Breakfast shop scheduling supervisor - the weekly schedule-building SOP: pull data Thursday, lock anchor shifts first, back-schedule the dawn crew, publish with confirmation - an archivable weekly procedure

Please organize our breakfast shop's weekly schedule-building procedure
into an archivable SOP for whoever takes over scheduling. Below is how
we actually do it. Structure it as a proper SOP; do not change the
procedure itself - only fill in each step's output, judgment criteria,
and red lines.

Procedure (in order, executed every week):
1. Thursday evening, pull three datasets: last week's order volume
   per half hour (POS export), the skill matrix (who can do what:
   soy grinding / dough mixing / wrapping / deep-frying / griddle /
   register), and next week's availability declarations (approved
   leave, student part-timers' class hours)
2. Set anchor shifts first: every operating slot, every production
   line (soy line / steamed line / fry line) has at least one person
   who can run it - anchors outrank headcount; better one short in
   the lull than an anchor missing
3. Back-schedule the dawn crew: from the first batch at door-opening,
   work backwards to each opening task's latest start, and set
   arrival times from that; arrival may only be earlier than the
   latest start, never exactly on it
4. Fill support shifts: rush-hour slots (drawn from last week's
   volume) get register support, packing, restocking; off-peak
   keeps anchors only
5. Hours check: part-time single shifts capped at [A] hours;
   full-time weekly hours capped at the shop's own limit [B];
   consecutive early days capped at [N] - split the shift if over
6. Keep one floater: one on-call backup per day, off the schedule -
   the list is maintained separately, with consent obtained in
   advance and callable windows noted
7. Publish Friday noon: post to the group chat, each person replies
   to confirm individually; anyone unconfirmed gets a phone call
   before Saturday's market; any post-publish change re-confirms
   everyone affected
8. Archive: file this week's schedule together with the "exception
   log" (no-shows, ad-hoc swaps, rush-hour gaps) for the month-end
   review

Organization requirements:
- Each step states: who does it, from what data, what the output is,
  and what to do on exception
- Red lines get their own section: any unmanned operating slot
  (table is void, rebuild), any single-skill dependency (must record
  who is shadow-learning and expected completion), consecutive early
  days over N
- Use shop-floor language; no HR jargon ("human capital," "flexible
  workforce")

Constraints:
- Do not add "employee care" or "team-building" steps the procedure
  does not have
- Do not invent time quotas for any step - durations are per-shop
  measured facts; leave [ ] variables in the SOP
- The SOP is for the person taking over: no industry background
  sections, no scheduling theory

Why it is written this way

The skeleton of this SOP is "anchors before headcount." Generic scheduling tutorials teach you to total the person-hours first and cut shifts second; a breakfast shop that does this fails every time: six people, adequate headcount, nobody who can grind soy - and no first pot at opening. Headcount is an off-peak concept; anchors are an opening concept. Step 2 deliberately places anchors before any filling, and only step 3 back-schedules arrival times, because once the order is reversed, whoever builds the table starts treating "headcount complete" as the finish signal. Within step 3, "arrival may only be earlier than the latest start, never exactly on it" builds in the buffer: proofing time drifts with temperature, and scheduling exactly on the line hands that drift to luck.

The confirmation regime and the on-call floater (steps 6 and 7) move the cost of the dawn no-show upstream. People who individually confirm their shifts genuinely no-show less - the act of replying is itself the reminder that "I have shifts next week." The floater list is the crisis insurance: consent obtained in advance with callable windows noted means the five-a.m. phone call does not become a crawl through the contacts list from the top. Together they prevent most dawn crises rather than rescuing them with heroic phone skills.

The constraints explicitly ban care-and-team-building additions and invented time quotas, and the reason is that an SOP gets taken literally: every paragraph of content the real procedure never had costs the document some of the executor's trust, until even the true parts stop being followed. Time quotas are the sharpest danger - an invented "20 minutes for dough" gets enforced as shop law, and the day the dough misbehaves, the rule's provenance turns out to be a language model's guess. Every duration and cap stays a [ ] variable for the person taking over to fill with measured numbers.

The exception log (step 8) is this SOP's only output beyond the schedule itself: no-shows, swaps, and rush-hour gaps jotted down weekly become the entire evidentiary basis for adjusting anchors, red lines, and the floater list at month-end. Without it, the review runs on memory - and memory only ever contains last week.

Variables