Café Visit Duration: Keep Unfinished Visits Visible

Define the visit and the observation window

Choose one outlet, service format and observed seating area. Write the date, window start, cutoff and timezone before collecting observations. In this guide a visit starts when the party is seated and finishes when its last guest leaves the seating area. Record arrivals and waiting separately: arrival at the door is a different starting point.

Include every eligible party seated from the window start up to, but not including, the cutoff. Keep anonymous visit IDs and party size. Parties already seated before the window started are outside this new-arrivals group; record their presence and the exclusion reason in the context log. Do not quietly add a pre-existing party with an invented seating time.

State how late arrivals joining a party, split parties, seat changes and departures followed by returns are handled. Keep the same rule throughout the observation. Table references can help locate a record, but a party remains one visit even if it uses several seats. Its elapsed minutes are counted once.

Use observed events rather than convenient timestamps

Record the seating event and its evidence as it happens. Record departure independently. Opening a bill, taking an order, receiving payment and clearing a table are different events; do not substitute one of them for a missing seating or departure time without changing and clearly naming the measure.

Use full dates and times on one fixed UTC-offset basis. The worksheet accepts whole-minute timestamps, so record the source precision and use one agreed method consistently. A same-minute completed visit has zero recorded minutes at that precision; that does not mean no visit occurred. Convert times across a clock change to one fixed offset before entry.

Keep only the information needed for the observation. Anonymous IDs, party sizes, role labels and protected evidence references are sufficient here. This is an operating record, not a reason to gather names, contact details or recordings of guests.

Give every included visit an outcome at cutoff

Use Departed only when the departure was observed by the cutoff and its timestamp is supported. Completed minutes are departure minus seating. Use Still seated only when a cutoff check confirms that the party remains seated; elapsed minutes to cutoff are a lower bound on its eventual completed visit, not its final duration.

Use Outcome unknown when the departure was not observed and continued presence at cutoff cannot be established. Keep the seating time and a missing-evidence reference, but leave completed and ongoing minutes blank. An unknown outcome is neither a zero-minute visit nor a party assumed to remain seated.

Reconcile the included visit IDs against the complete observation register before confirming the group. A record of five convenient visits from a busier service is not the full group merely because all five rows are filled in. If observation ends early, keep the original scope and identify missing outcomes, or create an explicitly revised observation record.

Worked example: five parties, three completed visits

In this fictional observation, the window runs from 12:00 to 14:00 IST on 8 November 2026. Five eligible parties were seated before the cutoff. The team has evidence for the first three departures, confirms V04 still seated at 14:00 and cannot establish V05's outcome.

Read the completed-only average with its missing outcomes

The three completed visits total 125 minutes. Their mean is 125 divided by 3, or 41.67 minutes. Report it beside the counts: five included visits, three departed, one still seated and one outcome unknown. The average for all five completed visits is not available.

Do not insert V04's 60 minutes into the completed average or turn V05 into zero. Nor should the completed-only average become the expected length of a new visit. A short observation can favour visits that finish quickly, while longer visits remain ongoing. The conditions and the unfinished records belong beside any comparison.

This is a party-level elapsed-time measure. Two parties can be seated at the same time, so their minutes are not a partition of the service window. Do not divide summed party minutes by the window and call it table utilisation. Seating capacity, shared tables, clearing, resetting and the next assignment require their own records.

Keep later evidence and follow-up separate

If a departure after 14:00 is observed later, keep it as linked follow-up. It must not replace the original fact that V04 was still seated at the 14:00 cutoff. The worksheet summary stays tied to its original window. A longer follow-up register needs its own endpoint and preserves the original selected group; do not extend the seating window to admit later arrivals or drop unresolved visits.

If later evidence establishes a departure that actually occurred before the original cutoff, preserve the first entry and record the correction, evidence, reviewer and date. Distinguish that correction from extending the observation. Keep each version so the basis of a changed summary is visible.

Review comparable services with their menu, party mix, staffing and unusual events recorded. Use the findings to investigate guest experience and service delays with the team. A duration record by itself does not justify rushing guests, changing staffing or promising that a different table layout will increase sales.

Keep expectations separate from measured visits

Cornell's publication abstract for How Long Should Dinner Take? describes research into customers' expected dining duration in a particular casual-restaurant setting. Expectations are a different question from the observed seating-to-departure times recorded here. This guide applies none of that study's numerical findings, preferred durations or revenue implications. Its observation procedure, example and worksheet are original.