Technique effectiveness

Every technique in the library claims to help with something. This insight is Therapy's attempt to check that claim against your own data, rather than asking you to take it on faith — did your mood, in practice, tend to move after you used a given technique?

What it actually is

Whenever you work through a technique with Claude (or the in-app chat), that's logged with a timestamp. Separately, your mood check-ins are also timestamped. This insight lines the two up: for each time you used a technique, it looks at your mood ratings in the roughly 48 hours that followed, and compares that to what your mood looked like without a recent technique in the mix. Do that across every technique you've tried, and a rough ranking emerges — which ones tend to line up with a lift, which tend to look flat.

How it's computed

This is closer to a statistics exercise than an AI-generated read: mood ratings before and after each technique use are averaged and compared, technique by technique, using your check-in data specifically — not data from other users. There's no model inferring intent or effort here, just numbers being grouped and compared. The 48-hour window is a fixed choice, not something that adapts per technique, so it will occasionally miss a slower-building effect or catch a mood shift that had nothing to do with the technique at all.

An honest note on sample size

This is the part worth being direct about: correlation from a handful of data points isn't a reliable signal, and this feature will tell you that plainly rather than dress up a thin result as a confident one. If you've used a technique once or twice, any "effect" you see is closer to noise than a finding — a good night's sleep or a stressful meeting the next day will swamp it completely. The dashboard marks results as low-confidence until there's enough usage behind them, and it would rather show "not enough data yet" than a specific-looking number built on three data points. As you use techniques more, the picture sharpens; a technique you've tried fifteen or twenty times, spread across different weeks, gives this insight something real to work with. Until then, treat any ranking as a loose hint, not a leaderboard.

How to read it

Use it as a nudge for what to try again, not as a verdict on what to abandon. A technique that shows a flat or even negative pattern in a small sample might still be exactly the right one for a specific situation — the data here is coarse, and it can't see context like how hard a week you were having when you reached for it.

Beyond the dashboard, your top-performing technique may now appear proactively in your chat greeting or in the lead section of your weekly or monthly progress report — a reminder grounded in the same computed statistics, never an invented claim. It's still subject to all the caveats here, including sample size and the limits of a 48-hour window.

What it can't tell you

It can't tell you why a technique helped or didn't — mechanism is outside what this measures. It also can't separate a technique's effect from everything else happening in your life during that 48-hour window. Read alongside themes & topics, it's a starting point for a conversation, with yourself or a professional, not a substitute for one.