Supply & Demand Fleet · Design partners
28% fewer stock-outs, on 23% less stock.
That is what our ordering did when we replayed eight months of a grocery retailer’s own sales against its own reorder points. Overnight, agents check every store order, chase the suppliers who are late and explain every number. Your planners decide, and nothing reaches SAP without a person.
- Fewer stock-outs
- 28%
- Less stock on hand
- 23%
- Pays for itself?
- See your payback →
Replay, not a live result: January to August 2026, 500 store × product lines in 25 stores of our first design partner, a grocery retailer running SAP, sampled across slow and fast sellers. Both policies ordered weekly through the same simulation; the retailer’s used its own reorder points, re-sized for weekly ordering. Its MRP as it actually runs, ordering daily, had fewer stock-outs but placed about five times as many order lines. Live figures come from the side-by-side run with each partner.
What I did overnight · Supply Manager
- Delivered 3 of 4 planning books by 07:00 (one stock file was late)
- Checked 1,860 order lines, 3,940 open POs and 2,300 forecasts
- Put everything that needs a person below, by deadline and owner
What I need from you
3 approvals are yours. Your team has 25 order lines to decide before 11:00, and DC North runs out of breaded chicken in 2 days unless the vendor delivers.
- 07:00Now
- 11:00Order cut-off
- 11:30SAP file cut
- 14:00DC causes due
- 16:00Escalate silent vendors
Before the 11:00 order cut-off
Later today
Payback
Does it pay for itself?
Put in your own numbers. The improvement defaults to our replay; change it to whatever you think is fair.
Your business
Assumptions you can change
Our replay: 28% fewer.
Our replay: 23% fewer days of cover.
The rest are bought as a substitute or later.
Every year
$1.4M
margin from sales you stop losing, plus the cost of stock you no longer carry.
- Margin recovered
- $560K
- Carrying cost saved
- $857K
- Cash released from stock, once
- $4.3M
The fleet pays for itself when it costs less than $1.4M a year, before counting the $4.3M of cash released or the planner time it frees.
An estimate from your inputs, not a quote or a promise. During the design-partner programme we measure the real figures on your stores, against matched stores, before anything goes live.
The loop
From last night’s data to this morning’s orders
Not one assistant waiting for a question. A team with roles, a deadline, and a hand-off at every step.
01Data Steward
Check the night’s data
Every extract is checked when it lands: which business day it covers, rows against a normal day, stores reporting. You know how fresh the data is before anything runs on it.
Being built02Forecast Steward
Forecast, and earn it
A forecast per store and product, graded against a simple average. Where it does not beat the simple method for a product group, the simple method is used.
In the pilot build03Replenishment Planner
Propose the orders
Order proposals under your rules: pack sizes, minimum stock, caps, supplier delivery days, promotion days. A line that breaks a rule is blocked, not sent.
Being built04Supplier Follow-up
Chase what is late
Late purchase orders are picked out and scored against what was delivered. The chase message is drafted for your own WhatsApp or email and goes only after someone approves it.
In the pilot build05Your planners
Decide
Line by line at the start. Exceptions only, product group by product group, once the record shows the agents agree with your planners.
Every line
Every number explains itself
Open any suggested order and see why: the cover it aims for, where stock stands today, the rules applied, and what happens to the shelf with and without it.
Why 24?
1 · Target
2 · Position
3 · Rules & rounding
Next 14 days at store 0412
— with this order- - without
Before go-live
Side by side with your process, then the verdict
You should not take the forecast on trust. Before go-live, the fleet runs next to what you do today and is graded every week, product group by product group.
Runs next to your process
For at least four weeks the fleet proposes every day while your MRP and your planners carry on as they do now. Nothing it produces reaches SAP.
Every number kept as it was made
Each night’s forecast and proposals are stored with the data they were built from. Your own system’s numbers are copied each night too, before SAP overwrites them.
A fair weekly comparison
Graded over the days an order has to cover, not just tomorrow. Days a shelf was empty are treated the same for both sides. The measure is agreed in writing before week one.
Then pilot stores against matched stores
After go-live, the result is the change in pilot stores minus the change in comparable stores, so seasons and growth cancel out.
| Product group | Volume | Forecast error, ours | Your system | Simple average | Stock-outs vs yours (simulated) | This week | Next week |
|---|---|---|---|---|---|---|---|
| Dry grocery | 38% | 31% | 36% | 33% | −0.4 pt | Better | Keep forecast |
| Dairy | 21% | 27% | 28% | 27% | ±0.1 pt | Same | Use simple average |
| Household | 17% | 44% | 41% | 42% | +0.2 pt | Worse | Use simple average |
Error over the days an order has to cover, not the next day. Days a shelf was empty are left out for every method alike. Illustrative values.
Data
It knows how fresh its data is
Each extract is tracked from expected to landed, checked, loaded and used. When a file is late or wrong, the run says so, and your planners hear it before the cut-off, not after.
| Feed | Expected · landed | Business day covered | Rows | Stores | State | Last 14 days |
|---|---|---|---|---|---|---|
| Store sales | 03:00 · 02:41 | Mon 11 Jan | 1.21M (normal 1.2M) | 99.6% | Validated · used | |
| Store stock | 03:00 · 03:26 | Mon 11 Jan | 2.1M (normal 2.1M) | 98.9% | Late · used | |
| Open POs | 03:00 · 02:30 | Mon 11 Jan | 3,940 (normal 3,900) | — | Validated · used | |
| Goods receipts | 03:00 · 02:31 | Mon 11 Jan | 3,610 (normal 3,550) | — | Validated · used | |
| PO delivery dates | 03:00 · 02:33 | Sun 10 Jan | 0 (normal 3,700) | — | Quarantined |
PO delivery dates covered yesterday, not today, so the run used Sunday's dates and flagged every affected line. Planners were told at 06:40.
The team
Seven roles, one fleet
Each agent has one job and hands its work to the next, the way a supply team does. Your people work with all of them in one app, the Control Tower, in English or Arabic.
Supply Manager
Runs the day. Briefs each person at 07:00 with what was done overnight and what needs them, holds your policies, and asks for more autonomy only with the record to back it.
Being builtData Steward
Watches every extract arrive, checks it, quarantines what fails, and tells planners by the cut-off when a run used yesterday’s data.
Being builtForecast Steward
Keeps the forecast honest: weekly accuracy against your system and a simple average, and the switch to the simpler method where it wins.
In the pilot buildReplenishment Planner
Builds store order proposals under your rules, with a plain explanation of every quantity.
Being builtSupplier Follow-up
Tracks open purchase orders against receipts, scores each supplier on delivery, and drafts the chase for a buyer to approve.
In the pilot buildAvailability Watch
Finds items at zero in the DC and stores, pre-fills the likely cause, and routes each one to the team that can fix it.
Being builtWatchdog
Checks every run finished, every number is in the audit log, and any day can be replayed. Raises the alarm when one is not.
Being builtControl Tower
One app for planners, buyers, DC managers and leadership: today’s deadlines, worklists, approvals and results.
Being built
Control
You decide how far it goes
Autonomy is a setting your team signs, product group by product group. The rules live in code, not in a prompt.
Proposes, never places
It starts propose-only: your planners place the orders. Sending approved lines to SAP is a later step, through a channel your SAP team chooses, and only after you sign it off.
In the pilot buildYour rules are hard rules
Pack sizes, minimum stock, caps and delivery days are checked on every line. A breach blocks the line.
Being builtAutonomy is earned and signed
A trust ledger per product group shows the record. Moving to exceptions only needs a second signature and can be taken back at any time.
Being builtMore stock stays with a person
Anything that raises stock or changes a price is always decided by someone on your team.
One audit log, one stop switch
Every suggestion, decision and undo is in one log. Every agent has a stop switch, and so does the whole fleet.
Being built
Design partners
Shape the fleet on your own supply chain.
We are taking a small number of retailers and distributors as design partners. You get the fleet on your data, and a say in what it does next.
- Who it fits: retailers and distributors with stores and DCs, ordering every day from an ERP.
- How it starts: a discovery call, your data extract, then a shadow run of at least four weeks before anything goes live.
- What you agree up front: the stores in the pilot and the stores they are compared with, one primary measure, and the bar it has to clear.
- Where it is heading: more ERPs, fresh and promotions as their own phases, and planners who only see exceptions.
Questions
Before you talk to us
Does it change SAP on its own?
No. It proposes and your planners decide. Sending approved lines to SAP, through a file, a staging table or an interface your SAP team picks, is a later step you sign off. Your MRP keeps running throughout.
What does it need?
A daily extract from your ERP, delivered by SFTP: store sales, store and DC stock, open purchase orders, purchase order history, and the item and supplier masters. And your ordering rules: pack sizes, minimum stock, caps, delivery days and promotion days.
Which ERPs does it work with?
SAP today, through daily extracts. Other ERPs are where it is heading, and design partners decide which come first.
Where does our data live?
Sales and stock data stay in the cloud region agreed with you, and the forecasting can run inside your own network. Supplier identities are kept out of the model data.
Does it work in Arabic?
The Control Tower is being built in English and Arabic, right to left, and supplier messages can be drafted in either.
What does it cost?
Pricing is agreed with each design partner before the pilot starts, and you decide then whether to go ahead.