← all features
007

Correlation Engine

19 auto-computed metric pairs with Pearson analysis and plain-English interpretations.

xeve automatically runs statistical correlation analysis across all your data streams. Sleep vs productivity. Music vs focus. Steps vs coding output. It discovers relationships you would never notice on your own and explains them in plain language.

01
01

19 metric pairs

Pre-configured pairs include sleep vs productive time, steps vs coding, music listening vs focus, heart rate vs app switches, and more.

02

Pearson correlation

Uses Pearson correlation coefficient (r) to measure the linear relationship between metric pairs. Values range from -1 (inverse) to +1 (direct).

03

Plain-English interpretation

Every correlation comes with a human-readable explanation. Instead of "r = 0.72" you see "Strong positive — more sleep strongly predicts more productive time the next day."

04

Auto-updating

Correlations are recomputed daily as new data flows in. Relationships become more statistically significant over time as sample sizes grow.

05

Minimum data thresholds

Results are only shown when there is enough data for statistical validity. The engine requires at least one day of overlapping data between metric pairs.

02

Correlations are computed via a Supabase RPC function (compute_correlation) that takes two metric names and a user ID. It joins daily_summaries with itself on date, extracts the requested metrics, and computes the Pearson correlation coefficient using the standard formula: r = Σ((x-μx)(y-μy)) / √(Σ(x-μx)² × Σ(y-μy)²). The function returns the coefficient, sample size, and p-value. The web dashboard fetches all 19 pairs in parallel and displays them ranked by absolute correlation strength.

03

Does correlation mean causation?

No. Correlation measures the strength of a relationship between two metrics, not whether one causes the other. A strong correlation between sleep and productivity suggests a relationship worth exploring, but other factors may be involved.

How many days of data do I need for meaningful correlations?

The engine can compute correlations with as little as one day of overlapping data, but results become statistically meaningful with 7+ days. After 30 days, you will see reliable patterns with strong confidence levels.

Can I add custom metric pairs?

Not yet. The 19 pre-configured pairs cover the most useful combinations across productivity, health, coding, music, and activity data. Custom pairs are planned for a future release.

early access

get started free

free during early access. no credit card required. install the mac app and your dashboard is live in seconds.