Developers love metrics. Uptime, latency, test coverage, lines of code — we measure everything about our systems. But when it comes to measuring ourselves, most of us are flying blind.
Over the past year, I've been experimenting with tracking different aspects of my dev life. Some metrics turned out to be game-changers. Others were pure noise. Here's what I learned about what's actually worth your attention.
The Metrics That Changed Everything
1. API Token Spend (Real-Time, Not End-of-Month)
If you're using LLM APIs — Claude, GPT-4, Gemini — you're probably checking your bill at the end of the month and wincing. That's like checking your bank balance once a month and wondering where the money went.
I started tracking token usage in real time using TokenBar, a menu bar tool that shows exactly what each API call costs as it happens. The insight wasn't just "I'm spending too much." It was where the waste was hiding: oversized system prompts, redundant context windows, and agents that loop without producing value.
Tracking spend per-request changed my prompting habits more than any optimization guide ever did.
2. What I Eat (But Not How I Think)
This one surprised me. I started photographing my meals and letting AI handle the nutritional breakdown instead of manually logging every ingredient. The friction reduction was massive — I went from tracking 2 days a week to tracking every single meal.
The metric that mattered wasn't calories. It was protein intake. Once I could see I was consistently under-eating protein by 40g/day, I made one adjustment (added a shake) and my energy levels during afternoon coding sessions noticeably improved.
Manual calorie counting is busywork. A quick photo that gives you the macro breakdown is signal.
3. Distraction Patterns (Not "Screen Time")
Screen time is a useless metric for developers. Of course I'm on my screen for 10+ hours — that's my job. The metric that actually matters is what pulls me out of flow state.
For me, it was always the same pattern: I'd open Twitter "for a second" to check notifications, then 20 minutes would evaporate. The app wasn't the problem — Safari is a legitimate work tool. The feed was the problem.
I started blocking feeds specifically (Twitter's timeline, Reddit's front page, YouTube recommendations) while keeping the rest of those sites functional. The difference was immediate. My longest uninterrupted coding sessions went from ~45 minutes to 2+ hours.
The Metrics I Stopped Tracking
Lines of code per day. Meaningless. Some of my most productive days involved deleting 500 lines.
Hours worked. Led to performative busyness. Sitting at my desk for 12 hours doesn't mean I shipped anything.
GitHub contribution streaks. Turned coding into a chore. Broke the streak intentionally and felt relieved.
Task completion rate. Optimized for easy tasks instead of important ones.
The Rule I Follow Now
A metric is worth tracking if it changes your behavior within a week. If you look at a number and it doesn't make you do anything differently, stop tracking it.
Token spend made me rewrite prompts. Protein intake made me change my diet. Feed-blocking data made me restructure my browser. Those are metrics that earn their place.
Everything else is just noise dressed up as data.
What do you track that actually changed how you work? Genuinely curious — drop it in the comments.
Top comments (0)