# Initiative frame

One page. Written by product, design, and engineering together, before anyone estimates
anything. Version it as facts change — a changed frame is learning, not failure.

Source: lesson 0039 · Turn a Product Brief into an Initiative Frame.

---

## Blank

**Initiative:** `[name]` · **Version:** `[0.1]` · **Date:** `[YYYY-MM-DD]` · **Lead:** `[name/role]`

| Field | What goes here |
| --- | --- |
| Why now | The trigger, the evidence behind it, and the cost of doing nothing this quarter. |
| Problem | For `[cohort]` in `[situation]`, `[observed condition]` causes `[impact]`. Baseline `[number]`, source `[system/report]`, measured `[date range]`. |
| Outcome | Move `[measure]` from `[baseline]` to `[target]` for `[cohort]` by `[date]`. A behavior change, not a shipped feature. |
| Guardrails | Measures that must not degrade. Each needs a threshold and an owner. |
| First boundary | The first cohort and situation you will serve end to end. Narrow enough to finish. |
| Non-goals | Each one prevents a specific plausible misunderstanding, not a generic "out of scope". |
| Constraints | Facts the design cannot negotiate. If you could change it by asking, it is a preference, not a constraint. |
| Assumptions | Top value, usability, feasibility, and viability beliefs, each with current evidence strength. |
| Risks | Impact × likelihood, the earliest signal, mitigation, owner, review date. |
| Decision rights | Who has the final call on outcome/scope, experience, architecture, rollout — and who must be asked first. |
| Next evidence | What the team will learn before the next expensive commitment, and by when. |

**Quality check.** A reader outside the initiative can answer: why now, for whom, what
measurable change, what is deliberately excluded, what cannot move, what might be false,
who decides. If they can only describe the proposed features, this is still a brief.

---

## Completed — returns resolution pilot

**Initiative:** Self-serve returns resolution · **Version:** 0.3 · **Date:** 2 October ·
**Owner:** Priya (product), written with Marco (design) and Dana (tech lead)

| Field | Entry |
| --- | --- |
| Why now | Returns are the second-largest source of support contact volume, and returns volume rises about 40% from mid-November. No support headcount is approved for the season. Doing nothing means the contact queue grows with the returns queue and agents spend the season chasing status instead of handling exceptions. |
| Problem | For signed-in domestic customers returning an unopened item within 30 days, a request waits a median of 72 hours for an agent decision, and 41% of those requests produce at least one support contact. Baseline: returns queue export joined to the contact log, 1 June – 31 July, n = 4,812; query owned by Priya and re-runnable. |
| Outcome | Move median time from *request submitted* to *decision communicated* for the eligible domestic cohort from 72 hours to under 4 hours, and the support-contact rate for that cohort from 41% to 31% or lower (a 25% relative reduction), with the pilot open to real customers by 30 September. |
| Guardrails | Fraud loss per 1,000 eligible returns at or below the 2.4 baseline (Risk). Manual review queue at or below 50 requests/day (Nils, support operations). WCAG 2.2 AA on the four critical flows (Marco). Carrier cost per resolved return at or below $3.42 (Priya). Personal data retained no longer than 30 days (Ola, privacy). |
| First boundary | Signed-in domestic customer, unopened item, delivered 30 days ago or less, refund to the original payment method. |
| Non-goals | Exchanges. International shipping. Guest checkout. Redesigning warehouse intake. Replacing the agent tool. Each is listed because someone plausibly assumed it was in scope. |
| Constraints | The order platform stays authoritative for order truth. The carrier contract is fixed for the pilot term. No payment card data enters the returns module. Personal data is retained for 30 days. The team is seven people and does not grow. The 30 September date is fixed by the season, not by preference. |
| Assumptions | Eight named beliefs, in `assumption-map.md`. The three that got a budget: policy is codifiable (A1), customers trust an automatic decision (A2), the carrier is fast enough to sit inside the request (A3). |
| Risks | R1 automatic decisions raise fraud loss (high × medium; signal: fraud rate in the exposed cohort versus a matched control; mitigation: fraud hold plus staged exposure; owner Risk; weekly). R2 the referred cohort reads a referral as a refusal and contacts support anyway, harming the very measure (high × medium; signal: contact rate inside the referred segment; owner Marco). R3 the carrier cannot meet the label latency budget (medium × high before the spike; closed by ADR 0003). R4 guest orders are 29% of demand and stay outside the pilot, so the headline outcome overstates the business effect (medium × high; owner Priya; decision date 15 November). |
| Decision rights | Outcome, scope, cohort: Priya. Experience: Marco. Architecture and technical risk: Dana. Rollout advance and reversal: Dana with Nils — and stop authority for any on-call engineer, acting first and explaining afterwards. |
| Next evidence | Comprehension retest at n = 12 (Marco, 15 September). Fraud rate in the 1% cohort against a matched control after 120 exposed requests (Risk, from 29 September). Guest-order opportunity sizing (Priya, by 15 November). |

### Version history — the part most frames delete

| Version | What changed | The evidence that forced it |
| --- | --- | --- |
| v0.1 · 25 Aug | Written from the brief in one 90-minute session. Outcome stated as "resolve eligible returns in under 4 hours and generate fewer support contacts". Success described as "automatic approval for at least 90% of eligible requests". | Nothing yet. It is kept rather than deleted so the team can see how far the framing moved on evidence rather than on opinion. |
| v0.2 · 4 Sep | The contact target gained a number and a denominator (41% → 31% for the eligible domestic cohort). The outcome event changed from "automatic approval" to "decision communicated", so a referral answered within four hours now counts as success. Added the non-goal "replacing the agent tool" and the carrier-cost guardrail. | The 200-case policy sample (A1) showed 15% of eligible requests are genuine judgment calls. An outcome defined on *automation* would have scored the correct product behavior — routing a judgment call to a human quickly — as failure. |
| v0.3 · 2 Oct | Guest orders moved from "non-goal" to "deferred opportunity, sized, decision date 15 November", with the 29%-of-demand figure written in. Mobile comprehension work moved ahead of the evidence-photo rule in the slice order. | Guest orders were re-argued in three consecutive weekly reviews because "non-goal" read as "refused forever"; naming the deferral with a date ended the debate for the cost of one line. The mobile reorder followed the first exposed week's 11-point completion gap. |

**Traces to:** nothing — this is the root. Every other artifact traces here, and the real test
runs in reverse: name any artifact in the pack and you should reach this page in one hop.