Graham McNicoll

It's entirely possible to run a test, see it win, ship the feature, and still damage your product. The metric moved. The experiment passed. Yet later you discover cancellation rate went up. Customer support volume went up. You just weren't measuring it. Short-term metrics can be easy to hit. That's the problem. You can make your numbers for the project, or for the quarter, and quietly erode the thing that keeps users around. This is how dark patterns get built. Not always with bad intent. Often just because the team was measuring the wrong thing and the data said yes. A complete picture means pulling in every signal you have. Clicks and monthly revenue are easy to measure. Support tickets, cancellation rates, user sentiment take more work to pull in, but they're what tell you whether the product is actually getting better. If your experiments can only see part of what's happening, they can only tell you part of the truth. What are the long-term metrics your experiments currently can't see?

Only one snapshot so far — the engagement-over-time curve appears once the daily scrape has captured this post at least twice.