Sextant
← Catalog
Block · intelligent

Cohort Insight

Lower-triangular weekly cohort retention grid with auto-detected best/worst cohorts. Pure CSS Grid, no Recharts.

State
Size
Mode

User retention by signup week

50%

overall mean retention · 8 cohorts × 8 weeks

Cohort May 22 showing 82.0% D7 retention — 32 pts above baseline.

Wk0
Wk1
Wk2
Wk3
Wk4
Wk5
Wk6
Wk7
Apr 17
100
62
38
24
15
9
6
4
Apr 24
100
63
40
25
16
10
6
May 1
100
64
41
27
17
11
May 8
100
66
43
28
19
May 15
100
67
45
30
May 22
100
82
65
May 29
100
69
Jun 5
100
Cells · weekly retention %
0%
100%

Analytics · updated hourly

Demo data is illustrative. Replace with your own typed data prop.

How it computes

Cohort × week retention grid with per-cohort and per-week means; best/worst cells excluding the trivial week 0.

grid mean = Σ values / count (sparse cells skipped);  best/worst = argmax/argmin over weeks > 0

Assumptions

  • Cells are [0,1] retention fractions; a missing cell means "not yet aged".
  • Week 0 is ~1.0 for every cohort and excluded from the extremes.
  • Weekly means use only the cohorts old enough to have that week.

Honest about

  • Best/worst deliberately skip week 0, where every cohort is trivially 100%.
  • Sparse younger cohorts are tolerated and simply omitted from later-week means — no interpolation.
  • A 0% cell still renders at 6% alpha so it reads as present rather than blank.