Full-service restaurant dinner-service observation checklist - sit through one entire dinner rush recording four countable lines (ticket times, table turns, waitlist, backfill), no scoring, findings converted same night
You are a field supervisor coach for a full-service restaurant chain. Produce a Dinner-Service Floor Observation Checklist for a supervisor to use once a month per store, sitting through one complete dinner service. Every recorded item must be countable (occurrences, minutes, table counts) - no scoring items anywhere.
Time windows (fixed):
- 30 minutes before opening the floor: station prep versus the 86 list, table setting completion, staff on shift
- Peak window (start/end times defined per store): four lines - ticket times, table turns, waitlist, backfill
- Wind-down: closing-side execution
Each observation item has three columns: what to watch / how to record it (the counting rule) / the judgment rule (phrased as "versus the same store, same window, last week" or "over X minutes" - X left blank for the supervisor to set per store).
The four lines (add items, never remove these):
1. Ticket line: tables whose first course ran over X minutes from fire; follow-up ("where is our food") counts and each gap from request to delivery; tables where the gap between first and second dish exceeded Y minutes
2. Turn line: minutes from guests leaving to the table reset (time 3 sample tables); count of dirty tables left over Z minutes; whether bussing and resetting are done by the same person
3. Waitlist line: parties waiting over W minutes; groups that left a phone number, walked away, and came back; whether anyone offered water or menus while they waited
4. Backfill line: count of runner/bussing staffing gaps during peak; count of cross-section backfills; count of bar or cashier stepping onto the floor
Two closing steps (same night):
- Convert the record into findings (standard visit-report format), facts and numbers only
- Tag every threshold breach "process suspect" or "staffing suspect" - the tag must cite a count from the shift, never an impression
Constraints:
- No scores, no stars, no "overall performance" wrap-up paragraph
- No employee names - record positions and actions, not people
- Do not set thresholds for me: every concrete number stays as [X] for the supervisor to fill per store
- Observation never interrupts the floor: the checklist records; it does not direct staff mid-service
Why it is written this way
Static checks and a live dinner service are different universes: cleanliness and signage hold steady all day, while rush-hour breakdowns - tickets stacking, bussing falling behind turns, runners disappearing - exist only inside a two-to-three-hour window a standard visit never sees. Sitting the full service is a hard requirement: watch only the peak hour and you miss the wind-down, cutting the problem list in half.
Everything countable, no scoring, because a score is an opinion and a count is a fact. When supervisors rotate, when a manager disputes the conclusion, when the regional manager asks for evidence, only counts hold up. "Seven follow-ups, longest gap 22 minutes" has no debate in it; "service can't keep up" has an hour of debate in it.
Judgment anchored to same-store, same-window comparisons rather than cross-store benchmarks closes the most common multi-unit mistake: different trade areas, table counts, menu complexity - holding store B to store A's ticket speed is meaningless. Comparison against its own history is the only honest baseline at the single-store level; cross-store comparison belongs to the region, not this checklist.
"Observation never interrupts the floor" is written for the supervisor personally: the moment you start directing, you are watching a directed service. Keep the urge to fix things until the debrief - only then is the record an unmodified sample.
Variables
| Variable | Notes |
|---|---|
| Peak window | The store's actual rush hours - set from POS time-of-day sales, not gut feel |
| Thresholds X / Y / Z / W | Minutes for late entrées, inter-dish gaps, dirty-table dwell, waitlist dwell - first month record without thresholds, set them from the distribution the following month |
| Sample tables | How many tables get timed for turns (default 3) - more is unrecordable, fewer is not a sample |