feat(statistics): holiday baseline — compare a holiday with the same holiday in earlier years #205

New issue

Open

opened 2026-10-02 15:41:49 +00:00 by gabogg · 0 comments

gabogg commented

2026-10-02 15:41:49 +00:00

Owner

Idea

Give each Holiday its own reference: the same holiday in earlier years ("this Christmas vs last Christmas"). Today a holiday can only be compared against the Usual Weekday Baseline. That shows the holiday differs from an ordinary day, but not whether it was a good holiday.

Context

Decided while triaging #202 (2026-10-02):

Holidays and Event Days are expected, not ordinary.
They stay in every total and trend, and stay out of the ordinary-day references (Usual Weekday Baseline, learning, spike check).
A separate Holiday Baseline is a future capability, outside the historical backfill milestone.

It needs about a year of history, which the historical import (#170/#171) could provide.

To settle

How a holiday is matched across years: by name, by fixed date, or by a movable-holiday rule. That depends on #204 for Carnival, Easter and the like.
The minimum history, and what to show when there isn't enough.
Eligibility, mirroring the Usual Weekday Baseline: counted flow, trusted, no ingestion gap. Do the same opening hours matter, and do imported days count?
Do Event Days get no such reference? #202 decided events get no "same event last year".
Where it shows: the day view, investor analytics, or exports.

Blocked by: #204 (movable holidays), for anything beyond fixed-date holidays.

Refs #161, #202.

## Idea Give each Holiday its **own reference**: the same holiday in earlier years ("this Christmas vs last Christmas"). Today a holiday can only be compared against the Usual Weekday Baseline. That shows the holiday differs from an ordinary day, but not whether it was a *good* holiday. ## Context Decided while triaging #202 (2026-10-02): - Holidays and Event Days are expected, not ordinary. - They stay in every total and trend, and stay out of the ordinary-day references (Usual Weekday Baseline, learning, spike check). - A separate Holiday Baseline is a future capability, outside the historical backfill milestone. It needs about a year of history, which the historical import (#170/#171) could provide. ## To settle - How a holiday is matched across years: by name, by fixed date, or by a movable-holiday rule. That depends on #204 for Carnival, Easter and the like. - The minimum history, and what to show when there isn't enough. - Eligibility, mirroring the Usual Weekday Baseline: counted flow, trusted, no ingestion gap. Do the same opening hours matter, and do imported days count? - Do Event Days get no such reference? #202 decided events get no "same event last year". - Where it shows: the day view, investor analytics, or exports. Blocked by: #204 (movable holidays), for anything beyond fixed-date holidays. Refs #161, #202.

gabogg added the

needs-triage

label

2026-10-02 15:41:49 +00:00

gabogg referenced this issue

2026-10-02 15:42:07 +00:00

feat(schedule): movable holidays defined by recurrence rules #204

gabogg referenced this issue

2026-10-02 15:42:51 +00:00

Should holidays be excluded from statistics, or are they expected traffic? #202