Broken Service Investigation
An unfamiliar multi-module service, no onboarding document, and a symptom report rather than a diagnosis — with at least two plausible root causes waiting in the evidence.
Due: 16 April 2027 · Weight: 25%
The service is failing. Here is the symptom a user reported. Find out why.
You receive a multi-module service with no onboarding document and a failure description written the way a user would write it, not the way a diagnosis would. The system is scoped so that more than one plausible cause turns up once you start looking — the assessment exists to test whether you can tell them apart, not whether you land on one by luck.
Writing substantial new functionality is out of scope. Any code you write exists to observe, instrument, test, or probe the system that’s already there.
What you submit#
A reconstructed model of what the system is meant to do; a record of at least two evidence classes you gathered and why; an explicit list of competing hypotheses, including one you eliminated and how; a diagnosis; and a plain account of what you still don’t know. A proposed intervention and verification check are expected, but a working patch is not the target of this assessment.
How it’s marked#
Handling of alternative explanations carries the single largest weight. Naming the right cause without showing how a real alternative was ruled out does not reach the top band; eliminating a genuine alternative and stating what remains uncertain can score well even with a lightly-sketched fix.
The spec
The brief poses the problem and leaves room for a response. The spec is the fixed part: what a marker checks. Some lines can be checked mechanically; the rest call for judgement.
- a reconstructed model of the system's intended behaviour
- at least two distinct evidence classes gathered, each with a stated reason it was sought
- competing hypotheses named explicitly, including at least one eliminated with its discriminating test
- a final diagnosis, plus a proposed (not required) intervention and verification check
- a plain statement of remaining uncertainty or unverified assumptions
How it is marked
| Criterion | Weight |
|---|---|
| Observation and problem framing | 10% |
| Evidence strategy — evidence classes chosen and why | 20% |
| Evidence interpretation | 20% |
| Handling of alternative explanations | 25% |
| Diagnosis quality and justification | 15% |
| Investigation record / communication | 10% |