How to promise a ship date you can actually keep

The question a customer asks is "when will it ship". The question most farms can answer is "roughly how busy are we". Those are not the same question, and the gap between them is where refunds, bad reviews and apologetic emails live.

Why a lead time is not a date

A flat "ships in 3–5 working days" is a hedge. It is too long when you are quiet, which loses you the sale, and too short when you are busy, which loses you the customer. Neither failure is visible until it has already happened.

What a forecast needs to know

• Every task still outstanding, and how long it takes on each printer that can run it
• When each printer actually becomes free, including the job on it now
• Filament changes, and which colours are mounted where
• The measured failure rate of each part, per machine
• Which orders have earlier promises and will jump the queue

With those, assigning work to whichever printer finishes it soonest produces a finish time for the last task in each order. That is a date. It moves when reality moves, which is the point.

The useful part is the slip, not the date

A forecast that only tells you the date is a nicer lead time. A forecast that is recomputed whenever anything changes tells you something better: that order #1182 is now going to miss its promise by two days, while there are still two days to do something about it.

At that point you have real options — reprioritise it, split it across more machines, or email the customer before they email you. All three are fine. Finding out on the promised day is not.

Be conservative on purpose

A forecast that assumes nothing fails is optimistic in exactly the cases where you most need it to be right. Including your own measured failure rate makes the date slightly later and dramatically more repeatable, and a date you hit is worth more than a date that sounds good.