Blog · Procurement and supply chain · Procurement
Why category-level compliance figures hide the real finding, how a procurement team breaks off-contract spend down by site and requester from the purchase ledger, the three patterns that account for most of it, and the conversation each pattern needs.
A category compliance figure of 86 percent tells the CPO something is leaking and nothing about where. The purchase ledger knows which site raised each order and, in most systems, who. Broken down that way, the 14 percent of off-contract spend is rarely fourteen percent of everyone; it is a few people in a few places, and the reasons cluster. This guide sets out the breakdown and the three patterns.
Per category:
Off-contract spend by site, and by requester within site, ranked by value Share of off-contract spend from the top five requesters
Per requester:
Off-contract share = off-contract spend ÷ total spend by that requester in the category
Supplier and requester identifiers only.
Pattern one: the site the contract never covered. A plant acquired after the contract was signed, or a country the agreement excludes. Every purchase there is off-contract by definition. The conversation is with the supplier, to extend coverage, not with the site.
Pattern two: the requester who does not know. A new requester, or one in a function the category manager never briefed, buying from the supplier they used at their last company. Off-contract share near 100 percent, small volumes, several categories. The conversation is a briefing.
Pattern three: the requester with a reason. An experienced requester buying a specific item from a non-contracted supplier because the contracted one cannot supply it, or supplies it late. Off-contract share concentrated in one item or one supplier. The conversation is with the category manager about the contract's gap, and it is the most valuable of the three.
One category, one quarter, $430,000 off-contract.
| Site | Requester | Off-contract | Their category spend | Share | Pattern |
|---|---|---|---|---|---|
| Monterrey plant | all | $190,000 | $190,000 | 100% | 1: site not covered |
| Ohio plant | R-0412 | $96,000 | $104,000 | 92% | 2: new requester, unaware |
| Leeds plant | R-0288 | $71,000 | $310,000 | 23% | 3: one item, supplier lead time |
| Others | 31 requesters | $73,000 | Tail |
Three lines explain 83 percent of the leak, and none of them is a compliance failure by the person buying. One is a contract scope gap, one is a briefing, one is a supplier performance problem the category manager needs to raise. The category figure said 86 percent and nothing else.
Requester missing. Ledger exports without the requester cannot break down beyond site. Get the PO export and join.
Contract scope not carried. Pattern one is misread as maverick spend and the site is blamed. Carry the sites the contract covers.
Read as a league table. A ranked list of requesters by off-contract spend, circulated, teaches requesters to hide purchases. The list goes to the category manager with the pattern beside each line.
Card spend excluded. The tail is where pattern two lives. Include it.
Mapped once, the ledger and the contracts produce the breakdown per category every month, with the pattern suggested per line. Covirage builds this from the export as it is. The procurement page describes the setup, and the contract compliance guide covers the category-level roll-up this sits under.
In most procure-to-pay systems the requisition carries the requester and the purchase order carries the buyer. Where the ledger export lacks it, the PO export has it, and the join is on the PO number. Where neither has it, site and cost centre are the fallback.
Tail spend on cards and one-off orders is real maverick spend and usually unmeasured. Include card data where the company has it, mapped to the same categories, and report it as its own line with the same site and requester breakdown.
The list goes to the category manager and the site lead, not to a scoreboard. The three patterns are the point: most off-contract spend is a coverage or awareness problem, not a behaviour problem, and the report says which.