[hcarvalhoalves]

I'm afraid the OP is jumping to conclusions.

You can't arrive to the conclusion that TDD reduced the number of bugs based on a percentage ratio between UI/server bugs. You could simply have a bigger share of backend bugs over time.

You need to see it in absolute numbers, using some meaningful metric (bugs vs. LOC maybe?).

[jdlshore]

Bugs vs. changed LOC could be a good one. I believe there have been studies showing that bugs are proportional to LOC, but don't quote me on that.

The nice thing about his data is that it's fairly well controlled (for software, anyway)--we can separate his contribution vs. others' and his code pre- and post-TDD.

I agree that his analysis isn't great, though. Also, it's hard to account for effect of legacy code. The big drop in bug count doesn't happen until 5 quarters after he started TDD.

[idunno246]

When you spend half your time writing tests you change half as much, and create half as many bugs. Admittedly a bit of exaggeration, as tests do help.