THE RESULT
What the data shows
The largest computed metric is 0.4208 for (30.0, 90.0]; the smallest is 0.09157 for Never previously contacted. Metric: rate (Subscription share).
The -1 sentinel is not a negative elapsed time. Small groups have wider uncertainty.
| recency band | events | n | rate | wilson lower | wilson upper |
|---|---|---|---|---|---|
| (-0.001, 30.0] | 28 | 188 | 0.1489 | 0.1051 | 0.2068 |
| (180.0, 365.0] | 786 | 4,416 | 0.178 | 0.167 | 0.1895 |
| (30.0, 90.0] | 223 | 530 | 0.4208 | 0.3794 | 0.4632 |
| (365.0, inf] | 188 | 643 | 0.2924 | 0.2585 | 0.3287 |
| (90.0, 180.0] | 680 | 2,480 | 0.2742 | 0.257 | 0.2921 |
| Never previously contacted | 3,384 | 36,954 | 0.09157 | 0.08867 | 0.09456 |
THE METHOD
From source to answer
Preserve pdays=-1 as a sentinel category before computing rate intervals.
The Python source exposes this study’s transformations. The complete project download includes shared preparation and evaluation routines.
THE NEXT DECISION
What follows from the finding
Keep never-contacted records distinct when evaluating engagement history.
Where the conclusion stops
Observational marketing records do not identify campaign uplift. Month and row order are not precise timestamps; repeated-client identifiers are unavailable. Models are research diagnostics, not financial eligibility or automated contact decisions.
Related studies may reuse observations or holdouts. These are historical analyses; associations and backtests do not demonstrate commercial impact. Further model tuning needs new, untouched evaluation data.