A customer calls with an order and asks when they can have it. In many small workshops, the answer is a lead time from memory: “three weeks, usually”. Sometimes it is right. The order book, the machine already booked solid, and the supplier who takes ten days on that one part are not consulted, because consulting them takes twenty minutes and the customer is on the phone now. So the date is a guess, and every guess that turns out wrong costs more than the twenty minutes would have.
There are two different questions hiding inside “when can I have it”, and they need different machinery to answer.
Available-to-promise: is it on the shelf?
Available-to-promise (ATP) is the older and simpler of the two. ETH Zurich’s logistics course defines it as “the uncommitted — that is not yet assigned to an open customer order — portion of a company’s inventory and planned production”, and notes that “the ATP quantity is maintained in the master schedule to support customer-order promising”. In plain terms: take what is on hand, subtract what is already promised to other orders, add what is due in from suppliers or from production that is already scheduled, and what is left in each period is what you can promise.
ATP works well for stock items. If you hold finished goods and replenish them, the question really is whether there are enough on the shelf after everyone ahead in the queue has taken theirs. For example, a hardware supplier selling standard bolts could use ATP to check if the 1,000 units a customer wants are available in the next two weeks based on current stock and scheduled replenishments. It says nothing about whether you could make more, because it assumes the production plan is fixed.
Capable-to-promise: could we make it by then?
Capable-to-promise (CTP) goes further. The same course describes it as a technique that “develops the classic MRP technique for checking availability by adding in not just stock levels but also capacity and other resources, possibly even for suppliers”, and the Institute of Business Forecasting & Planning glossary calls it “an advanced order promising technique used to determine when a new, unscheduled customer order can be delivered” that “uses finite scheduling techniques, as well as such things as material and capacity constraints across multiple nodes in the supply chain”. ETH’s verdict on the trade-off is blunt: multilevel ATP “gives quicker results, but CTP gives more accurate results”.
For a make-to-order business, CTP is the one that matters. Stock is usually low or zero for the finished item, so ATP just says “no”, and the useful answer is the date the shop could finish it. For example, a custom machinery manufacturer quoting a delivery for a non-standard pump would need to check:
- Whether the steel plate required is in stock or needs ordering
- If the CNC machine has capacity to cut the parts before the requested date
- If the assembly team has time to build it between existing jobs
CTP answers these by exploding the product into components, checking stock levels, calculating material lead times, and then scheduling operations on finite capacity work centres. The result is a date that accounts for real constraints, and it is only honest if it comes from the same scheduler that plans the real jobs; a promise engine running its own simplified model will disagree with the schedule, and once sales stop trusting the date the exercise is pointless.
What a promise has to check
A promise that is going to be kept has to look at four things, in order:
1. Stock
Finished goods on hand, minus what is already allocated to other orders. Expired lots and quarantined stock do not count, whatever the on-hand figure says. For example, a food manufacturer might have 500kg of dried fruit in stock, but 200kg is past its shelf life and 150kg is reserved for a standing order from a supermarket chain. Only 150kg is truly available for a new customer request.
2. Material
For a made item, each component: on hand, on order with a due date, or neither. For “neither”, a supplier with a lead time; if no supplier is linked to the part, the promise cannot be dated at all, and it should say so rather than guess. A furniture maker quoting a delivery for a custom table must check not just the wood in the warehouse but also whether the special-order hinges are on a purchase order with a confirmed delivery date.
3. Capacity
The route’s operations placed on their work centres behind the jobs already there, respecting the working calendar, holidays and any downtime. The finish date comes out of that placement, not out of a lead-time field. For instance, a welder booked solid for the next 10 days cannot start a new job until the 11th, even if the theoretical lead time for welding is two days.
4. Transit
The customer asked for a delivery date, not a despatch date. A shipper that takes two days door to door turns a despatch on the 18th into a delivery on the 20th, and it is the 20th the request should be judged against. A machinery exporter quoting a delivery date for a machine in Germany must factor in the three days required for sea freight, not just the date the machine leaves the factory.
The output should be a date and the reason for it. “Earliest despatch 18 Sep, delivery 20 Sep; the constraint is the CNC cell, which is booked until the 16th” is a promise a salesperson can negotiate with. “Three weeks” is not.
Where promises go wrong
The promise ignores the queue
It costs the job from an empty shop and gives a date that would only be true if nothing else was booked. Every promise is then optimistic by the length of the queue. For example, a sheet metal fabricator promises a three-week lead time for a batch of panels, but the laser cutter is already scheduled for four weeks. The customer’s job cannot start until the fifth week.
It counts stock that cannot be used
Expired batches, stock on hold, and material belonging to another customer are on hand but not available. A pharmaceutical company might show 1,000kg of an active ingredient in stock, but 800kg is quarantined pending quality checks. Promising 500kg would tie up the remaining 200kg available for other orders.
It has no supplier lead time
A missing part with no supplier linked has no date, and a system that silently assumes a default lead time is guessing on your behalf. A bicycle manufacturer quoting a delivery for a custom bike might overlook that the carbon fibre frame has no supplier set up in the system, so the system cannot calculate when it will arrive.
It judges the request on despatch
The customer wanted it on site on the 25th; you despatched on the 25th; the promise was “met” and the customer is unhappy. A construction materials supplier delivers concrete to a site on the 25th, but the customer needs it by the 23rd to meet a project deadline. The two-day delivery window breaks the promise.
Quotes are promised as if they were orders
Tentative work should be visible in the plan at some probability, not booked as firm; otherwise every quote you send pushes every real order later. A job shop sends five quotes for CNC machining work, all promising dates based on zero existing workload. When three are accepted, the real jobs have to be squeezed into an already full schedule.
Where to start
Take the last ten orders where the delivery date was missed and, for each, write down which of the four checks would have caught it. In most shops it is capacity: the date was quoted as if the machine were free. That tells you what the promise needs to look at first, and whether a lead-time field on the product will ever be enough.
Brytebuild runs the promise on the sales order and quote lines while the customer is on the phone, using the same finite scheduler that plans the real works orders, and shows the earliest despatch, the delivery date with the shipper’s transit time, and the constraint that set them. See sales orders and lead times in the docs.