Single-source risk flagging

The job is to stop an award that would leave a category, a spec, or a plant-critical part sitting on one supplier or one producing site, and to say so while the event is still open.

A flag belongs on the award file: spend and critical volume on this winner if you sign, whether that volume also sits in one region or one plant, and whether a second qualified producing site exists. Dual-source names a second plant that can already make the spec, or it names the gap. It is not a savings claim.

This sits in select, before the PO. After award, supplier risk continuous monitoring watches the supplier you already chose. That is a compensating control when single-sourcing is deliberate. It does not undo a concentration you just created.

Spend and sourcing suites in the SAP Ariba, Coupa, and Jaggaer class already hold the POs, invoices, and the event. Supply-chain risk tools in the Riskmethods class hold site and region maps when you have them. The model is a join across those records plus the pending award, not a second buying channel.

Compute two numbers on the live event, then read them against criticality. They often disagree.

Spend share. Trailing paid spend to this supplier, or this legal-entity group, in the category, plus the volume you are about to award, divided by category spend over the same window plus that pending volume. Historical share alone misses the award that turns a large incumbent into the book. Event share alone misses the rest of the spend already with them.

Do not use a single global cutoff. A modest share of office supplies is noise. The same share of a qualified aerospace forging is a problem. Set the threshold with the business that owns the disruption cost, not inside a procurement dashboard.

Site and region share. Map volume to the producing site, not to the billing entity. Two vendor-master numbers that ship from the same mill, industrial park, or floodplain are one source. If you cannot name the plant, you do not have geographic diversification. You have two invoices.

Criticality is the third axis. A small-spend supplier that is the only qualified plant for a spec is more concentrated than a large-spend commodity vendor with five plants. Pull it from the item or category record you already use for business-continuity planning: unique spec, long qualification, no substitute grade, or a part that stops a line. If that record is empty, say so on the flag. Do not invent a criticality score from spend alone.

Include committed volume that is not yet invoiced: open POs, blanket balances, consignment. Exclude one-off tooling or NRE that will not recur. Category trees that dump unlike items into one node will manufacture fake concentration or hide the real kind. Fix the node before you trust the share.

When spend share or site share crosses the threshold you published, the output is not "diversify." It is one of three statements:

1. A second producing site is already qualified for this spec, at this quality bar, in this region constraint. Name the plant, the last qualification date, and the capacity you can actually use.
2. A candidate plant exists in discovery or on the vendor master, but it is not qualified. Name what qualification would take (audit, PPAP, first-article, ESG file). Do not award it on this event.
3. No second plant is known. Say that. Do not fill the gap with a logo from AI supplier discovery that has not survived a spec check.

Discovery is how you look, not proof that a second source exists. Run it against the spec, the geography you will accept, and the quality bar. If the longlist is empty or full of traders, the honest flag is "single-source, no qualified alternate."

The common false dual-source is a distributor or converter that buys from the same mill as the incumbent. Their DUNS is different. Their invoice is different. The melt, the resin, or the wafer is the same. Until you can name two producing sites that do not share that upstream plant, you are still single-sourced. Riskmethods-class maps help when you have tier data. Vendor-master names do not.

If ESG supplier screening flags the only qualified plant, that is a shortage risk, not a dual-source opportunity. Dropping that plant because a screen went red, without a second qualified site, leaves you with zero sources. Hold the award, open the evidence, and give the supplier a path to correct the record. Do not move volume onto an unqualified name to clean up a dashboard.

This walkthrough is illustrative, not a case study.

A category manager is awarding a two-year resin conversion contract for a food-contact grade used on three packaging lines. The preferred bidder, Northvale Polymers, holds most of the book. The event would take them from a large incumbent share to nearly the whole category. Their producing site is a mill in the same river basin that flooded two winters ago.

The model flags supplier concentration (pending award on top of trailing spend) and site concentration (one mill, one basin). It also lists Harbor Conversion Co. on the vendor master, a converter the plants use for overflow. Harbor invoices from a different legal entity. The warehouse is in another state. The award pack almost called that dual-source.

Plant quality had Harbor's incoming COAs. The resin pellets are bought from Northvale's mill. Harbor does not melt. They bag and ship. Two logos, one producing site.

Discovery returned three more converters. Two also buy from Northvale. The third melts a near-grade in another region, but has never run food-contact for this company: no audit, no first-article, no line trial. The honest recommendation is not "split the award with Harbor." It is: concentration is real; Harbor is not a second plant; the only other melt is unqualified; do not award the whole book to Northvale without a named qualification plan, and do not move this week's volume onto Harbor.

They almost parked Northvale on an ESG media hit about the river basin and invited an unqualified trader as the ESG option. That would have removed the only qualified melt.

Qualification of a second plant takes audits, samples, line trials, and often a customer or regulator notification. None of that happens between a Tuesday flag and a Friday award. Splitting this week puts unqualified material on the line, or onto a distributor on the same mill.

When a second plant is already qualified, a split is a commercial and operations decision: capacity, MOQs, tooling, freight, and whether the spec is identical. Put concentration into the multi-criteria decision matrix as a scored dimension with the weights the committee already signed. Do not let a sidecar model invent a split ratio.

When no second plant is qualified, the award file should say single-source, with the site named, the compensating controls (safety stock, monitoring, an open qualification), and who owns the disruption cost. Competitive tension is a different job. Competitive pressure analysis can still use should-cost and peer quotes to keep a single source honest on price. That is not dual-source.

Do not publish a dual-source savings figure from the flag. Recommending a split has not saved anything, and two setups can cost more. The quality outcome is that concentration was visible before award, and that any dual-source sentence named a real plant or named the gap.

Run the join in shadow on one category first: a process input with a known mill, or a qualified part with a long PPAP. Show the category owner spend share, site share, criticality, and the three-way recommendation (qualified second plant, candidate not qualified, none found). Count how often the "second source" was a trader on the same melt. If that rate is high, stop recommending dual-source until the plant map is trustworthy.

Coupa, SAP Ariba, and Jaggaer can hold the event and the spend. Riskmethods-class tools can hold site maps. Neither will refuse the award. The committee still signs. The model's job is to put concentration on that file in time to change the award, or to record that you accepted it with eyes open.