How Zentallio turns four lenses of your operation into one live decision engine — walked through from start to end.
→ or Play to beginSpace play / pause
01 · The framework we build on
The Balanced Scorecard, as Kaplan & Norton drew it.
Four perspectives, linked by strategy and read from the base up — each translated into Objectives · Measures · Targets · Initiatives. For three decades this is how operators turn strategy into numbers.
Financial
“To protect margin as we scale — what must each outlet return?”
ObjMeasTargInit
Customer
“To turn first orders into regulars — how must we show up?”
ObjMeasTargInit
Vision & Strategythe centre
Internal Process
“To serve every order on time and on-spec — what must we master?”
ObjMeasTargInit
Learning & Growth
“To improve shift after shift — how must the operation learn?”
ObjMeasTargInit
The base — Learning & Growth — is the enabler that sustains every perspective above it.
02 · Why food retail strains it
Food retail moves faster than a quarterly review.
A multi-outlet food business is a measurement problem in constant motion — and a scorecard reviewed each quarter, built on people learning, can't keep the pace.
Thin margins
A two-point food-cost drift at one branch can pull the whole group below plan.
Perishable stock
Waste and mis-forecasting convert directly into lost margin.
Outlet variance
No two outlets behave alike across dayparts, markets and menus.
Real-time data
Every till receipt, delivery and review is emitted live, all shift long.
Classic cadence
Reviewed quarterly · paces on people learning
→
What the operation needs
Caught every shift · paces on system learning
03 · The proposition
Swap the base: from people learning to system learning.
Kaplan & Norton define the base not by a stakeholder group but by a question — how do we sustain the ability to improve? In an AI-run food operation that faculty is the system learning. So AI Learning takes the base — and unlike the old base, it is directly measurable.
cause → effect
Financialthe outcome
Margin protected as the group grows.
Customerwho pays for it
First orders become loyal regulars.
Operationalinternal process
Every order on time and on-spec.
AI Learninglearning & growth · the measurable base
Detects, forecasts and acts — sharper with each cycle. forecast accuracy · models live · drift caught
04 · The operating model
One engine per outlet — metered once.
Every perspective is AI-instrumented — the exhaust of the operation. One engine ingests it, grounds each decision in the outlet's own strategy, authors the plan and tracks it to outcome — read by a single meter: is it getting better?
AIFinancialmargin & cost
AICustomerrepeat & NPS
AIOperationalon-time & prep
AI Learning · the engine
Detect → forecast → explain → act → verify
92.8%next-day forecast9models live2drift caught
ObjectivesMeasuresTargetsInitiatives
↻ initiatives reshape activities → fresh data
05 · The scorecard, live
Health at a glance, quarter by quarter.
Q3 FY 2026 · trajectory shows Q1→Q4 · click any KPI to drill into markets
KPI
Owner
Target
Now
Δ tgt
Trajectory
FinancialProtect margin as we grow
Net margin
Finance
≥6.0%
6.2
+0.2
Food cost
Kitchen
≤30%
30.8
+0.8
Labour cost
Ops
≤28%
27.4
−0.6
CustomerTurn first orders into regulars
Repeat orders
Marketing
≥40%
43
+3
NPS
CX
≥+35
+41
+6
Complaint rate
CX
≤2%
1.6
−0.4
OperationalOn time, on-spec, every order
On-time delivery
Ops
≥95%
95.4
+0.4
Avg delivery (min)
Ops
≤30m
27
−3
Prep accuracy
Kitchen
≥98%
97.2
−0.8
AI LearningGet sharper with each cycle
Forecast accuracy
Data
≥90%
92.8
+2.8
Actions auto-drafted
Data
≥80%
84
+4
Models live
Data
—
9
+1
Market detail
—
Ask Iris · use your own numbers
Paste your KPIs in the format below — one per line. Zen builds your scorecard and reads it back to you. Nothing leaves your browser.
Format › Lens, KPI, dir, target, Q1, Q2, Q3, Q4 — dir = min (higher is better) or max (lower is better)
Iris · Ask Iris
Lens
KPI
Target
Now
Δ
Trend
06 · Layer 1 · Zen Rules
Zen Rules — the guardrails.
Deterministic thresholds you set once. The instant a lens breaches target — food cost over plan, delivery over 30 minutes — a rule fires. Zero latency, zero ambiguity.
Rule
Threshold
Now
State
Net margin
≥ 6.0%
6.2%
clear
On-time delivery
≥ 95%
95.4%
clear
Food cost · Austria
≤ 30%
32.0%
fired
Delivery time
≤ 30m
27m
clear
Complaint rate
≤ 2%
1.6%
clear
1 rule fired · routed to Layer 2 for the forecast
07 · Layer 2 · Zen Models
Zen Models — the forecast.
Gradient-boosted models learn each outlet's rhythm and project where every lens is heading — before the shift ends. Nine models live at 92.8% forecast accuracy.
Model
Predicts
Accuracy
Food-cost drift
margin risk by item
93.1%
Demand forecast
covers per daypart
94.2%
Delivery ETA
on-time risk
91.6%
Repeat & churn
who lapses next
90.4%
Labour demand
staffing gaps
92.0%
9 models live · 92.8% blended accuracy · margin forecast to recover 5.6% → 6.2% by close
08 · Layer 3 · Ask Iris
Ask Iris — the answer, in plain language.
Iris reads all four lenses together, names the cause, and recommends the action. Ask in words — “which branch is trending over food cost?” — and get the answer, not another chart.
YouWhich branch is trending over food cost?
Iris Ask IrisAustria — 32% against a 30% plan. It's the one branch pulling group net margin below 6.2% this week.
YouWhy, and what do we do?
Iris Ask IrisCause: portion drift on 3 items. Recommend: re-issue the portion guide, cap the cheese scoop to 95 g. I've drafted the kitchen notice — approve to send.
09 · Live
Cause → action, watched live.
Net margin · this period6.2%
↓ Austria · food cost 32% vs 30% plan
↑ Recovers to 6.2%
Iris Ask Iris
Cause: portion drift on 3 items in Austria. Recommend: re-issue portion guide, cap the cheese scoop to 95 g.
Action sent to Austria kitchens · 09:42Food cost corrected · margin recovering
10 · The loop
One closed loop, running every shift.
Detect → forecast → explain → recommend → act → verify → learn. Each pass trains the models, so the next drift is caught earlier. The scorecard doesn't just report — it acts, then checks its own work.