I have seen Acceptance Criteria treated like a formality and like a real craft. One produces green statuses, the other produces confidence people can explain.
The most common mistakes I see around Acceptance Criteria are rarely caused by laziness. They come from time pressure, fuzzy ownership, and the comforting idea that past success will repeat itself. It gets expensive when everyone agrees on the story title while quietly imagining different behavior underneath it.
A weak QA habit often hides inside work that looks efficient on the surface.
Mistake One: Testing the Shape Instead of the Risk
Teams mirror the implementation too closely. They test the visible steps, but they do not test the part that could do the real damage. With Acceptance Criteria, that usually means the team can demo the feature but has not really challenged shared expectations before coding begins.
Mistake Two: Trusting Default Conditions Too Much
Friendly data and stable environments create a polished story that reality does not honor. A seemingly simple requirement becomes three different implementations depending on who reads it is exactly the sort of thing that disappears when setup is too clean.
Mistake Three: Writing Down the Result Too Late
Teams often discover the right insight but never capture it well enough for the next decision. By the time sign-off starts, nobody remembers which uncertainty was tested and which was only assumed away.
What I Do Instead
- Name the most expensive failure in plain language before testing begins
- Pull in the right product, design, development, and QA when the risk depends on business context
- Record the few facts that made the decision easier, not every action that happened
- Treat unclear evidence as its own finding instead of polishing it into confidence
Those habits keep Acceptance Criteria grounded in outcomes rather than ceremony. That is usually when confidence becomes visible enough to share, not just feel.