Nobody could answer my engineering options until I rewrote them as plain questions.
This week I built an end to end role regression suite for a client's marketplace. Every role in the app: visitor, buyer, seller, agent, auctioneer, professional, admin, superadmin, plus cross account. 272 tests, green twice, split into 219 passing, 34 tied to known defects, 6 skipped, and 13 waiting on a product decision.
Writing it surfaced around 70 suspected defects. I filed 25 as new issues. It caught three previously fixed issues that had quietly regressed. That alone paid for the work.
But here is the part I keep thinking about. The suite surfaced 11 things that were not bugs at all. They were product decisions nobody had made. Should an admin be able to strip a superadmin. Can users self assign roles. Should an owner be able to delete their own draft listing.
I wrote those up as engineering options. Careful, precise, full of the tradeoffs. Sent them to the client's builder. He could not parse them. They just sat there.
So I rewrote every one as a plain question with a recommended answer. Should an owner be able to delete a draft listing? I recommend yes, here is why. Same eleven decisions. He moved through all of them the same day.
That is on me, not him. I dressed a business question up in engineering clothes and then wondered why the business could not answer it.
Most of the decisions that stall a build are not hard. They are just phrased for the wrong reader. If you want an owner to decide fast, hand them a yes-or-no with your recommendation attached, not a menu of tradeoffs.
A good engineer translates. He does not just report.
AI Diagnostic | All insights