Practice case — a constrained replay platform decision

All people, numbers and events below are fictional teaching inputs, not production evidence.

Decision packet

A team has 120 engineering hours for the next two weeks after leave and routine meetings. Support/on-call needs 32, a required incident repair needs 24 and uncertainty reserve is 16. There is no authorized overtime or borrowing from other teams.

Option A: a 36-hour synthetic-data replay automation pilot. If useful and adopted, it may save 1–6 operator hours weekly and require two hours weekly maintenance. The immediate proposal is a pilot, not permission to run production replays.

Option B: a 60-hour service extraction with benefits unmeasured. The receiving team has not accepted operational ownership. A smaller discovery spike is possible but has no supplied estimate; scope and estimate it explicitly.

Option C: retain manual work this cycle, collect task-level effort and improve the runbook. Estimate this work rather than treating it as free.

Support staff fear a new escalation burden. Security requires synthetic data for the pilot. Product wants a delivery date; neither adoption nor benefit has been established. The existing ethical-impact exercise still applies.

First submission

Write a two-page memo choosing an option or a justified smaller experiment. Show arithmetic, assumptions, displaced work and a falsifiable benefit hypothesis. Include an outcome metric, collection plan, at least two safeguards, owner and stop/review conditions. State what you can promise now and what depends on evidence. Present the strongest objection fairly.

Create a delegation brief for the pilot or discovery, with decision rights and escalation. Reuse this in the original guided labs, including two actual pre-reads and a consenting peer delegation. The fictional case alone does not satisfy interpersonal evidence.

Change card — open after your initial decision

Support demand rises from 32 to 52 hours. A pilot sample reports 40 tasks/120 minutes before and 20 tasks/40 minutes after; difficult tasks were diverted to an unmeasured queue. One proposed convenience feature would widen data access.

Revise the capacity plan and benefit claim. State who must resolve the access decision, what work pauses, and which evidence is missing. Preserve your original memo and show changes. Do not infer saved money, safe production operation or causal improvement from this sample.