Online Orders Stop Being Re-Typed

The problem

Selling your own brand direct makes the store the front of the plant — and it is not connected.

With direct-to-consumer and HORECA sales, orders arrive through the online store already structured: SKU, quantity, delivery date. But if the store isn't wired to production, someone copies it by hand.

• With direct-to-consumer and HORECA sales the order already arrives structured — SKU, quantity, delivery date, channel — which is what makes retyping it particularly absurd.
• But if the store is not wired to production, somebody copies it by hand: a gift box of three references becomes three lines, and a mistyped one becomes a wrong pack shipped under your own name.
• The store keeps selling whatever its stock field says, and that field is only as true as the last time somebody updated it from the cold store.
• And the three channels do not behave alike: a consumer order is a parcel tomorrow, a HORECA order is a delivery slot, a retail order is a pallet with a lead time — all planned together, by hand, out of an inbox and a spreadsheet.

How it fits the IRIS system

Connect between the store and the ERP — one API, one flow, three channels.

iLEAN Connect stitches the online store to the production ERP via API. What happens in the store is automatically reflected in the system that plans the batch. The moment payment is confirmed, the order flows straight into the ERP, with the right SKU, channel (B2C, HORECA or retail) and delivery date already in place.

This is the case that only exists when the brand is yours: a plant that packs for somebody else never has it. The moment a payment is confirmed the order is already in the system that plans the batch, with its SKU, its channel and its delivery date attached — and planning stops being a reconstruction of what the store sold yesterday.

See the full IRIS architecture →

Before and after

Today's order flow versus the connected store

Aspect	Today	With iLEAN Connect
From payment to production	Somebody copies it by hand	Automatic, in seconds
A gift box or assortment	Exploded into lines by hand	Exploded into its SKUs automatically
Channel of the order	Worked out from the address	B2C, HORECA or retail, attached
Store stock	Whatever was last updated	What is actually in the cold store
A campaign weekend	Orders queue up unentered	Volume changes nothing
SKU errors on an order	Found when the parcel arrives	Removed at the source

Orders double-keyed, risk of SKU errors → One automatic flow from payment to production.

Impact estimate

Impact estimate — to be validated with your numbers.

The block below is an estimate to be validated against your plant's actual data. We put it forward so the committee has an order of magnitude; we refine it during the assessment.

• Estimated payback 3-6 months, from fewer SKU errors on multichannel orders and the hours now spent copying them by hand.
• One automatic flow from payment to production, instead of a queue of orders waiting for somebody to have time to key them in correctly.
• Fewer wrong parcels shipped under your own brand, which in direct sales means a public review and not just a credit note.
• And a production plan that sees consumer, HORECA and retail demand in the same place, at the same time, instead of in three separate inboxes.

Payback 3-6 months · double-keyed orders → one automatic flow. Estimated payback is 3 to 6 months, from fewer SKU errors on multichannel orders and the time currently spent copying them by hand. Estimate to validate.

And the fair question from the production manager

"Our store and our ERP were never designed to talk to each other." — they rarely are, and that mapping is the work: the store's order structure is mapped once onto the ERP's, gift boxes and assortments included, so they explode into component SKUs for production. Where a field has to be interpreted rather than mapped, it is an anchored task with the best models below 1.5% error [1], and the order still surfaces on the tablet before it enters the plan.

^[1] OpenAI paper "Why Language Models Hallucinate", 2025 — on the reliability of AI in anchored tasks.