Wholesale ordering that runs itself
A Mile End roastery selling to sixty cafes was taking wholesale orders by text message and transcribing them by hand every Monday. We built them a storefront with a separate wholesale portal, tiered pricing and standing orders.
The problem
Boreal roasts on Tuesdays and Fridays, and everything about the business is timed around those two mornings. The problem was that nobody knew what to roast until the orders were counted, and the orders arrived as sixty separate text messages spread across a weekend.
One person spent most of Monday reading messages, typing quantities into a spreadsheet, and calling back the cafes whose orders were ambiguous. Mistakes were expensive in a way that was hard to see: a bag short meant a same-day drive across the city, and a bag over meant coffee that went stale on the shelf.
What we did
We kept the public storefront deliberately simple, because retail was never the bottleneck. The work went into the wholesale side.
Each cafe got an account with its own negotiated pricing, its own order history, and a one-tap reorder of whatever it bought last time. Standing orders let a cafe set a recurring quantity once and forget it, with a reminder the day before the cutoff so they can adjust. Everything closes automatically at the Sunday and Wednesday cutoffs, and the roastery gets a single production sheet grouped by bean and grind.
We also built the account signup so Boreal approves it, not the software. Wholesale pricing is a relationship, and they did not want it handed out automatically.
What changed
Monday order entry went from around three hours to about twenty minutes, and that twenty minutes is now spent checking a production sheet rather than transcribing messages.
Standing orders were the surprise. Just under half the wholesale accounts set one up in the first two months, which made demand predictable enough that Boreal could commit to green coffee earlier and at better prices. Short-shipments effectively stopped, because quantities are now entered by the person who wants the coffee.
More work
- E-Commerce2025
Maison Papier
Fewer steps, more finished orders
Checkout abandonment fell by 41% after we cut the flow from five steps to two.
View project - AI Automation2025
Clinique Sherbrooke
An intake assistant that books the right appointment
Booking moved off the phone for most patients and no-shows fell by roughly a third.
View project - Web Development2024
Atelier Marchand
A catalogue that loads before you notice it
A 400-piece catalogue that opens in under a second on a phone, on hotel wifi.
View project