Playbooks

    Small, time-capped rituals for turning an intuition into a decision you can defend. Each one is a full afternoon at most. Most are less.

    Pick one. Run it this week. If it does not produce a written decision, it did not run.

    The Throwaway Probe

    Time cap

    One afternoon. Hard stop.

    Use when

    You have an intuition about a behavior, a market, or a workflow — and you are about to spend a week (or a quarter) building the 'real' thing to test it.

    1. 01

      Name the smallest observable behavior

      Not 'do users want this feature.' Something you could watch happen or not happen — a click, a signup, a reply, a share, a request for pricing.

    2. 02

      Build the ugliest surface that could produce that behavior

      A landing page. A Loom. A hand-typed email. A Notion doc with a form. If it takes more than an afternoon, you are already building the thing, not probing it.

    3. 03

      Write the kill number before you launch

      'If fewer than X of Y people do the thing in Z days, I will not build this.' Written down, dated, shared with at least one person.

    4. 04

      Put the probe in front of the right people, not any people

      Distribution matters more than polish. Ten right people beat a thousand random ones. If you have to synthesize traffic, the probe is not testing what you think.

    5. 05

      Read the trace, not the total

      Watch what happened between arrival and action. The number of conversions is a headline; the sequence of events is the actual data.

    6. 06

      Kill, iterate, or promote — same day

      The probe expires at end of day. Sitting on it becomes sunk cost. Decide, write the decision and the reason, and move.

    Output

    A one-page note: what you tested, what you saw, what you decided, and why. Filed by date, findable in six months.

    Fails when

    The probe grows. If you catch yourself designing the 'real' version while running the probe, stop — you have already made the decision the probe was supposed to inform.

    The Pre-mortem Sprint

    Time cap

    45 minutes. No slides.

    Use when

    You are about to commit to a bet that would be expensive or embarrassing to reverse — a launch, a hire, a partnership, a rebuild.

    1. 01

      Assume it failed

      The decision is made, the work is done, six months have passed, and it did not work. That is the premise. Not 'might fail.' Did fail.

    2. 02

      Everyone writes the obituary silently for 10 minutes

      In writing, alone, no discussion. What broke. Who saw it coming. What the room refused to say out loud in this meeting. Silent writing beats group discussion on candor by a wide margin — Gary Klein's original work on pre-mortems is the source here.

    3. 03

      Read them aloud, no commentary

      Round-robin. No responses, no defenses, no 'well actually.' Just the failure modes on the table.

    4. 04

      Cluster the failure modes

      Three to five clusters, no more. Most pre-mortems produce a long tail of one-offs and two or three fears that appear in almost every write-up. Those two or three are the real risks.

    5. 05

      For each cluster, name the earliest observable warning

      What would you see, and when, that would tell you this failure is starting to happen? If you cannot name it, you cannot monitor it, and the pre-mortem was theater.

    6. 06

      Assign one owner per warning

      Not a committee. One person, one warning, one review date. Put the review date on a calendar before leaving the room.

    Output

    A short list of specific, dated, owned warnings — and, sometimes, the realization that the bet is not worth taking.

    Fails when

    Nobody actually believes it could fail. If the room is performing pessimism, the exercise is decorative. Invite an outsider who has permission to be blunt.

    Signal-to-Noise Audit

    Time cap

    One quiet morning per quarter.

    Use when

    Your dashboards are green, your team feels vaguely unsettled, and nobody can articulate why. Or the opposite: dashboards look bad and everyone is calm.

    1. 01

      List every metric you look at weekly

      All of them. Product analytics, revenue, ops, support, engineering, hiring. If it appears in a standup or a review, it counts.

    2. 02

      For each metric, answer three questions

      What decision does this metric change? When was the last time it actually changed a decision? What would we do differently if it moved 20% tomorrow?

    3. 03

      Delete or demote every metric with three weak answers

      Metrics that inform no decisions are consuming attention. Attention is the scarce resource — not data.

    4. 04

      Now list every real observation from the last month that no dashboard captured

      The customer call that felt off. The support ticket pattern nobody has named. The engineer's quiet comment in a 1:1. The competitor move nobody flagged.

    5. 05

      For each, ask: is this a one-off, or the first instance of a pattern?

      You will not know for sure. Flag the candidates. Give each one a review date. Some will decay to noise; the ones that recur are the leading indicators nothing in your stack is measuring yet.

    6. 06

      Publish the delta

      Share what you stopped tracking and what you started watching. The team's model of the business is now more honest than it was this morning.

    Output

    A shorter dashboard, a longer watch-list, and a quarterly ritual that keeps the two in tension.

    Fails when

    You add without subtracting. If the audit only produces new things to watch, you have not audited — you have expanded surveillance.

    The Reversibility Test

    Time cap

    15 minutes before any meaningful commit.

    Use when

    You are deciding whether to make a decision now or wait for more information.

    1. 01

      Ask: is this a one-way door or a two-way door?

      Bezos's frame, still the sharpest: one-way doors deserve deliberation; two-way doors deserve speed. Most decisions are two-way doors dressed up as one-way doors.

    2. 02

      Estimate the cost of undoing it

      In dollars, in time, in reputation, in team morale. If the answer is 'small' on all four, the decision is more reversible than it feels.

    3. 03

      Estimate the cost of waiting

      Waiting is not free. Waiting is a decision to keep the current state. Name what the current state costs per week.

    4. 04

      If reversible, decide now and set a review date

      The review date is the important part. A reversible decision without a review date quietly becomes irreversible through accumulation.

    5. 05

      If irreversible, run a pre-mortem first

      See above. Do not enter a one-way door without the failure modes on paper.

    Output

    A calibrated sense of which decisions deserve friction and which deserve to be made in the next ten minutes.

    Fails when

    Everything gets classified as one-way. That is not caution — that is paralysis wearing caution as costume.