4) I can't tell if you have a plan for this. But a website that depends on a feedback loop seriously needs to overcome it somehow. Even PlentyOfFish for all the errors when I registered, I still tried to use it.
4) Start in a small geographical area. A city. On the basis that most people are looking for a date close to them. Saw within 20 miles? Get it right there, then move to the next few closest cities. Use geographically targetted advertising for this - facebook perhaps, for online. Local papers could work for offline....or even maybe LoveFilm subscription drop-in-fliers?
I don't think improving on the current system is the way to go. There are too many fundamental mistakes. The entire system desperately needs an overhaul.
You don't have an email in your profile but if you'd like to chat shoot one over this way.
Cool - are you on twitter? I'm http://twitter.com/oceanician
interesting because it models the likelihood someone is going to reject a suggested person after the first few didn't work out.
This is recycled spam.
If you want the reality of online dating, look for the OKCupid blog post about how you're statistically more likely to find a partner and get married by not using online dating. Takeaway: men outnumber women, women are overwhelmed, men get no responses and become increasingly desperate. Of course, after Match.com bought OKCupid, the blog post was removed.
If you want the reality of online dating, look for the OKCupid blog post about how you're statistically more likely to find a partner and get married by not using online dating. Takeaway: men outnumber women, women are overwhelmed, men get no responses and become increasingly desperate. Of course, after Match.com bought OKCupid, the blog post was removed.
Is this the one?
http://web.archive.org/web/20101006104124/http://blog.okcupi...
Yes, exactly!
http://paulgraham.com/submarine.html
Similar article from 2008: http://www.nytimes.com/2008/01/29/science/29tier.html
Similar article from 2003: http://money.cnn.com/magazines/business2/business2_archive/2...