Company Information / Message from the CEO
Out to Change
I really appreciate everyone who helped with this project and all the members who joined and communicated.
eChemistry, as a project, is ending and the energy that went into maintaining it is being focused
entirely on personality and romantic chemistry research. You may read through the debrief of what happened below
or you may jump directly to the new website to see how the ongoing personality research is going. As of today,
eChemistry as a website is here mainly for historical reference.
The good news is that nearly 170,000 people took the personality test here, providing a huge cache of data
with which to pick apart the accuracy of the personality test questions, and a huge amount of personality feedback with which
to hone the personality research.
Why No More Dating
It's really hard to know where to start, so, in no particular order, here is a sampling of why this project is changing focus.
Running a dating site involved
a HUGE amount of server and network maintenance. I overbought computers up front because I was so sure we'd need to handle
so many users at once, and I should've put this website on one dedicated server to start and then fixed that problem when the user load
got too high. Instead I had more than 20 servers from the first day which took forever to set up and then were constantly
having problems. Multiplying the servers only multiplied the headaches and therefore decreased the quality of the user experience.
It was a total shock how many users required daily, personalized customer service emails to discuss their
lives and how many matches they were getting and how we should change the system to focus more on people's horse preference because
that is 'really obviously critical' to matching. Users were certain they'd been 'black balled' if they didn't get a match on
any random day and wanted to know why we hated them. If I had responded to every customer service email with people monologuing about
their lives and wanting to be heard and reassured, there would have been no other maintenance of this project at all. To be clear,
for those who wonder, there was no 'black ball' system. Some people just didn't get many matches if their preferences were too tight or
if no one near them signed up that day.
The website actually made it really easy to become a member. The cost of $10/month was only to make people appreciate
the value of a membership.
Any person who paid for even one month got a free lifetime membership, although we never said that anywhere.
There was nothing in the code which
reverted people's membership status after their term ended. $10 was so much cheaper than most of the competition, and so far below
the actual cost of what went into each membership, that I thought it was small enough to be the tiniest hurdle, but the website never
had a month where costs were covered by membership fees. $10 was way too high and it kept too many people from communicating.
The planet is huge and people's desired distance to a match is amazingly tiny. If you do the math, in order to
have enough members so that there is exactly one member in every ten mile radius, so that any new joining member has at least one
other human joined who might be the wrong sex and personality and age and religion, it takes 182,000 members just to cover the
land surface area of the planet. We had 170,000 members, and we screened for sex, age, and then screened out most potential matches
using personality, so we were left with very, very few matches per user and the numbers just weren't enough to keep people engaged.
I had a little too much faith in the ability of the data mining math that we were using to help with getting
people's personalities right. The personality test on this website was the best that has ever existed, without a doubt, but
getting that many
traits right for both sides of a romantic match, all using a self-reporting written personality test let just enough variation in
that the resolution didn't get high enough fast enough with this version.
The math and programs are incredible and definitely helped, but accurately categorizing multiple personality traits required
more personality research
more than mathematical, statistical analysis to solve. Calling people to verify their personalities showed
just enough error in the test all along that it was never quite delivering
what we hoped because the problem it's trying to solve is so incredibly complex.
This is a big reason why the next obvious step is to keep working on the personality part, and take away the distractions
of running an online dating site with over a hundred thousand users, many of whom want to chat every day.
I paid for everything here myself. It was all my savings. I did that so that no one was ever at risk but me.
For example, shutting down this website today requires consultation with no one. All of the memberships have run out and I can turn
off the server and no one lost a penny but me, who lost an incredible mountain of pennies. With that said, I know some people
will argue that more advertising should've happened to bring in the critical mass of users, which is apparently somewhere over 170,000,
but the conversion rate I experienced was SO low that I don't see how spending more would've resulted in a profit at all,
much less enough to hire people to work on all the problems. I know theoretically
the tipping point would be hit and there would be enough users so more would pay, but I'm not sure there was a real-world number
where that would've been anything more than lost money. On the day this website launched HUGE money was going into a competitor, True.com,
and on the day this website is shutting down True.com has been gone as a dating website for a really long time. I think I just
lost less money than they did by spending less on advertising. Both Plentyoffish and OKCupid get by with only advertisements, but I showed
up after they did and therefore had to advertise to get anyone to this website, and then people got too few matches because of the personality
filter and therefore didn't stick around enough for matching to really catch on here.
The personality work continues at the new website and it's going incredibly well now
that the dating website isn't a distraction. Maybe eChemistry will come back someday, but the problem of human personalities
is compelling and endless enough that it currently doesn't look like it'll be me who brings it back to life if it does return.
To be clear, the pattern of who falls for whom, personality-wise, is still really clear. The hard part is automating the correct
acquisition of people's trait profiles. It turns out that even with fantastically accurate classifications, getting people to understand and report which one they
are is still an incredibly difficult task. Then collecting enough accurately profiled people to get matches, and then getting them to give another human a real, honest chance just compound
what really should be a straightforward process.
In the meantime, I'm grateful to everyone who gave this website a try. I'm grateful to everyone who filled out
the personality profile. I'm grateful to everyone who did long interviews with me on the phone and in person to verify their
personality results. I'm grateful to everyone who volunteered to help with this project. Hopefully romantic chemistry will be resolved
soon for everyone, but it looks like it might be easier to teach everyone how to do it than to do all of it myself. It'll take
some time to really make the research clear, but eventually I intend to get it all out so that whomever wants to pick it up and help
can do so.