Heatmaps vs session recordings: pick your poison

Heatmaps vs session recordings: pick your poison
Needed to know WHY my pricing-page variant flopped. Day 2, two tools open.
Hotjar heatmap: aggregate. 1,000 users, one pretty scroll map. Told me 60% never reached the CTA. Fast diagnosis, zero detail.
Session recordings: I watched 12 actual humans. One rage-clicked a fake button (it was just bold text). THAT was the killer. Heatmap never showed it.
Tradeoff is time. Heatmap = 2 min, directional. Recordings = 40 min, but you find the weird thing.
My flow: heatmap to find WHERE it dies, recordings to find WHY. Not either/or.
Next failed test, don't just rerun it. Go watch 10 recordings of the losing variant first. Report back.