Week 10 Session: No One Left to Ask

How do you investigate a system when the people who built it, and much of its context, are simply gone? Scaffolding drops substantially from this session onward.

3 May 2027

Before the session#

The system this session is large, loosely documented, and assembled from multiple components. What you receive is operational only — partial runbooks, incident tickets, stale comments — with no author to ask.

The incident#

Northfield’s storefront platform now spans a catalog service, a cart service, and the checkout flow that hands off to fulfil, each maintained at different times by different, now-departed people. Customers occasionally report that items they added to their cart are simply gone by the time they reach checkout — never all of them, never predictably, and only reported during high-traffic periods. Nothing in any single component’s partial runbook mentions this. No ticket in the incident history names a subsystem as responsible. You are not told whether the fault sits in the cart service, in how checkout reads from it, in something shared between them, or somewhere else.

Choosing where to start#

There is no assigned first file, first log, or first component this week. The instinct to open the largest or most obviously broken-looking service first is not automatically the productive one — a system this size makes “start reading” an expensive, low-yield move unless it’s aimed at something. The more useful question is which single observation, gathered first, would most reduce your uncertainty about where the failure actually crosses a boundary — cart to checkout, checkout to fulfil, or somewhere inside one service alone — before you commit to reading any of them closely.

In the session#

No single evidence class is curated to be sufficient on its own, and the system is larger than one person can fully read. You choose which evidence classes to reach for yourself, and produce a model, a diagnosis, and an explicit account of what remains unknown given what was actually available.

Afterwards#

Nothing here is a new investigative skill — it’s Weeks 3–9’s evidence toolkit, used without being told which piece to reach for.