The lead time in the system is a negotiating position, not a measurement
Module · Supplier Performance and Continuity
A scorecard tells you a supplier was late. It does not say how late, how often, or what to change.
Prophesee Sourcing turns every promise and receipt into a distribution the plan can use, and prices a disruption the day it appears.
Supplier reliability is not improving. Not measurably.
Across 101,000 suppliers and two years, the published reliability index moved less than one point. A quarterly average is not a number a plan can use.
Sources: SourceDay, Supplier Reliability Index Report Q1 2025, Apr 2025 (101,000+ suppliers, 100m annual purchase order updates, $70bn direct materials spend); the publisher sells purchase order management into this problem · McKinsey, Supply Chain Risk Pulse, Dec 2025 (n=100 global supply chain companies) · Sphera, supply chain transparency study, Apr 2025 (n=250 chief procurement and supply chain officers) · 3RDi review of published detection benchmarks across the supplier risk category, Aug 2026.
From scoring suppliers to planning around them
The planning lead time was typed in at go live and never tested against a receipt. Every order date downstream inherits the error.
The measured lead time and its spread by supplier and part, from survival analysis on promise to receipt, refreshed as orders close, so the plan uses a distribution rather than a number.
The report lists what is late. It does not rank by what the late part stops, so the buyer works it from the top and the line stops anyway.
Every open order scored by the works order, the customer order and the revenue behind it. When a confirmed date moves, the consequence is calculated and routed to a named owner.
A second source is argued, not modelled. Nobody prices the service gained against the qualification effort, so the decision waits for a shortage.
Model the service, cost and qualification effort of a second source before you commit, watch the projected curve update, then track the actual against the modelled plan.
The list holds tier one. The part that stops the line is bought by your supplier from a company you have never contracted with.
Suppliers resolved into one record across ERP, procurement and quality, with the chain followed past tier one and a stated confidence on every inferred link.
Turning supplier challenges into decisions
The planning lead time was typed in at go live and never tested against a receipt. A quarterly average hides the spread the plan needs.
A newly qualified supplier has no delivery record, so the plan uses the lead time on the contract until reality corrects it.
The receipt is recorded and the promise it broke is not, so the gap between them never becomes a correction.
The list holds tier one, while most incidents originate two or three hops out where nothing is mapped.
A subscription tells you something happened near a supplier. Nothing tells you which parts, orders and revenue it stops.
14 AI applications that could be relevant
A sample of what becomes possible on the decision layer, not a fixed list: each application draws on the same data foundation and audit trail, and new ones are configured on the engines, not built from scratch.
The measured lead time and its spread, set against the number in the master.
The distribution of when a supplier actually delivers against the promise, by part and site.
A starting distribution borrowed from comparable suppliers, replaced by measurement as orders close.
What a supplier event does to your plan, estimated from precedent as a range with a confidence bound.
Candidate supplier events detected from feeds, documents and your own order telemetry.
A confirmed date moves and the consequence is calculated the same day.
A supplier trending down, flagged before it becomes a shortage.
The service and cost of a second source modelled before the qualification effort is committed.
Model the service, cost and qualification effort of adding a second source.
Move volume between suppliers in a model and watch service and cost update.
Which supplier event classes have earned automatic application, and what each one still needs.
Parts, plants, orders and revenue exposed to a supplier or region, with confidence on every link.
The same supplier reconciled across ERP, procurement and quality into one record.
The promise, the changes and the receipt assembled for the supplier review.
A day when the promise is data
Today: A scorecard produced quarterly from data nobody trusts.
Every part shows its measured lead time and spread against the master. Twelve are wrong enough to matter.
Today: A late delivery discovered when the goods do not arrive.
A confirmed date moves. The works order, the customer order and the revenue it stops are on the alert.
Today: A second source argued for in a meeting and never costed.
Two suppliers modelled on service, cost and qualification effort. She funds the one that moves the curve.
Today: A lead time in the system that has been wrong for a year.
Corrected lead times go back to the planning run. Next week's orders are placed against reality.
Yesterday's promise becomes today's plan.
Buy the lead time you were promised
We agree the metric and the baseline in week one, and measure the result on your data.