A Full Calendar. A Plan That Governed Nothing.
- Meridian Grocery Co. runs 68 stores across a regional multichannel grocery chain — in-store plus delivery, heavily weighted toward perishables.
- On-Shelf Availability had sat at 91% for over a year. The failure mode that mattered wasn't the warehouse — it was the shelf, and nobody had a system that looked there.
- Forecast accuracy (WAPE) ran at 34% at SKU-store-week, with a bias direction nobody had actually diagnosed — some categories over-forecast into waste, dairy under-forecast into real stockouts.
- A full monthly S&OP calendar already existed — Category, Demand, Supply, Financial Review, Reconciliation, Executive S&OP — every meeting held, every month, for two years. None of it moved the number.
- The COO's question wasn't "do we need more meetings." It was "prove one integrated planning cycle can be run, repeated, and owned — in 90 days — without dissolving into another set of meetings that decide nothing."
That's the exact condition FlightDeck90 is built for: a business that has institutionalized the cadence but not the decisions, with more suspected root causes than anyone has traced back to a single locked panel.
What "success" had to mean here
The KPI Panel Contract locked one distinction before Day 1: a fuller calendar isn't the deliverable. Three KPIs, each with a floor and a stretch, is.
- Primary (financial) — Working Capital: $1,000,000 inventory investment baseline · floor 8% freed · stretch 15% freed.
- Service KPI (OSA): 91% baseline · floor 95% · stretch 97%.
- Accuracy KPI (WAPE): 34% baseline at SKU-store-week · floor 25% · stretch 20%.
Everything below was built against that panel — and the Day 90 verdict reports all three honestly, including where the trajectory fell short of stretch.
Three KPIs, Locked Before Day 1 — One Designated Primary
Diagnostic is the only phase off the 90-day clock. The three KPI slots were filled from the Retail & Grocery row of the technical library, not a generic KPI set — the failure mode that mattered for Meridian was the shelf, not the warehouse, so On-Shelf Availability stood in for a generic fill-rate metric.
| KPI | Definition | Baseline | Floor | Stretch |
|---|---|---|---|---|
| Primary — Working Capital | Shrink-adjusted inventory investment | $1,000,000 | 8% freed | 15% freed |
| Service KPI | On-Shelf Availability (OSA) | 91% | 95% | 97% |
| Accuracy KPI | WAPE at SKU-store-week | 34% | 25% | 20% |
Owner of the number: VP Merchandising & Supply Chain — one named executive, not a committee. Start date: Day 10, upon KPI Panel Contract sign-off. And one rule that made the panel more than three separate scoreboards: Working Capital was designated Primary — when Reconciliation later forced a trade-off between freeing cash and protecting the shelf, the Primary KPI's floor could not be silently breached to buy a better OSA number.
Dimension Discovery and the atomic decisions are how the panel moves. The KPI Panel Contract itself doesn't move — it's the fixed line everything downstream gets judged against.
An Honest 2.1, Not a Hopeful 3
The Planning Maturity Score was taken across People, Process, Technology, and Finance — and it came back at 2.1, "Anticipated": cadence exists and is followed, exceptions are triaged, forecast has a named owner. That score, not appetite, set the starting AI tool tier: templated dashboards, rules-based alerts, and a basic statistical forecast — agentic exception triage stayed switched off until the cadence itself could carry it.
Pilot scope: Dairy — the cold-chain category chosen specifically because it stress-tests the weekly S&OE Loop hardest, not the easiest category to show a quick win.
The misfit team and its financial-accountability enforcement
- Category Manager (Dairy) — holds the "No BS" veto to stop a shipment or a plan step on imminent waste or uncorrected phantom inventory.
- Demand Planner — owns the Demand dimension and the Tuesday demand-sensing slot.
- Supply Planner — owns Supply and the cold-chain capacity read.
- Finance Lead — owns Financial Integration and facilitates Reconciliation.
The financial-accountability half of the "No BS" model was confirmed the same day: an unplanned or poorly forecasted promotion's emergency-freight cost charges against the Category Manager's own budget, not absorbed silently by logistics.
The Filter Gets Tested Before It Ever Makes a Live Call
Week 2: the S&OE Loop's exception filter is run against last month's POS and inventory feed — a rehearsal, not a live decision. It's the same discipline the wider industry is racing to retrofit after agentic pilots got bolted onto cadences that couldn't yet hold them; FlightDeck90 builds the shadow run in from Day 11.
| Day | Session | Result |
|---|---|---|
| Week 2 | Shadow run — last month's data | Filter tuned, zero live decisions made |
| Week 3, Monday | First live S&OE Loop session | 11 SKU-store combos flagged past the >15% deviation threshold |
| Week 3, Tuesday | Demand-sensing slot | Push quotas recalculated for the Dairy pilot |
The first monthly S&OP Cycle ran with Creativa facilitating all six steps end to end — Category & Portfolio Review through Executive S&OP — heavily scaffolded, so Meridian's team saw the whole cycle work before being asked to run any piece of it themselves. Shadow pricing translated the Dairy volume scenarios into margin impact automatically, ahead of Finance Review.
Day 40: The First Batch, AI-Drafted, Human-Approved
The first batch of atomic decisions is generated across the confirmed dimensions — each one AI-drafted from the scored evidence, and signed only by its named human owner.
| Dimension | Atomic decision | Owner | Status |
|---|---|---|---|
| Demand | Dairy SKU-family forecast adjustment approved, entered into plan of record | Demand Planner | Made |
| Supply | Cold-chain carrier capacity confirmed for the period | Supply Planner | Made |
| Inventory | Dynamic safety-stock change applied to the Dairy SKU set, by day of week | Category Manager | In progress |
| Financial Integration | Revenue assumption for Dairy reconciled and signed off by Finance | Finance Lead | Made |
Meridian Leads. Creativa Only Coaches Reconciliation.
Day 41: the Category Manager runs Demand Review directly; the Supply Planner runs the constraint-solver scenario read unassisted. Creativa steps back to coaching, facilitating Reconciliation only. The GUSTO Gate goes live for the first time to decide which conflicts actually earn executive airtime.
| Conflict | GUSTO score | Verdict |
|---|---|---|
| Raise Dairy safety stock vs. cold-storage capacity ceiling | 74/100 | Cleared — reaches the executive |
| Shift a promo calendar slot to protect the cold-chain route | 68/100 | Cleared — reaches the executive |
| Bring forward a national ad push by two weeks | 41/100 | Rejected — documented and cut, not softened |
A scenario-generation agent proposed 2–3 resolved options per conflict, GUSTO-scored, for the human arbitration that actually happened in Reconciliation — the agent drafted the shortlist, the executive sponsor decided.
Frozen Joins the Pilot — Capacity Allowed It
Capacity proved sufficient to expand the pilot scope to a second category, and the Portfolio dimension activated after Cycle 1 showed it carried more leverage than originally assumed.
| KPI | Day 60 reading | Trajectory |
|---|---|---|
| OSA | 95% | On pace — floor threshold reached |
| WAPE | 24% | Ahead of pace — past the 25% floor |
| Working Capital | ~9% freed | On pace — floor within reach |
No assumption needed correcting this cycle — the recalibration log stayed empty, logged as a deliberate non-event, not an oversight. A full cycle of runway remained before the client-led close.
Creativa Observes. Meridian Runs the Whole Cycle Alone.
Day 71: all six S&OP steps and the weekly S&OE Loop run without facilitation. Creativa attends Reconciliation only, to observe — no coaching, no drafting the brief.
The Absorption Test asked one binary question: is Cycle 4 already scheduled and staffed to run without Creativa's participation?
| Check | Result |
|---|---|
| Cycle 4 calendar invite | Sent and accepted — zero Creativa attendees |
| Named agent owners | Every AI step keeps a human sign-off and a visible audit trail past Day 90 |
Floors Cleared. Stretches Not Yet — Reported Honestly.
The final KPI Panel reading and the Day-90 Planning Maturity re-score land together. All three KPIs cleared their floor. None reached stretch — and the verdict says exactly that, the "factual not yet" FlightDeck90 is built to tolerate rather than paper over.
On-Shelf Availability, Day 0 → Day 90
Monthly KPI Panel reading. Floor cleared by Day 60; the last four points are the honest distance still open to stretch.
The Traceability Map — KPI back to atomic decision
| Link | Movement |
|---|---|
| On-Shelf Availability | 91% → 96% |
| ← Demand, Inventory dimensions | forecast adjustment + dynamic safety-stock change |
| WAPE at SKU-store-week | 34% → 21% |
| Working Capital freed | ~11% — floor cleared, stretch (15%) not yet |
| Planning Maturity Score | 2.1 → 2.9 — Anticipated toward solid Consistent |
A Cadence That Held Without Us — Even Before It Hit Every Number
Three KPIs moved for ninety days. Two cleared comfortably past floor; the third landed exactly on it. None reached stretch. The system didn't need every number to hit its ceiling to prove the cadence was real — it needed the client running the whole cycle alone, and a verdict that said precisely what happened.
Day 0 → Day 90, at a glance
Named at the KPI Panel Contract on Day 1, held all the way to the Day 90 verdict — nothing in the final scorecard that can't be traced back through a dimension to a signed atomic decision.
Three KPIs, not one, forced a real trade-off decision. Designating Working Capital as Primary meant Reconciliation had a tie-breaker the moment a Dairy safety-stock increase threatened the cash floor — no KPI could quietly win at another's expense.
The Day 10 Gate prevented scope creep before it started. Culture & Behavior was named as a candidate dimension and deliberately held back — activating it without evidence would have diluted the six dimensions that actually traced to the panel.
Client-led Cycle 2 and Cycle 3 proved ownership transferred, not just that a cadence existed. The GUSTO Gate went live under the client's own facilitation in Cycle 2; by Cycle 3, the "No BS" veto held under real pressure with zero coaching.
AI drafted. It never signed. Every scenario, every variance narrative, every executive brief came out AI-drafted — and every one of them carries a named human owner's signature before it counted as a decision.
The verdict stayed honest about what didn't happen, not just what did. All three floors cleared. None of the three stretches did. Publishing that distinction — instead of rounding "96%" up to "essentially 97%" — is the discipline FlightDeck90 is built to protect.
The Absorption Test, not the scorecard, was the real proof. A number that holds only while the consultant is present proves nothing. Cycle 4 scheduled and staffed with zero Creativa attendees is the actual evidence the cadence is Meridian's now.
One 90-day cycle. A grocery chain with a full calendar and no locked KPI, three months of AI-drafted, human-signed decisions across six dimensions, and a Day-90 report that says exactly what moved and exactly what didn't. The methodology doesn't promise every KPI clears its ceiling in one cycle — it hands over a traceable panel, a cadence proven to run without Creativa's involvement, and a verdict honest enough to plan Cycle 4 around.