Short answer. Capacity is a calendar per resource and per day, shift hours, rate, planned stops, leave, written by the manager and changed without development. Load is the sum of routing durations of launched and promised orders, per resource and per day, recomputed at every launch and corrected by shop-floor clock-in. The bottleneck is the period’s most loaded resource, and it sets the date: the promise is computed by checking components, available or expected at the measured lead time, finding the first free slot on the bottleneck, placing the routing around it, and adding the round that delivers the zone; it is written on the order with its computation and sent to the customer. An emergency moves an unpromised order or adds hours, with the cost shown and a written decision; moved customers are warned beforehand. Real clock-in against the routing corrects durations the following month, so that load stops being optimistic. According to the common definition, the “available to promise” answer rests on stock, planned receipts and orders and requirements already committed; this guide adds the bottleneck’s capacity.
This guide is for the maker who sells to order, juice, cosmetics, joinery, furniture, packaging, parts, the wholesaler who packs, the workshop that gets emergencies every day, and the founder based in France who wants to know, before answering a customer, which date is true. It does not repeat how a production order is launched and costed, described in the manufacturing guide, nor how a made-to-measure product is configured, described in the configurator guide; it describes what, between the order and the launch, answers the question “by when?”, and it extends our software engineering page.
Promising a date, in six beats
The bottleneck slot is the accented beat: the most loaded resource sets the date, and the rest is placed around it.
Receive the order: the quantity, the product or configuration, and so the routing and the BOM, are known at entry, whether the order comes from the counter, the portal of the B2B ordering guide or the configurator. Check components: available in the ledger of the stock guide, reserved by other orders, or expected by a reception at the measured lead time of the replenishment guide; the latest component date is the first bound. Find the bottleneck slot: on the most loaded resource, the first day where remaining capacity covers this order’s routing duration, after the component bound.
Place the routing: upstream operations, pressing, cutting, go before the bottleneck slot, downstream operations after, with their routing durations. Add the round: the first round of the rounds guide serving the customer’s zone after the end of the last operation; for a fitting, the work order of the field service guide. Write the promise: the date is written on the order with the four bounds that produced it, and sent to the customer through the notifications guide; the seller sees a date, not an estimate.
Scheduling, in five layers
Load is the accented layer: it is not estimated, it is the sum of routings, corrected by clock-in.
Capacity. A calendar per resource: shift hours per day, rate, planned stops, leave, declared breakdowns; written by the manager, changed without development, because capacity is not a constant. Load. The sum of routing durations of launched and promised orders, per resource and per day, recomputed at every launch by the jobs of the background jobs guide; shop-floor clock-in described in the manufacturing guide replaces routing duration with real duration as it goes, and remaining load falls or rises continuously. Promise. Components, bottleneck slot, placed routing, round; date written with its computation.
Rescheduling. An emergency, a breakdown, a supplier delay: the application shows the options, move an unpromised order to a date, add hours on the bottleneck, and their cost; a promised order does not move without a written decision by whoever has the right under the roles guide, and moved customers are warned beforehand. Measurement. Dates kept, load per resource, overtime, orders moved, and the gap between routing duration and clock-in, in the manager’s dashboard; that gap corrects routings the following month.
Load against capacity: three resources, ten days, one bottleneck
The figure shows a fortnight’s load plan. The bottleneck is not the loudest resource, it is the most loaded.
Three resources: the press, 8 hours a day; the bottling line, 8 hours at 400 bottles an hour; joinery cutting and assembly, 16 hours over two shifts. Each cell is the sum of routing durations of orders launched that day, against the capacity written by the manager. The press is at 61 %, cutting at 48 %; the line exceeds its capacity on 18 and 19 November, 110 % and 105 %, and it sets every date of the fortnight. The first free eight-hour slot on the line is the 24th.
The manager sees the load plan every morning and can act on it: add two hours on the line on the 18th, which costs overtime and frees 0.8 hour; or let an unpromised order slip to the 24th. A breakdown on a Tuesday sets the day’s capacity to zero, and every order on the line is rescheduled with customers warned before the time. Resources, capacities, orders and hours are a constructed example.
«Load is not estimated: it is the sum of routing durations of launched orders, per resource and per day, recomputed at every launch and every clock-in.»
Six real situations, and what the application does
Every row is a sentence heard in a make-to-order workshop, and the answer is a computation, not caution.
The seller promises the 18th, we deliver the 25th: the date is computed from the bottleneck, components and round, and written with its computation; the seller no longer guesses. Everyone is urgent: promised orders are protected; an emergency moves an unpromised order, with a written decision and a warned customer. The line is full, the press is empty: load is read per resource, the bottleneck sets the date, and upstream aligns on it instead of producing juice that waits in a tank.
The supplier delivers late: components are checked at promise time, with the expected reception at the measured lead time, not the promised one. A breakdown on a Tuesday morning: the day’s capacity goes to zero, orders are rescheduled, and customers are warned before the time, as in the rounds guide. The routing says 6 hours, the shop takes 8: real clock-in is compared with the routing, and durations are corrected the following month; an optimistic load is a wrong load.
An order arrives on 11 November: the computed date
The figure follows an order of 3,000 bottles received on a Tuesday. The seller would have said the 18th; the computation says the 23rd, and says why.
Materials: 4,200 kilos of oranges are needed, 3,100 remain; the supplier’s reception is expected on the 14th, at the measured lead time of its last six deliveries; first bound. Press: six hours, first slot on the 16th. Line: seven and a half hours; the line is busy until the 19th with two orders already promised, the new order is placed on the 20th. Round: the North round leaves on Monday, the 23rd. The promised date is therefore Monday 23 November, written on the order with these four bounds, and the customer receives it on WhatsApp.
If the customer wants it sooner: to deliver on the 20th, order PO-0415, which is not promised, must slip to the 24th; the application proposes it, the manager decides, and the other order’s customer is warned. To deliver on the 18th is impossible: no oranges before the 14th, and the line at 110 %; the only option would be two hours of overtime plus emergency materials, and the application says so rather than letting it be promised. A deposit collected at order time, following the quote and invoice guide, freezes the promise. Order, quantities, dates and slots are a constructed example.
«The promised date is computed from three known things: components, free slots on the bottleneck and the round that delivers; it is written on the order with its computation.»
Scheduling and promising in 2027, in weeks
Seven weeks, the first without code, and two for capacity and load, because the promise is only a reading of those two layers.
- What the manager sees: the morning load plan, the week’s bottleneck, the day’s promises with their bounds, orders moved and why.
- What the seller no longer does: guess a date, call the workshop to find out, promise the impossible, call the customer back to apologise.
- What is measured: dates kept, load per resource, overtime, orders moved, routing-vs-clock-in gap, average time from order to delivery.
Week 1, rules: resources, their daily capacities, shifts and stops, the protection rule for promised orders, who decides a move; with the manager and the shop-floor supervisor. Weeks 2 and 3, capacity and load: the capacity calendar, load from the routings of launched orders of the manufacturing guide, the load plan screen, the computed bottleneck; compared with the last three real weeks to check that load tells the truth. Week 4, promise: components, bottleneck slot, placed routing, round, date written on the order with its computation, sent to the customer.
Week 5, rescheduling: emergencies, breakdowns, supplier delays, options and costs, customers warned. Week 6, clock-in and correction: real durations against routings, monthly correction, remaining load continuously. Week 7, measurement: dates kept, load per resource, overtime, orders moved, Monday review. A service workshop, a salon or a garage booking slots follows the same logic with the booking guide; here the resource is a machine and the duration comes from a routing.
- Where does the promised date come from?
- From four bounds: the date when all components are available, the first free bottleneck slot covering the routing duration, the end of the last placed operation, and the first round serving the zone. The date is written on the order with those four bounds, so anyone can re-read it.
- Why the bottleneck and not every machine?
- Because the most loaded resource sets the date anyway: if the line is full until the 19th, pressing juice on the 12th only fills a tank. The bottleneck is computed every day, and it can change from one week to the next.
- What happens when an emergency arrives?
- The application shows the options and their cost: move an unpromised order to a precise date, or add hours on the bottleneck. A promised order does not move without a written decision by whoever has the right, and the moved customer is warned before the time.
- What if the routing is wrong?
- Shop-floor clock-in, start and end per operation, is compared with the routing duration; recurring gaps correct the routing the following month. A load computed on optimistic routings gives optimistic dates, and it is the first gap to watch.
- Who decides capacity?
- The manager, per resource and per day: shift hours, rate, planned stops, leave. It is a calendar, changed without development; a declared breakdown sets the day’s capacity to zero and reschedules.
- Does it work for made-to-measure?
- Yes: the accepted configuration of the configurator guide gives the routing and the BOM, hence the load and the components; the promise is computed the same way, and the quote can show it before acceptance.
- How long to set this up?
- Seven weeks in the typical calendar, the first without code and two for capacity and load. A workshop that already has the routings and orders of the manufacturing guide takes less; one starting from a whiteboard takes all seven.