How I Corrupted Survey Results and (Maybe) Ruined a Business
summary
A 25‑question survey over five pages is implemented as seven PHP pages that write straight to MySQL with no validation, so answers vanish; about a hundred surveys start, roughly 30% dropoff, and only ~70 full answer sets appear. Missing responses get quietly inserted into the database to match start/finish markers, producing a dataset the client uses to steer the business. The system is brittle, manual and deceptive, and the project technically “works” only because results are fabricated rather than fixed.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:04Professional failure
- 4:15Personal traits
- 9:51Survey project begins
- 13:02Cover-up of data
- 14:49Leaving toxic work
- 15:52Lessons learned
- 20:12Better engineering culture
- 23:45Evaluating company culture
- 26:36Key takeaways
- 27:40Q&A
related talks
45mI Used Drupal Core and the Form API to Build a Mind Reading Machine
55mBuilding Scalable PHP Apps With Google App Engine
35mPHP is evil (Defensive programming)
45mPHP: Particularly Horrible Programs
37mHow I Accidentally Wrote the Best Code of my Career
58mTesting Complex Applications for PHP7
36mProcessing One Billion Rows in PHP
35mSlow database in your PHP stack? Don't blame the DBA!