Reflection on Effort Estimation

04 May 2026

---

Where did the time go?

My effort estimates were mostly based on how hard I thought each task would be and whether I had done something similar before. For simpler things like UI pages, the estimates were usually closer, but for anything involving authentication, databases, or integration, I underestimated pretty badly. A good example is user auth and courses, where the actual time ended up being much higher than expected because of debugging and edge cases . I didn’t use any formal historical data, just past experience and a rough guess of complexity.

Even though the estimates were often off, they were still useful. Having a number in mind made it obvious when something was taking way longer than expected, which helped me realize where the real difficulty was. Tracking actual effort was also helpful because it showed a pattern of underestimating anything that wasn’t straightforward coding, especially debugging and setup. I tracked time manually in minutes, and while it wasn’t perfect, it was accurate enough to see trends. If I did this again, I’d adjust my estimates to account for debugging upfront instead of just the coding time. I didn’t really use AI for estimating or tracking effort, so it didn’t play a role in this process.