Y Combinator Challenge #14 - Tools for Measurement | Hacker News Reader

Y Combinator Challenge #14 - Tools for Measurement

astartupaday.wordpress.com

18 pointsinglorian16 comments

Its not a bad idea, but I think it misses the potential that YC was suggesting.

Read the book 'Making Things Talk' - think about the potential of measuring things in the analog world, temperature, weight, flow, force, sound, etc, and correlating these things with enterprise data.

Maybe you can start off by measuring things like heartrate, blood glucose levels, physical mobility, dietary intake, etc. For instance, if someone is stressed all day, this probably is going to correlate with productivity.

That books looks sweet, thanks for pointing it out. Reminds me of a good post by Seth Godin I read the other day: http://sethgodin.typepad.com/seths_blog/2008/07/let-me-see.h...

I started down this path, looking at things like: - number of people who Emailed an individual asking about a specific topic -percent of time spent in Outlook vs Visual Studio vs WoW (RescueTime-esque stuff) - number of IMs initiated from someone at least one level higher than them

But at the end of the day, can that really tell you quantitatively that Bob is better than Suzie? Even the example above about being stressed - I know people who are super chill and excel at their job, and others who are stressed all the time and also excel (and vice versa).

Maybe the answer is to pull in hundreds of these variables and run a neural net against actual performance rankings to see if any statistically significant correlations pop out?

Hell yea its sweet! :) I initially was pointed there by this great presentation: http://www.tomtaylor.co.uk/talks/delighting-with-data

I think you are right about pulling in hundreds of these variables to correlate with both intuitive and quantitative tests that we have to define 'productive' employees. Then you can send an electric shock whenever someone's theta waves get too low or something..

Seriously though, our first level of productivity is dictated by things biological, and if you can create models to paint a strong enough picture of a productive worker then it gives you a feedback loop. And don't forget, not everyone is a knowledge worker like us, people still drive trucks for a living, pick fruit, do construction, etc. That is a huge market.

I was just thinking about the same things earlier today. I thought of doing an experiment where different volunteers are fed different amounts of food, and then give them standardized tests to see what the effect (if any) is of food on brain function that same day.

It's probably been done. If anyone knows of something like this, please let me know.

I would also like to mention an anecdote; a teacher once told me a student of his bet he'd get positive results from taking LSD (or something, it might have been extasy) on the day of the PSU, the once-a-year standardized test that determines admission to Chilean colleges. He flunked.

This idea if implemented would measure niceness (political sleekness) but not productivity.

Not sure about that. There was a guy who knew his stuff and ran the build and version control system at my last job. You had to kiss his butt to get things done (=pay karma), but I wouldn't call him nice.

What it overemphasizes is jobs that benefit multiple people in visible ways. At my last job, I wrote code used by the internal QA team of 50 people that made their lives easier. I imagine that would earn a lot of karma. On the other hand, when I worked on a bug for an external customer, only my manager would care.

I think his second assumption that people know who the high performers are is wrong. Often, recognizing competence requires a great deal of competence.

Also, I think d0mine's right that kinder people who are easy to work with but not necessarily great performers will get lots of points.

Well, being easy to work with is in itself a valuable trait in a co-worker.

By "easy to work with" I mean qualities such as being amiable, outgoing, funny, or good-looking. Those are valuable traits, but pointing out who's the nicest guy in the room's not the problem here, and it's not even a difficult problem to solve. Everyone always knows who the popular people are.

Right, but the proposed solution wouldn't involve asking people which of their co-workers they like, it would be based on hopefully meaningful work-related interactions (emails). Obviously, the model breaks down if people give the thumbs up to forwarded joke emails or to mass "oh crap my review is coming up, can you give me a thumbs up" emails.

I think this implementation would work at least as well as state of the art modern management techniques do. It's a decent idea.

I've daydreamed about something similar. Not sure whether it would be better or worse, but my implementation would limit the amount each person can give, per month. I also struggled with the idea of making it a token economy. Thoughts?

I think the token economy thing is a good idea. You could also do a market approach of rewarding people who pick the top performers, but that gets really tricky.

All the research on token economies shows they don't work, at least according to Punished by Rewards.

That's an interesting looking book. I'm familiar with the whole intrinsic/extrinsic motivation arguments. I guess it depends on the definition of "work". In this case the economy would be intended as a measurement tool more than a motivational tool. However, I suppose you could still make a case that it would become demotivational, which would be at cross purposes.

Tell folks that you're setting up a "bet your job" project and you'd like to know who else they want on the team.

This implementation would be utterly trivial to game.

Fail.

Reply on news.ycombinator.com