Week 3 Queue Lab: Recompute with and without the outlier

Rehearse the Queue Autopsy's core technique on the Office of Everything's data, then apply it to your own observed system before writing it up.

8 March 2027

This lab separates changing a summary from changing a queue. Part A prepares the Queue Autopsy; Part B introduces later policy comparisons.

Before the lab#

Bring your week-1 and week-2 notes, or an approved accessible observation record. If there is no conspicuous outlier, bring the longest recorded wait and say that it was not unusually different from the others.

In the lab#

Part A: change the summary#

Reuse the ten waiting durations from week 2. Compute their mean and median, then omit the single 26-minute wait and recompute both. Identify the changed denominator. This checks the sensitivity of a reported statistic; it does not show that removing that person would improve anyone else’s experience. Keep the full sample as the primary result.

Part B: change a service duration#

Office of Everything file — synthetic case material. A separate log, Monday of week 3, not a reconstruction of week 2. A new dispute and nine later tickets arrive at minutes 0, 1, 3, 4, 5, 7, 8, 9, 11, 12. The dispute starts at minute 0 and finishes at 26; each later service takes 2 minutes. Service is FCFS and non-preemptive. A ticket cannot start before arrival; an idle desk waits for the next arrival.

Compute each later ticket’s start and wait, and their mean wait. Then recompute with the first service shortened to 2 minutes and every arrival and other service unchanged. Compare the mean and the total waiting minutes across the nine later tickets. This is a counterfactual within this model, not an observed improvement or deletion of one row from a report.

Apply Part A to your observation#

Report your real sample in full, then recompute its mean with one longest recorded wait omitted. If several tie, omit one and explain which. With no conspicuous outlier, say so: the sensitivity calculation still has value. Do not claim the omitted person caused other waits. Part B is not required for your real system without enough arrival and service data to support it.

Afterwards#

You now have the technique the Queue Autopsy asks for: observation notes (week 1), a wait-time calculation that separates typical from atypical cases (weeks 2–3), and a written account of what is measured versus assumed. The Queue Autopsy is due at the end of this week and asks you to bring these three things together for the one system you have been watching all along.