Somewhere in your order flow today, a system promised a customer a date. It did so with an available-to-promise check: current inventory plus planned inbound supply, netted against existing commitments, covers the request, therefore promise. The logic is decades old, runs deterministically, and is treated as plumbing, necessary, invisible, and nobody's strategy.
Look closely at what that lookup just decided. It made a bet: the planned inbound it counted on is a forecast, and forecasts miss. It priced a risk. If the inbound slips, the choice between a broken promise and an expedite fee has real costs, none of which entered the calculation. And it allocated scarcity. The units it committed to this order are now unavailable to every later one, including the strategic account's order arriving tomorrow.
A bet, a price and an allocation, executed thousands of times a day by a lookup that understands none of the three.
What first come, first served actually optimises
Under comfortable supply, naive promising is harmless; everyone gets their date. Its costs concentrate exactly when the decision matters, under scarcity, and they compound:
- The reliability cost. Deterministic ATP treats planned supply as certain, so promise reliability degrades precisely when supply gets volatile, which is when customers are watching reliability hardest. The promise that had a 60% chance of holding was issued with the same confidence as the one that had 99%.
- The allocation cost. First come, first served allocates your scarcest units by ordering speed. The distributor with an automated ordering bot outdraws the strategic account with a quarterly cadence, and the system executes that outcome faithfully, order after order, with no one ever deciding it.
- The margin cost. Under scarcity, the marginal unit could go to its highest-value use: the contract with penalties, the customer at churn risk, the product with the richest margin. A value-blind allocator donates that optionality, silently.
Nobody in the building would let a junior employee hand out the last units of constrained supply by queue position. The ERP does it at scale, and it is called standard functionality.
Promising as the decision it is
The upgraded version makes the three hidden decisions explicit:
- Probabilistic promising. Every candidate date carries odds, computed from the actual reliability history of the supply it depends on: this date at 97% confidence, three days earlier at 80%. Which risk to quote becomes a policy by customer tier, not an accident of netting logic. And the odds are backtested. Of the dates promised at 97%, about 97% should have held.
- Value-aware allocation. Under scarcity, the question changes from "is supply available?" to "where does this unit earn or protect the most?", scored on margin, penalty exposure, strategic priority and churn risk, within rules the business sets openly. Allocation becomes a policy someone can read, defend and tune, rather than an emergent property of order timing.
- Repromising as a managed event. When supply moves, the system knows every promise at risk, ranks the exposure, and proposes the least-damaging reshuffle before customers discover it, converting a week of firefighting into a half-hour of decisions.
The commercial framing deserves to be said plainly: order promising is a pricing-grade lever wearing an IT costume. Firms tune pricing with committees and science, then hand the physical allocation of scarce product to a first-in queue. Treating the promise with the same seriousness as the price is one of the few margin levers most companies have never pulled.
Every promise is a bet, a price and an allocation. Make all three on purpose.
Probabilistic promise dates, value-based allocation under scarcity and managed repromising are what the Prophesee Supply Chain Suite's fulfilment module provides. See what your promises have been deciding. Start here.