Pixel-lytics11.07.13 · go
Before graduate school I worked at Caterpillar’s engine division. I learned a lot about product development and after four years, the smell of diesel fuel always makes me smile.
Caterpillar encourages a data-fearing culture—and for good reason. When you’re designing multi-million dollar machines, you can’t eyeball it. As a result, prototypes are heavily instrumented with accelerometers, load cells, GPS, thermocouples, microphones—the works.
But whether you wire up 50 sensors or 100, no one is ever satisfied. Powertrain wants higher sampling rates. Chassis demands more strain gauges on critical joints. And Electronics will just die without at least 15 leads on the ECM.
For my team, in charge of pollution & exhaust controls, data was a luxury. We were an internal startup without a big budget, test priority, or 80+ years of history to lean on. We had to do more with less. That meant simplifying our sensor map down to the most critical channels and focusing on specific failure modes. Our focus paid dividends in the field - our product performed excellently with few warranty claims.
Since then, I’ve traded CAD for plaid & micrometers for MacBooks, but the lesson remains. You don’t need to measure everything just because you can.
The pixel server, written in Go, can reside anywhere and logs every request.
```sh # start server on port 8080 go run pixelog.go
2013/11/05 22:33:50 188.8.131.52 /t.gif?param1=X¶m2=Y “Mozilla/5.0 (Macintosh; Intel Mac OS X 10_7_5) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/30.0.1599.101 Safari/537.36” http://app.server/pixel_test/ ```
In one line, you’ve got the answers to 5 key questions:
- When did the user access the page? (date & time)
- Who are they? (user ip)
- Where are they on the site? (request url)
- What browser are they using? (user agent)
- How did get here? (referrer)
While instrumentation costs for web & mobile (Google Analytics, MixPanel, etc) are effectively zero, measuring everything isn’t a panacea. Meaningful analyses take time & money, and the learning curve is steep. Every time I open Google Analytics I feel like a deer caught in the headlights. Instead, focus on one particular problem and gather the data which validates or disproves your hypothesis.
Join the discussion