Build a Weekly Warmth Report That Drives One Decision

Card 1: what materially changed

List only changes that affect a user promise, operating risk, acquisition path, cost, or decision. Examples include a deployed conversion repair, new payment failure, updated pricing, support backlog, material content revision, provider disconnect, ownership change, or completed recovery test. Exclude routine activity unless it created evidence or an exception.

Link each change to the authoritative receipt: commit, deployment, public URL, provider ID, ticket, event, or approved record. State the effective date and environment. This keeps the report from becoming a narrative detached from what actually shipped or failed.

• Material change
• User or operating impact
• Date and environment
• Evidence link

Card 2: the current highest risk

Choose the risk with the greatest combination of consequence, likelihood, time sensitivity, and weak control. A broken checkout or ownerless security inbox usually outranks an overdue social post. State the affected user, observed condition, known evidence, uncertainty, current mitigation, and decision deadline.

Do not average the risk into a general health score. A project can have nine warm channels and one critical payment or access incident. Preserve that exception at the top until it is fixed, accepted, transferred, or consciously stopped.

• Observed condition
• Affected user or asset
• Consequence and urgency
• Current mitigation
• Decision deadline

Card 3: proof quality

Summarize which high-consequence paths have current receipts and which remain unverified. Distinguish success, failure, and unknown. A signed-out provider shell, stale screenshot, or historical event is not current proof. Link to the exact check and blocker so the reviewer can inspect it without opening every dashboard.

Show evidence decay. A safe checkout from today may be current; a recovery exercise from last year probably needs review. The report should not reward the number of checks completed. It should reveal whether the evidence needed for the week's decisions is strong enough.

• Current receipts
• Failed checks
• Unknown checks
• Expired evidence
• Exact blockers

Card 4: trend with definitions and caveats

Choose a small number of outcome measures such as qualified visits, activation, retained use, support age, successful delivery, paid conversion, or recovery time. Define the numerator, denominator, window, source, and observation status. Compare like with like and annotate changes to instrumentation or product flow.

Never coerce missing analytics to zero. Use first-party application or payment truth for activation and revenue where available, with analytics as corroboration. A trend should help distinguish noise, structural movement, and measurement failure. If the data cannot support that distinction, say unobserved and fix the measurement contract.

• Metric definition
• Window and source
• Observation status
• Comparison
• Instrumentation caveat

Card 5: owner and exception queue

List the few exceptions that require a person, each with accountable owner, next action, evidence required, and due date. Separate review-ready work from blocked work. For blocked items, preserve the exact missing authority, credential, account identity, provider control, or external state rather than assigning a generic in progress label.

Cap the queue. If every recurring task appears, the report becomes another task manager. Show incidents, stale high-risk evidence, pending approvals, and structural improvements. Routine green checks can be summarized with a link to receipts.

• Exception
• Accountable owner
• Next action
• Evidence required
• Due date or blocker

Card 6: one decision and follow-up test

End with one decision stated as a verb: repair, pause, continue, stop, reassign, investigate, or accept. Record the evidence considered, important alternatives, owner, deadline, and what result would confirm or challenge the decision. A decision to make no change is valid when it is explicit and supported.

Schedule the follow-up evidence before closing the review. If the team repairs attribution, specify the controlled journey and payload that will prove context survives. If it pauses a channel, specify when and what evidence would justify restarting. This turns the report into a learning loop rather than a weekly archive.

• Decision
• Evidence and alternatives
• Owner and deadline
• Follow-up test
• Revisit condition

A 20-minute review agenda

Spend three minutes on material change, five on the highest risk, four on evidence gaps, three on trend, and five on the decision and owner. Read the report before the meeting when more than one person is involved. Open source receipts only for disagreement, high consequence, or weak evidence.

After the review, publish the decision and update the queue. Do not rewrite the report into polished marketing. Its value is operational clarity. Once a month, inspect which cards repeatedly fail to produce decisions and remove or redesign them.

• 3 minutes change
• 5 minutes risk
• 4 minutes proof
• 3 minutes trend
• 5 minutes decision

Put this into practice

Keep the report to six cards and one decision. Link claims to receipts, preserve unobserved data, define trends, and show only the exceptions that need a person. Archive each report with its decision and follow-up result so the team can see which assumptions improved, which risks repeated, and which checks produced no useful action. The report succeeds when a named owner leaves with a clear next action and a follow-up test—not when every available metric fits on one page.