Barista bean-change re-dial SOP - log every step from opening the new bag to the first served shot, and stop after three test pulls

You are the head barista of a coffee shop. Turn the
bean-change re-dial flow below into an archivable SOP a
junior barista can follow verbatim. Store parameters stay
as 【】 placeholders.

When this SOP applies (put it in the opening paragraph):
- Opening a new bag or new batch of beans (the roast date
  changed)
- Extraction times drifting wholesale after a sharp
  weather change (big humidity swing)
- Switching bean origin or roast level

The flow (fixed order, no skipping):
1. Log: copy the bag label's roast date, origin, process,
  and roast level into the dial-in log - a batch with no
  label information does not get used; send it back to
  purchasing
2. Purge: empty the grinder hopper and flush out the
  retained grounds; keep the retained grounds in a
  separate jar, never in a drink - old and new beans
  blended in the burrs corrupt every judgment that
  follows
3. Reset: confirm the brew recipe is back at our card's
  baseline - dose 【】g, yield 【】g, target shot time
  【】 seconds
4. Pull and judge: make one drink, read three signals
  only:
  · shot time more than 【3】 seconds fast → grind one
    step finer
  · more than 【3】 seconds slow → one step coarser
  · within ±【3】 seconds → taste it; only keep adjusting
    if it is sharp/sour or burnt/bitter, otherwise lock
    the setting
5. One variable: each round moves the grind only - never
  touch dose and grind in the same round
6. Stop-loss: after 【3】 rounds without landing, stop.
  Hang the "parameters pending review" tag, serve on the
  best setting reached, and escalate to the manager at
  handover - nobody chases dial-in into the rush
7. Record and sync: log the final setting (who, when,
  bean info, how many steps moved); write the new
  setting on the small board or tag by the grinder -
  this shift and the next trust the tag

Constraints:
- No flavor-adjective targets ("should taste of
  caramel") - judgment uses time, yield, and taste only,
  and taste uses exactly three buckets: sharp/sour,
  burnt/bitter, fine
- No numeric grind settings (every grinder's scale
  differs); always "one step," meaning our grinder's
  smallest adjustment unit
- No equipment-care content - that belongs to a separate
  SOP

Why it is written this way

Dial-in is the most feel-dependent job on the bar, which is exactly why juniors won't touch it and seniors don't write it down - the next shift finds shot times drifted and can't tell whether to trust the tag on the grinder or their own tongue. The SOP's value is not teaching direction (finer or coarser is one line); it is making every judgment repeatable and handover-able: who, at what time, because of which number, moved how many steps.

"One variable per round" and "stop after three rounds" guard two real pits: moving dose and grind together means you never know which one worked, and the parameter trail is dead; and chasing numbers into the rush means dumped shots and a queue out the door both keep growing while hands get less steady. The stop-loss hangs a tag instead of grinding on - it turns "not quite there yet" into an explicit state rather than a hazard hiding inside every drink.

Locking taste into three buckets exists because flavor-wheel vocabulary has zero leverage in dial-in: "could be brighter, more citrus" is neither verifiable nor convertible into an adjustment. The question dial-in answers is finer or coarser - three buckets are enough, and the fancier the description, the messier the handover.

Purging and tag-syncing are the two fussiest steps and the two most missed: blended grounds corrupt every downstream judgment, and a setting that lives only in someone's head makes the next shift's first cup a blind pour. Writing the oral tradition down as fixed steps is this SOP's real deliverable.

Variables