You did not buy
too much. You put it
in the wrong counter.
Short-dated stock standing where it turns slowest, and the deep range missing from half the estate — two problems, one cause.
Illustrative · not a client counter
The AI that
Repeat-purchase rate, shade availability, expiry exposure and counter productivity — on one board, narrated by Iris.
The date. The shade curve.
The refill.
Beauty is the one retail category where stock has a clock on it, a range behaves like a size curve, and the customer is guaranteed to come back — if you reach her in the right week.
Batch and shelf life tracked per line, so nothing sells past its date and nothing sits past its margin.
Iris clears it before it expires, not after.
Which shades break first, by store and by season — because a foundation range sells like a size curve, not like a product.
Iris knows which shade goes first, and where.
When a customer actually runs out, and the prompt timed to land a week before rather than a month after.
Iris knows when the bottle runs out.
Every counter
on one board.
Rules fire. Models predict.
Iris explains.
| Rule | Threshold | Now | State |
|---|---|---|---|
| Gross margin | ≥ 53% | 57.1% | clear |
| Repeat purchase rate | ≥ 54% | 58% | clear |
| Shade availability | ≥ 90% | 94.2% | clear |
| Replenishment accuracy | ≥ 88% | 91.8% | clear |
| Model | Predicts | Accuracy |
|---|---|---|
| Shade curve | break order by shade | 92.5% |
| Refill timing | days to run-out | 91.3% |
| Expiry risk | cover vs shelf life | 93.0% |
All four rules read clear, and 0.9% write-off against 2.4% last year looks like a solved problem. The expiry model is what sees that two batches carry ninety-one days of stock against thirty-eight days of life. Iris then states the only choice actually available: move them while they are still worth full price, or book the write-off later.
Pick one. It opens
the real screen.
Repeat-purchase rate, shade availability, expiry exposure and counter productivity on one board.
Point of Sale
A beauty till sells a shade, a routine and a return visit. At 124 baskets an hour it also has to know which batch just left the counter, because that is what makes traceability possible at all.
The batch leaves with the sale. Which batch went to which customer is captured at the till, which is the only reason a recall can be a phone call rather than a press release.
The shade is the customer record. A recorded shade is what makes the next visit a two-minute sale instead of a twenty-minute match.
Iris knows what completes the routine. The prompt is the step missing from what she already uses, not a generic add-on — and refill capture at 44% is how it is judged.
Balanced Scorecard
Margin, repeat purchase, shade availability and replenishment accuracy on one live board — with Iris naming the batches that will not clear before their date.
Write-off belongs next to margin. 57.1% gross margin and 0.9% expiry write-off are the same conversation, because in beauty the second quietly funds the first.
Repeat purchase is the category's real KPI. 58% and climbing says the routine is working; refill capture at 44% says how much of it you are actually collecting.
The fourth lens audits the AI. Replenishment accuracy sits beside margin, because a counter is only stocked as well as the model behind it.
Numerus · CFO
Trial balance, P&L, balance sheet and cash flow on a ledger that ties out to the penny — across 63 counters, many of them concessions where the settlement is somebody else's arithmetic.
Short-dated stock is not worth its cost. £46k sits inside ninety days; the ledger values it against the date rather than against the invoice.
Concession settlement is reconciled, not accepted. A host store's statement is checked against your own till record, which is the only way a 57.1% margin survives contact with sixty-three of them.
Ask Iris for any number. Any counter, any batch, any line, any period — answered in the conversation rather than a week after close.
Nexus · Supply Chain
Beauty stock has an expiry date, which turns replenishment into a race rather than a level. The question is never only how much, it is how much and how long has it got.
Cover is read against shelf life, not against a reorder point. Ninety-one days of stock is healthy or fatal depending entirely on what the batch has left.
Batch is the unit, not the SKU. Two deliveries of the same line are two different clocks, and treating them as one is how a write-off happens quietly.
Moving beats marking down. Iris prices a transfer against a discount, so the cheap answer is chosen before the expensive one becomes the only one.
Motus · Operations
The wall, the testers and the hygiene standard behind them. A beauty counter is judged on how it looks from two metres away and how clean it is from two centimetres.
Planogram compliance is a stock fact. A shade missing from the wall and a shade missing from the stockroom look identical to a customer and identical in the sales data.
Testers are consumable inventory. They are counted, replaced on a hygiene schedule and costed — not treated as marketing spend that nobody measures.
Iris watches the exception, not the routine. The counters carrying the full range stay quiet; the thirty-two that do not are named.
Manus · Workforce
Beauty advisers are the product. A shade match, a routine built at the counter and a customer who comes back in eight weeks are all one person's work — and often that person is paid by somebody else.
Concession staffing is a shared roster. Host-store hours, brand hours and your own sit on one record, because the counter does not care who employs whom.
Repeat purchase is coachable. 58% across the estate hides the spread between counters, and the spread is almost entirely who is standing there.
Hours, payroll and compliance on one record. Who worked, where, at what rate — without a second spreadsheet per host store.
Every solution,
on the real screen.
Grouped by what they act on. Pick one and it runs.
How Iris enforces the rules.
Thresholds that fire on the shift the breach happens (the “Zen Rules” layer) — the date on the batch, the tester on the wall, the concession statement and the ingredient file. Four solutions run on this layer.
Two skincare batches carry ninety-one and seventy-eight days of stock against thirty-eight and forty-four days of life — and they are standing in the three slowest doors on the estate.
Batch, expiry & shelf-life control
| Batch | Line | Expires | Cover | State |
|---|---|---|---|---|
| B-2291 | Hydrating serum 30ml | 38 d | 91 d | will not clear |
| B-2288 | Night cream 50ml | 44 d | 78 d | will not clear |
| B-2310 | Vitamin C ampoule | 62 d | 54 d | tight |
| B-2276 | Lip oil | 81 d | 39 d | clears |
| Other 41 batches | — | >120 d | — | clears |
Two batches have less than the cover they need to sell through — 38 days of life against 91 days of stock. Worse, they sit in the three slowest counters. Moved to KaDeWe and Selfridges, which turn skincare twice as fast, both clear at full price instead of becoming a write-off.
Tester & sampling consumption
Testers are the only stock in retail that is bought to be given away, and they are almost never counted. In a category running 57% gross margin, an untracked tester line is margin leaving the counter one pump at a time.
Testers are held as inventory with their own consumption rate, not written off at delivery and forgotten.
Consumption per counter is compared against sales of the same line — a tester burning fast next to a shade that is not selling is a signal, not a coincidence.
Hygiene replacement runs on a rule with a date, because a tester past its life is a compliance issue before it is a cost one.
Sampling is attributed to the sale it produced where the customer is known, which turns a marketing cost into a measurable one.
Tester stock carries the same batch and expiry discipline as sale stock, since it comes from the same production run.
Counter & concession settlement
Most beauty counters sit inside somebody else's store, on somebody else's till, settled on somebody else's statement. Contribution per counter is only knowable if that statement is checked rather than accepted.
Concession terms are held as rules — commission tiers, deductions, staffing recharges and settlement timing — rather than as a contract in a drawer.
The host store's statement is reconciled against your own till record line by line, and the difference is a query raised inside the window.
Contribution is reported per counter after all of it, which is the only number that answers whether a door is worth keeping.
Owned counters and concessions sit on one board, so sixty-three doors can be compared on the same basis.
Settlement timing is modelled into cash flow, because in a concession estate the sale and the money are weeks apart.
Ingredient & recall traceability
Cosmetics is a regulated category applied directly to skin. A recall is not a stock exercise — it is a question about which batch reached which customer, and it has to be answerable in hours.
Ingredient and allergen data is held against the batch, so a regulatory change can be assessed across the range without opening a single file.
Because the till captures the batch, an affected sale is identifiable — a recall becomes a contactable list rather than a public notice.
A flagged batch is blocked at the till by rule across all sixty-three counters at once, not by an email asking staff to check.
Market-specific ingredient restrictions are enforced per country, which matters across a European estate.
The audit trail is a by-product of trading, not a document assembled after an incident.
How Iris predicts the wall.
What the range and the routine do next (the “Zen Models” layer) — modelled early enough to act on, with the cause named and the move costed. Two solutions run on this layer.
Shade-curve demand
| Shade | Share | Cover | Counters out | State |
|---|---|---|---|---|
| 14N Medium neutral | 11.4% | 2.1 w | 9 | breaking |
| 16W Medium warm | 9.8% | 2.6 w | 6 | thin |
| 22N Deep neutral | 7.1% | 1.4 w | 14 | broken |
| 04C Light cool | 3.2% | 9.8 w | 0 | over-stocked |
| Other 18 shades | 68.5% | 5.4 w | — | healthy |
The deep range is not under-selling, it is under-distributed — only 31 of 63 counters carry the full set, and where they do it hits plan. 22N is out in 14 counters while 04C sits on 9.8 weeks of cover. That is an allocation decision made on the wrong data.
Replenishment cycle prediction
Beauty is the rare category where the repeat purchase is close to certain and only the timing is in question. A bottle empties on a schedule; the retailer who knows that schedule owns the refill.
Days to run-out is modelled per customer per line at 91.3% accuracy, from size, purchase history and how she actually uses it — not from an average cycle.
The prompt lands a week before the bottle is empty. A month after, and she has already replaced it somewhere else.
Refill capture at 44% is the number this model exists to move, and it is reported next to repeat purchase rather than hidden in CRM.
Predicted refills feed counter replenishment as base demand, so the shade she is coming back for is on the wall when she arrives.
The same cycle view is what makes 91.8% replenishment accuracy possible — known demand is easier to stock than guessed demand.
Ask Iris · the counter
Not a screenshot — the actual agent, reading the same estate the rules and models write to.
Twelve core POS solutions,
in every sector.
The sector solutions sit on top of these. They are not an upsell and they are not configured per customer — every Zentallio retail deployment ships with all twelve.
Implementation,
run by Iris.
You configure a sector playbook, not a custom project. Iris applies it herself — agentically, from day one.
Onboarding is agentic. Iris connects the batch file, the shade range and the concession agreements directly — no manual data mapping.
Customer purchase history is imported, not retyped. Past shades and sizes carry in, which is what makes refill timing useful in month one rather than month twelve.
Live in weeks, learning from day one. Go-live is a configuration, not a project plan — and every batch already on hand has a date running.
Support starts
with Iris.
The first line of support is agentic — Iris resolves most of it herself. Our engineers pick up from there.
Layer 1 — Iris, 24/7. Configuration questions, anomalies and routine issues resolved directly, instantly.
Layer 2 — our engineers. Anything Iris can't close escalates automatically to a Zentallio engineer.
No blank tickets. Every escalation arrives with Iris's own diagnosis — engineers start from an answer.
Cleared, not written off.
while it is still stock
shelf life
not under-selling
a week early
Where good stock
quietly becomes a write-off.
Cover measured against a reorder point instead of against a date, so ninety-one days of stock on thirty-eight days of life reads as healthy.
A deep range judged as under-selling when only half the estate carries it — the wrong conclusion drawn from the right number.
A customer who was always going to repurchase, reached a month after she already replaced it somewhere else.
Every batch carries its own clock, testers are counted as stock, concession statements are reconciled and a flagged batch blocks at all 63 tills at once.
Shade break order at 92.5%, days to run-out at 91.3%, cover against shelf life at 93.0% — so the cause is named, not guessed.
“Why, and what do I do?” — move them now, and they clear at full price instead of becoming a write-off.
The same spine,
a different floor.
Cosmetics & Beauty is sector six of nine in Fashion Retail. The layers, the six products and Iris herself are the same ones running across ten Food & Beverage sectors.
on your counter.