Portfolio PDLC — Strengthen

Outcome

One card upgraded from status-report material to steering instrument: outcome-framed, carrying leading indicators a sponsor can actually steer by, riskiest assumptions tagged and matched with a proportionate derisking plan — plus, when the work taught something durable, exactly one learning captured for the operating system.

Pick-one rule

One card per invocation. Committed money outranks ideas: given a choice, strengthen the execute/rollout card before the explore card (the umbrella leverage table encodes this).

Workflow

1. Read the card cold. Title, hypothesis, indicators, assumptions, evidence log — plus its source material if reachable. Where on the input→impact ladder does it actually steer? (input → activity → output → outcome → impact.)
2. Rewrite for outcome orientation (an OKR-rewriting skill composes here if you have one; the rules below suffice on their own):
   • Title says what becomes true for whom — not the solution being built. Lightweight first move when a full rewrite is premature: fix the name.
   • outcome_hypothesis in testable form: if [user/stakeholder] attains [benefit] via [change], we expect [business impact].
   • Card-format alternatives are legitimate (OKR card, A3, Lean Canvas, DIBB bet) — match the org's language; the invariant is outcome + evidence, not the template.
3. Leading indicators — Goldilocks or nothing. Each indicator must give feedback in reasonable time AND plausibly correlate to the outcome. Replace activity counters ("12 features shipped") with sense-and-respond signals. One Metric That Matters for the current riskiest assumption beats five vanity numbers. Include current values or n/a honestly.
4. Sharpen the riskiest assumptions. Tag every claim evidence or opinion across Desirability / Viability / Feasibility. Confident-but-unevidenced items move to the top of the discovery agenda. Name watermelon risk plainly where clarity is asserted deep in the lifecycle without evidence.
5. Fix the derisking fit. Match approach to risk: low → just-do-it; medium → ship & measure; high → discovery/testing or seek-alpha (deliberately attack the scariest unknown). Both mismatch directions are smells — heavyweight discovery on the certain, "just build it" on the speculative. Update risk_level / derisking_approach and the Derisking plan section; note in the card if the mismatch implies a stage conversation (that's an advance move, not yours).
6. Record. Evidence log entry (what changed and why), orientation updated only if the card now genuinely steers on outcomes, board regenerated.
7. Capture — at most one learning. If this rewrite exposed something durable (a recurring anti-pattern, a template gap, a language fix that landed), write ONE learnings/<slug>.md (contract format). If it implies changing the operating model, also capture an improvements/ card — the learning holds the evidence, the card holds the bet. Zero learnings is a fine outcome; forced learnings are noise.

Rules

• Improve the card, not the initiative's plan of record — scope/commitment changes are decisions for humans, prepared via advance.
• Preserve the author's voice and the org's vocabulary; this is sharpening, not ghostwriting a consulting artifact.
• Never upgrade orientation as a courtesy. The field follows the words, not the intent.
• Indicators without current values are flagged n/a, not invented.
• One card, at most one learning, board regenerated — the strengthen loop stays small on purpose.

Quality Gates

• The card's title + hypothesis pass the ladder test at outcome level (or the brief says why not yet, and what's missing).
• Every riskiest assumption carries an evidence/opinion tag.
• Derisking approach matches risk level, or the mismatch is escalated as a candidate move.
• ≤1 learning captured, in contract format, only if it changes future behavior.

References

• Ladder + Goldilocks/OMTM: skills/sniff-test/references/derisking-and-indicators.md (this repo).