[encoderer]

Another peril with making engagement a north star kpi is that you quickly get into a local design maxima. I saw so much of this in my time at a well liked consumer tech company. It becomes almost impossible to evolve around new business requirements because the current thing has been perfectly “optimized” for “engagement”.

Here’s an example.. imagine a photo carousel with a critical button beside it that drives revenue for the business. At some point a PM spent a few months “optimizing” the photo carousel. Subsequent changes all tanked engagement on that magic button. Experiment after experiment all thrown out because they didn’t result in enough button clicks, despite huge improvements in photo experience.

Eventually I took over the team that owned those little revenue buttons and even more eventually I figured it out:

A lot of users fly through the photo carousel clicking the right-arrow very fast. At the final photo the right arrow disappeared and there was a small layout shift and it put a portion of the revenue button to where a portion of the “>” button used to be. Many many users accidentally clicked that revenue button.

You can’t a/b test yourself to PMF and organic growth, even when used appropriately it is only accurate but not precise.

[ant6n]

If u want to maximise revenue, wouldn’t it make sense to measure the revenue coming from those button clicks as well as the button clicks? Perhaps that would be more precise.

[encoderer]

2-sided marketplace, revenue not directly attributable