March 12, 2013

Hear Ye! Hear Ye! HackDay v0.9 hath officially concluded. If you don't know, PayScale HackDay is when the product team gets together, chugs coffee, eats donuts (or bagels (thanks Kayle)), and spends the day building whatever we want (as long as it has some relevance to PayScale). Sound interesting? Read on for all the details.

PayScale Welcome Screen

Since I write these blog posts, I thought I would highlight my own project. John Lynch and I had the idea to help people in the office "meet" and get to know our coworkers. So, we hooked in to our company's Yammer account to pull their picture, bio and interests. Our goal will be to install this app on a big screen in the lobby of the office.

(Sorry about the dialog in the background… I recorded the screen when my wife and I were watching the fourth installment in the Twilight Trilogy movies).

Do You Know What You're Worth?

Testing out Loggly

Engineer Chris decided to test out Loggly.com for the several gigs of log files we generate every day. We currently have a home-built solution for searching our log files, but services like loggly could be really good for us (because it means our engineers wouldn't have to work on our log file tools anymore)

Mobile Error Log Viewer

The technical people on our product team take turns being on-call in case the site goes down (or there is some other issue that needs to be sorted out). When it wakes you up in the middle of the night, we all think it would be really nice to find out what is going on before getting out of a comfy comfy bed. Emmett built a tool to view our error logs from your smart phone. Thank you!

Cleaning up Old Answer Values

We have a bunch of answer values with bad casing and too many or too few white spaces (commas without spaces around them). These basically only ever show up if you stumble upon them, but they look pretty bad. Data Analyst David wrote some python code to check all of our answer values for these common errors, as well as code to generate a suggested new version of the string. Even then, he had to go through and check them by hand, since there are valid answer values with abnormal casing (see: "PayScale"). The code I wrote cut the group of potential answer values to screen by hand down from 250k to 10k. In total, I identified 6,569 answer values with some sort of obvious syntax error that needed to be fixed (that’s about 3% of all answer values).

SalaryGuesser

Our newest engineer, (another) Adam, wanted to do some real-time-ish stuff with Node.js, so he made SalaryGuesser. It is a simple game where people can connect and play against each other to see who can most closely guess what PayScale knows to be the median salary of a given job title. The server randomly picks a job title, then tracks guesses. After enough time has elapsed, it posts scores and updates the connected clients in real-time, then picks a new position to quiz players on.

Determining Compensability with R

In the world of compensation, we’re not sure what questions we ask will end up being able to predict pay. So we ask questions we think might be compensable, and then after we collect a lot of data for that question to see if these answers have any predictive power of pay.

Say we add the question “How long does it typically take to close a deal after your first meaningful contact?” Once we’ve collected enough data, we need to analyze how this question affects pay overall and by individual jobs. Doing this type of an analysis can be tedious since you have clean and order your data, move it into a statistical software program and then start typing formulas over and over again. If you’re testing lots of questions this can be very tedious.

Ryan, a data analyst, created a tool that does all the leg work for us saving us hours of time doing tedious work so we can focus more on the “Is this question in its current form predicting pay?” part of the problem.

Rich Researh Center Search Results

Joe added rich info to the employer and job search results. We can now show salary data, popular employers, and job satisfaction for job search results and relative pay, employer satisfaction, and logo for employers.

PayPong

What software startup would be complete without a ping-pong table and a healthy dose of competitivness? Scott built an internal page for tracking who wins and loses, and lets you set up challenges (sending the email for you!)

Shower Hax0rz!

The lip at the bottom of our shower just barely blocked water, and only if you were lucky. Geary's hackday project was to solve that. He made a bigger barrier at the bottom. Thank you!