Office Briefings

Office Briefings are paid decision aids for specific throughput and cost-control questions. They exist to reduce deployment risk by making assumptions explicit, tying recommendations to measurable evidence, and defining verification and rollback conditions.

Required structure

1. Question (H1): one sentence; falsifiable.
2. Direct answer: one paragraph; state conditions.
3. System model: topology + constraints + the unit being optimized (cost per successful completion, p95 latency, success rate).
4. Evidence: time buckets + error/retry causes + one baseline-vs-current table with definitions.
5. Mechanism: name the control and the bucket it changes (cap, reuse, backpressure, timeout, retry budget).
6. Risks: what can regress and what to watch during rollout.
7. Verification plan: instrumentation, acceptance criteria, rollback conditions.
8. Scope boundary: explicit inclusions/exclusions; no future promises.

Evidence requirements (non-negotiable)

• Any numeric claim must be derived from a defined metric and a defined time window.
• If the metric cannot be computed from logs, the briefing must say so and specify what must be instrumented first.
• “Improvement” must be expressed as a ratio or delta of measured quantities, not as an ungrounded percentage claim.

Publishable vs rejected

Publishable:

• Clear question, direct answer, explicit assumptions.
• Evidence table with definitions.
• Mechanism tied to measurable buckets.
• Verification plan includes acceptance + rollback.
• Institutional tone; no hype; no vendor promises.

Rejected:

• Any invented benchmark, percentile, or cost number.
• Any recommendation without a stated measurement plan.
• “Trust me” claims or urgency framing.

Internal links

• Office index
• Flagship artifact