Skip to main content

Critical Path — 90-day execution

The journey from data-blind Monday to auto-Monday — owner-named, gate-named, kill-switch-named

Every week the Monday number doesn't land costs you. The path below changes that — but only if you act on it this week.

$6K–$20K

per week

cost of staying still

7–13 hrs

per week

lost to reconciliation

16 weeks

from Stage 1 start

to first auto-Monday

§1

5 Decisions — Start Here

Today

Schedule the Week 8 kill-switch review meeting

Bounded bet requires a named exit date. Without the meeting on the calendar, Stage 1 cannot start.

Owner: CEO / Managing Director — schedule it on your calendar before you close this tab

Day 3

Name the IT contact accountable for confirming POS API access

Highest-uncertainty INFERRED input — resolve fastest to remove Stage 1 budget risk

Owner: TBD — name someone now

Day 7

Approve Stage 1 budget of NZD $54,500 with named kill switches

Starts the predict-before-you-invest discipline immediately

Owner: CFO + CEO sign-off

Day 14

Name 5 Documentation Sprint owners (supplier history, pricing logic, margin floor, catalogue selection, store-mix)

Institutional knowledge encoding is on the Stage 3 critical path — start naming owners in Stage 1

Owner: TBD — one owner per knowledge domain

Day 14

Pick BI substrate from shortlist (Metabase / Hex / Power BI Premium)

Vendor commitment is the largest single Year 1 cost line

Owner: CTO / Tech lead

Why this page is designed this way

This page is a behaviour-change instrument, not a report. The pieces below are deliberate — and naming them is how trust gets built. You see the seams; nothing is hidden.

  • Cost hook (top, amber): loss-aversion + fear-sells. Losses feel 2× as heavy as equivalent gains. Loss Aversion ↗ · Fear Sells ↗
  • TODAY decision (§1, brand colour): hyperbolic discounting. People discount the future steeply — a deadline today outweighs a benefit next quarter. The Week-8 review goes on the calendar before this tab closes. Hyperbolic Discounting ↗
  • Locked stages (§2): Zeigarnik effect. Open loops stay in working memory until closed. Stage 2 and Stage 3 are visibly locked so the mind cannot file them as “someday.” Zeigarnik Effect ↗ · IKEA effect — you are co-building the path, not receiving a deck. IKEA Effect ↗
  • Positioned vs Exposed (§4): loss-aversion re-applied to the horizon. The five predictions are claims you do not control — but your readiness for them is. Each gets a binary state so the gap is visible.

Disclosing the persuasion is itself the differentiator. Manipulation hides the mechanism. Coordination names it.

§2

The Journey: 4-Stage Path

Stage 0NOW

0–0.5 mo

$0

4 GO conditions → Stage 1 launch

Any condition unmet → STOP

Stage 1

1–4 mo

NZD $108K

UC1 Unified Intelligence + UC2 saleId ROI

Monday misses W8 / API blocked D30 / over $65K W12

LOCKED

Stage 2

4–9 mo

NZD $187K

UC4 Stock-Position + UC5 FX Tracker

Unlocks when: Stage 1 kill switch not fired — Monday number lands on time for 4 consecutive weeks (W8 + W12 checkpoints)

LOCKED

Stage 3

8–12 mo

NZD $333K

UC3 AI-Assisted Buyer Brief (pilot)

Unlocks when: Stage 2 complete + Documentation Sprint complete (all 5 knowledge domains owned and documented)

§3

90-Day Execution Timeline

W
W2
W4
W6
W8
W10
W12
W14
W16
Documentation Sprint
Institutional knowledge
BI Substrate
Data ingest pipeline
Monday report loop
UC1 Business Analysis
Weekly report automation
Dual-run → auto
UC2 saleId ROI
Attribution pipeline
Catalogue optimise
Gates
GO / NO-GO
First auto Monday
Kill check #1
Budget gate
Prediction confirmed
DecisionKill checkBudget gateMilestone

GO conditions

POS API access confirmed in writing (IT lead named, Day 3)
Stage 1 budget of NZD $54,500 approved with named kill switches (CFO, Day 7)
BI substrate vendor selected from shortlist: Metabase / Hex / Power BI (CFO, Day 14)
5 Documentation Sprint owners named — supplier history, pricing logic, margin floor, catalogue selection, store-mix (Owner, Day 14)

Day 14

GO / NO-GO

4 conditions checked

Day 35

First auto Monday

Dual-run with manual

Day 56 (W8)

Kill check #1

4 consecutive on-time reports OR stop

Day 112 (W16)

Prediction confirmed

Monday number auto-delivered

§4

What's Coming — Positioned or Exposed?

Stage 1 complete = positioned for all 5. Without Stage 1: exposed on all 5. The predictions are not yours to control — your readiness is.

#1Discount tier of NZ retail will continue expanding share at 1–3 percentage points/year through 2027

⚠ Exposed

Source: Stats NZ Retail Trade quarterly + The Warehouse Group NZX disclosures

If you wait

Your customer base grows while your decisions stay data-blind — competitors with BI substrate take share

If Stage 1 ships

Stage 1 BI substrate live — you see category mix shift in real time and act on it

#2NZ-listed retail will increase AI/data investment 30–50% over 24 months

⚠ Exposed

Source: Industry pattern in NZ + AU retail transformation programmes 2024–2026

If you wait

Analytical gap between mid-market and large-format widens — talent and capital favour the prepared

If Stage 1 ships

Stage 1 demonstrates data-driven discipline — Stage 2/3 attract better hires and partners

#3FX volatility against major import currencies will remain in elevated band for 12–18 months

⚠ Exposed

Source: RBNZ monetary policy outlook + global currency volatility post-2024

If you wait

Margin erosion goes undetected until quarterly close — too late to hedge

If Stage 1 ships

Stage 2 FX tracker (UC5) — daily margin exposure visible, hedging decisions become routine

#4AI-assisted retail buying (RAG over supplier history) will become standard in AU mid-market retail within 24 months

⚠ Exposed

Source: Wesfarmers, Bunnings, Coles published data investments 2024–2026

If you wait

Buyer briefs remain individual judgement — institutional knowledge walks out the door on resignation

If Stage 1 ships

Stage 3 AI buyer brief pilot — supplier history queryable, buyer onboarding compresses 6mo → 6wk

#5NZ Privacy Act enforcement on consumer data will tighten progressively

⚠ Exposed

Source: Privacy Commissioner NZ public guidance 2023–2025

If you wait

Customer data sprawl across spreadsheets and tools — audit-grade response impossible under pressure

If Stage 1 ships

Stage 1 unified governed substrate — compliance becomes a query, not a fire-drill

§5

Your Next Step

Ready to commit to Stage 1? Two closing instruments:

The decision summary has the 4 GO conditions to check by Day 14.
The one-page plan has the commercial model, fees, and the kill switches in one screen.

Put this to work

Pressure-test this plan with your own AI assistant

For the Owner / IT Lead

Copy this prompt. Paste into Claude, ChatGPT, or any AI assistant. The page context is already loaded — send it and get analysis tailored to your role.

I need to pressure-test a 90-day technology implementation plan for a 15-store NZ discount retail chain.

THE 90-DAY PLAN:
Weeks 1-4: Documentation Sprint — 5 named knowledge-domain owners (supplier history, pricing logic, margin floor, catalogue selection, store-mix).
Weeks 1-5: BI Substrate data ingest pipeline (POS + ERP + e-commerce + catalogue → BigQuery).
Weeks 3-8: UC1 Monday report automation — currently 7-13 hours manual, target ≤15 min auto, 3 consecutive weeks.
Weeks 4-12: UC2 Catalogue saleId ROI Dashboard — currently zero attribution data on which promotions drive revenue.

KILL GATES:
Day 14: GO/NO-GO on 4 conditions (highest-risk: POS API access in writing).
Week 5: First automated Monday report, dual-run with manual reconciliation.
Week 8: Kill check — if Monday number has not landed 4 consecutive times, stop at NZD $30-40K sunk cost.
Week 12: Budget gate — stop if spend exceeds NZD $65K.
Week 16: Prediction confirmed when 4 consecutive Mondays land automatically by 8am.

HIGHEST UNCERTAINTY: POS system API access is INFERRED, not VERIFIED. This is the single biggest schedule risk.

WEEKLY COST OF STAYING STILL: NZD $6K-$20K per week (CFO + Finance Manager time + 2-4 day decision lag on every weekly call).

What should happen in the first 48 hours of this project to de-risk the POS API access question? What is the most common failure mode for a 90-day BI implementation, and how does the kill gate structure above address or fail to address it?