A discipline is not a deck. It is something a team does, on a cadence, whether or not anyone is watching. The Predictive Planning Loop — Scan, Story, Stake, Steer. This is the ninety-day plan to install it the first time, in a single operating unit, with executive backing.
Ninety days is not arbitrary. It is the shortest window in which a team can complete one full revolution of the loop, including the thing most installations get wrong: actually making a stake decision and steering against it. Anything shorter is a workshop. Anything longer is consulting theater.
Pre-conditions
Before Day 1, three things must be true. If any one is missing, do not start — go fix it first.
An executive sponsor with real authority.This is the person who can say "we are committing this much capital to this scenario" at the end of the ninety days. Without that authority in the room, the loop becomes another planning exercise that produces nothing.
A bounded operating question.The loop runs against a specific decision space — a market entry, a capacity expansion, a product bet, a workforce reshape. Not "the future of the industry." Pick one decision that will actually be made inside the ninety-day window.
A small core team. Four to six people, cross- functional, freed from a meaningful share of their normal calendar. This is not a side project. The team should include the operator who will own the eventual decision, not only analysts.
One Signal, End to End
One signal, end to end — Tuesday morning to a steward's read in under four hours.
Days 1–30: Scan
The first thirty days build the signal layer. The goal is a working inventory of the inputs that move the operating question — and a cadence for refreshing them. By Day 30, the team should be able to say, in a half-page, what changed this week.
Week 1 — Scope and source
Define the operating question precisely. Not "what should our pricing strategy be" but "should we hold or break our list price in Region 4 in the next two quarters." Identify the categories of signal that actually inform that question. There will be fewer than you think — usually six to ten.
For each category, name the sources you trust. Mix internal data, paid feeds, public data, and human signal — what your sales team is hearing, what your suppliers are quietly preparing for, what your competitors' job postings reveal. Human signal is the most underweighted input in most firms.
Week 2 — Standing instrumentation
Set up the lightest-possible standing instrumentation. A shared document is fine. A spreadsheet is fine. A dedicated tool is not required at this stage and often slows the team down. The discipline is more important than the tooling.
Establish a weekly Scan rhythm: a thirty-minute standing meeting where the team reviews what changed across the signal layer in the last seven days. Designate a rotating Scan owner each week.
Weeks 3–4 — Pattern read
With three weeks of data, begin pattern reads. Which signals are strengthening, which are decaying, which are uncorrelated to the rest. Where is the noise concentrated. Where is the signal-to-noise unusually high. By Day 30, the team should produce a one-page Scan Brief that summarizes the operating environment in plain English.
Days 31–60: Story
Days 31 to 60 convert the signal layer into scenarios. Not twenty scenarios. Three. The discipline of constraining yourself to three forces honest pressure-testing.
Week 5 — Identify the cleavages
From the Scan Brief, identify the two or three uncertainties that most determine the answer to the operating question. These are the scenario cleavages. Most operating questions resolve along one cleavage of demand and one of supply, or one of cost and one of regulation. Be specific.
Week 6 — Construct three scenarios
Build three scenarios across the cleavages. Not "good, base, bad." That is forecasting in costume. Build scenarios that are qualitatively different worlds — different customer behavior, different competitor response, different regulatory posture. Name them so they can be referenced in conversation. "The Pull-Forward," "The Long Drift," "The Snap."
Week 7 — Pressure-test
For each scenario, write the indicator set: what would we expect to see in the next sixty days if this scenario is the one converging. These indicators become the watch list for Steer. A scenario without indicators is a story without consequences.
Week 8 — Implications
For each scenario, name the implication for the operating question. Under "The Pull-Forward," should we hold or break price. Under "The Snap," should we hold or break price. The implications are usually not symmetric — and the asymmetry is the point.
Days 61–90: Stake and Steer
The last thirty days are the part most installations skip. They get to scenarios, write a deck, present it, and stop. That is not the discipline. The discipline runs through to a commitment and a cadence for adjusting that commitment.
Weeks 9–10 — Stake decision
Convene the executive sponsor and the core team. Make the call. Resource it. Use the Stake Sizing Matrix to set the magnitude. Hedge deliberately where the cost of being wrong is structural. Document the decision in one page: what we are committing, on which scenario, with which hedges, with which reversibility.
The discipline at this stage is to make a real commitment, not to produce another option set. Optionality is a tool, not a posture. A team that cannot move from scenarios to a commitment is not yet running the loop.
Weeks 11–13 — Steering cadence
Establish the standing Steer rhythm. The Scan loop continues, but now the indicators are tracked against the active scenarios. When an indicator triggers, the team reviews the implication for the commitment. Adjust, don't restart. The point of Steer is incremental correction, not periodic relitigation.
By Day 90, the team has completed one full revolution: Scan to Story to Stake to Steer. The next revolution is shorter. The third is shorter still. Within a year, the loop runs at the cadence of the operating environment, not the cadence of the planning calendar.
Where installations fail
Three failure modes account for almost every botched installation.
The team produces a deck instead of a decision.The symptom is a beautiful Scan Brief, three crisp scenarios, an elegant implication map — and no commitment. Fix it by holding the executive sponsor's calendar for the Stake decision in advance. The decision date is non-negotiable.
The signal layer is over-engineered. The symptom is a custom dashboard nobody opens after Week 4. Fix it by starting on a shared document and only graduating to tooling when the discipline is already running.
The scenarios collapse to a base case.The symptom is "good, base, bad" with the base case obviously favored. Fix it by forcing scenarios to live on different cleavages, named distinctly, and pressure-tested by an outside reader who is empowered to call them indistinguishable.
What you will have at Day 90
A working signal layer. A small set of named scenarios. A stake decision in motion. A steering cadence the team owns. And — critically — a team that has done the loop once and now knows what the loop feels like. The first revolution is the most expensive. The second is half the cost. The fifth is the new operating system.
Ninety days is not enough to turn an organization predictive. It is enough to prove that a single team can run the loop, deliver a decision, and steer it into the world. That is the unit of installation. From there, the discipline scales the way any operational discipline scales: by adjacent teams seeing it work and wanting in.