Café Queues: Observe Waiting Before the Order

Define the queue and what one entry means

Choose one visible join point and one service-start point. At a café counter, joining might mean taking a place in the ordering line; service starts when the order taker begins attending to that entry. The wait ends there, even though payment, production and collection may take longer. Use the kitchen workflow guide for work after order acceptance.

Choose a consistent observation unit. The example counts one ordering party as one queue entry, whether it contains one person or several. Do not mix a headcount of everyone standing nearby with a count of ordering parties. Record shared lines, priority access and collection-only traffic so their boundaries are clear.

Give every entry an anonymous observation ID. If a party leaves and later rejoins, record a new entry and link the history in a note; two entries are not two unique customers. Avoid names, photographs or other unnecessary personal information.

Fix the join window and the observation cutoff

Record the date, start time, timezone and clock used. Select entries that join from the start of the window up to, but not including, its end. Someone already waiting at the start is a carry-in entry: keep that observation separate because their earlier wait is outside this cohort.

Choose the cutoff at which outcomes will be reported. It may be the end of the join window or later if the observer continues following those entries. Do not add later arrivals to the original cohort just because observation continues. A second window needs its own record.

The worksheet uses elapsed minutes from one recorded start, not clock strings or an automatic midnight calculation. Zero means the actual start of observation. Use the same clock and convert source times consistently; retain the original observation notes. The join window must be positive and the cutoff cannot precede its end.

Record what happened, including what is unknown

Give each selected entry one outcome at the cutoff: Served, Left, Waiting or Unknown. Served means service began; it does not mean the drink was collected. Left means the observer saw the entry leave before service. Waiting means the entry was still observed in the line at the cutoff.

Use Unknown when the observer loses track of an entry or cannot establish its outcome. Do not turn that gap into a departure or a short wait. If service or departure happened, record its elapsed event minute and retain the supporting note. The event cannot precede joining or follow the reporting cutoff.

For Waiting, leave the event time blank: the worksheet shows time waited so far using the cutoff. For Unknown, leave it blank and do not infer a duration. Keep interruptions, queue switching and any assistance in the observation notes. An outcome label without evidence is not enough to explain a change.

Worked example: four entries at a ten-minute cutoff

This fictional observation selects parties joining in minutes 0 through just before 10 and reports their status at minute 10. The four IDs below are queue entries, not a count of every person present.

Read completed waits without hiding unfinished ones

Two of the four entries reached service, one left and one was still waiting. The mean wait to service is (3 + 5) ÷ 2 = 4 minutes, describing only those two served entries. The four-minute wait before departure answers a different question. The six-minute unfinished wait is not a completed service wait.

Do not average all four durations and call the result customer waiting time. That would combine different endpoints and treat an unfinished observation as finished. Nor does the observed departure establish a lost order, a reason for leaving or an eventual abandonment rate for the still-waiting party.

The sheet reports counts and the mean among served entries only after the scope and register are confirmed complete and all entered rows are valid. Unknown outcomes remain visible as a count. A confirmed empty register has zero counts and no mean; a cohort with no served entries also has no mean. Missing observations are never replaced by zero-minute waits.

Test one practical change and keep the conditions

Read the notes beside the times. A line that stalls while guests find the menu calls for a different trial from one that stalls during payment. A clearly marked ordering point or a menu visible before reaching the counter may be worth testing when the observations support it. Assign an owner and make the change specific.

Record staffing, menu version, service setup, disruptions and demand conditions for both observations. Use the same join and service definitions, but keep each window in a separate copy of the sheet. A later mean based on fewer served entries should not be presented as improvement without looking at departures, unknowns and unfinished waits.

Keep the change, evidence and decision together. These observations can guide another trial; a small before-and-after comparison does not establish causation or forecast future demand. If a queue creates an immediate access or safety concern, the responsible team should address that directly rather than waiting for the observation exercise to finish.

Where the distinction comes from

MIT OpenCourseWare’s queueing lecture distinguishes the waiting queue from the service facility and discusses their different measures. This guide uses that general distinction to frame a manual observation. Its cohort rules, four entries and workbook are original teaching choices; no queueing forecast, steady-state assumption or lecture numerical example is applied.