Queue Autopsy

A close analysis of one real waiting system, separating what you observed from what you assumed, and accounting honestly for its slowest case.

Due: 12 March 2027 · Weight: 20%

Purpose#

Weeks 1 through 3 gave you a method for looking at a queue honestly: watch it before theorising about it, compute more than one summary statistic, and check how sensitive a summary is to the longest recorded wait. The Queue Autopsy asks you to apply that method properly, in writing, to one real waiting system of your choosing — not the Office of Everything.

The task#

Produce a written analysis of the system you have been observing since week 1. The analysis has two distinct jobs, and they must not be merged: an observation record (what you actually saw, when, and how) and an analysis (what you conclude from it, including any assumption you had to make because the system didn’t show you everything). A reader should be able to tell, line by line, which section they are reading.

Within the analysis, include a wait-time calculation for a representative case and your longest recorded wait. Report the full sample mean, then recompute it with one longest wait omitted, as in Part A of week 3’s lab. State both sample sizes. If there is no conspicuous outlier, say so and use the longest wait without relabelling it unusual; if several tie, omit one and identify it. Honest absence of an outlier is not penalised. This check does not establish causes or what removing a customer would change.

An observation record approved under the accessibility policy may replace personal observation and satisfies the observation requirement. Identify its source and what you did not observe yourself. The same marking criteria apply; approval does not require inventing first-hand experience. General literature alone is not an observation record.

What you submit#

A single written document (a few pages is enough) containing the observation record, the analysis, and your wait-time calculations shown as working, not just a final number.

Scope boundaries#

This is an analysis of one system you can actually watch, not a survey of queueing systems in general and not a proposal for fixing anything. Do not collect anyone’s personal information while observing — timing and counting people is enough; see the policies page for what “observing without collecting personal data” means in practice. An outside source may help you interpret something you observed, and if you use one, cite it — but it cannot prove what happened in the system you watched. A claim about your specific system has to come from your own observation or the approved observation record. If that evidence cannot support a claim, say so and scope the claim down rather than borrowing authority from somewhere else.

Relationship to other work#

This is the only assessment built entirely from a system you chose yourself, rather than the Office of Everything’s synthetic case. Its method — separate observation from assumption, account for outliers rather than averaging past them — is the same method the Policy Trial and The Honest Queue both assume you already have.