Customer card and missing-item alerts for phone sales reps
A food wholesaler's sales reps knew from memory who orders what regularly. A fast, rule-based customer card now shows it before the call and flags what is missing from the order during the call.
- Client
- Food wholesaler, two sites, approx. 80 staff
- Area
- Development
- Status
- Demo complete, pilot in preparation
Starting point
The wholesaler has about 1,200 active regular customers and the vast majority of orders arrive by phone. Sales reps call customers from the next day’s route list and record an order of 5 to 40 lines during the call. Who regularly buys what, which day they order, and which delivery rules apply to whom is a matter of experience and memory. With sick leave or a new colleague, that knowledge is lost.
While taking a 30-line order by dictation, nobody notices that the weekly margarine is missing this time. That turns into a re-order, an extra delivery or lost revenue. An earlier helper feature had been switched off because it slowed the work down, so speed was a precondition.
The solution
- Customer card before the call: on one screen the note, the last and open orders, the 15 usual products, declining products and the current promotions relevant to the customer, with stock.
- Missing-item alert during the call: what is regular but not in the order, in a live-updating list, addable with one click at the customer’s usual quantity.
- Weekly skip warning: only fires once the customer’s usual ordering day has passed. The literal rule would paint more than half of the customers red every Wednesday.
- Tunable rule: the reps set what counts as regular (time window and threshold) and see the effect immediately.
- Product search: across 4,600 items by name or item number fragment, with the customer’s price and site stock, discontinued items marked.
- Stock by the customer’s site: only sellable warehouses count, not blocked or transit stock.
Deliberate limits. The demo does not write into the ERP, does not save orders, adds nothing on its own and uses no language model: it works with fast, predictable rules and runs offline. It was built on fictitious data, so no customer data left the company during development. The final decision and approval always stay in human hands.
How we worked
- On-site survey. Observing the phone sales reps’ work, recording the real screens and habits.
- Rules. Refining the client’s description, filtering out noise-producing cases, every rule tested.
- Realistic demo. 1,200 customers, 4,600 products, 12 months of orders with seasonality and built-in alert situations.
- Measurement and demo. Speed measurement on the full data volume, presentation to management and the sales team.
Next steps
A real export sample from the ERP, a pilot with 2–3 reps on an internal server, finalising the rules based on experience, then daily automatic data refresh. Business results will be measured after the pilot; until then we only publish numbers actually measured on the demo.