Category opportunity identification

The job is a rolling list of named opportunities the category owner can accept or kill: consolidate, renegotiate, or switch. Each row names the cluster, cites evidence from your posted spend and contracts, and proposes one action. An opportunity is not booked savings.

Finance will ask for a pipeline total. Do not give them one. Nothing is saved until the owner accepts the row, the work happens, and invoices change. Booking identified opportunities as cost reduction is how procurement spends the year defending a number that never reached the P&L.

Refresh the list on a cadence the team can actually work. Waiting for the annual category review is how drift sits for months. An unowned queue is a longer to-do list, not a better analysis.

Group transactions that are the same buy: spec, grade, unit of measure, pack size, and a comparable ship-to or Incoterm. Category codes route the row to the right owner. They are not the cluster key.

Mixing unlike SKUs into one opportunity is the usual false finding. A "packaging" cluster that blends RSC cartons, folding carton, and stretch film will show a tidy supplier count and a large total. It is not a consolidation event. Prefer a smaller, spec-tight cluster. When in doubt, split and let the owner merge.

Fragmentation only means something inside a tight cluster. Same 32 ECT RSC, four converters, three plants: that scatter is a candidate. Same family code, mill board, a spot trader, and a display-shipper converter: that scatter is a taxonomy problem.

Demand aggregation signal clusters live requisitions inside a hold window. This job clusters posted spend so you can change the next contract, catalog, or panel. Recurring intake clusters that never graduate belong here as a consolidate row.

If a usable catalog or contracted channel exists and people still buy around it, maverick spend detection is the sibling view. Do not write a switch for leakage that is actually a punchout nobody uses.

Once the cluster is the same buy, score three signals and keep them separate.

Price drift versus contract. Compare invoiced unit price to the contracted rate that should have applied, same spec, Incoterm, and volume band. Strip expedites, samples, and credits. Persistent drift on a covered item is a renegotiate candidate. A one-off premium is not. No contract means uncovered spend, not drift. Spend analytics and savings tracking is where realized price versus contract belongs after a change. Do not copy a drift gap into that savings file as if the conversation already happened.

Fragmentation. Count suppliers and volume share inside the cluster, by legal entity and site. Many suppliers on the same spec, none with a volume band, is a consolidate candidate. One qualified mill plus a trader on the same commodity code is not a fragmented panel. It is a qualification mix.

Market movement. Use a named series or a last-comparable-award cohort you can open. Market price benchmarking is the should-cost stack: index, last POs, freight. If you cannot name the series, geography, and date, you do not have a market signal. Do not invent a packaging index from an unrelated PPI line.

Then propose one action per cluster. Three verbs on one row is a meeting agenda. The owner cannot accept that.

• Consolidate when the spec is the same, too many suppliers or sites buy separately, and a volume band, blanket, or catalog would hold.

• Renegotiate when coverage is right, price has drifted versus the contract formula or a named series, and you still intend to keep the incumbent. Time it to notice periods from contract renewal alerting.

• Switch when a qualified alternate exists and the incumbent failed on price or performance against the spec you buy. Food-contact, mill qualification, capacity, and lane still have to hold. A cheaper unmatched quote is not a switch.

Do not auto-email suppliers from this list.

This walkthrough is illustrative, not a case study. It claims no measured savings and no capture rate.

A category manager owns corrugated converting for a food manufacturer with three plants. Posted spend, clustered on spec rather than on the word "packaging," shows two honest shapes and two false ones.

Same-buy RSC. 32 ECT, C-flute, two-color print, DAP, same board grade, all three plants. Four converters appear, plus mill-direct board on a formula contract. Two converters already hold food-contact print qualification. The other two show up on overflow and on one plant's preference. Proposed action: consolidate repeating RSC volume onto the two qualified converters with a volume band. Evidence is the spec-matched spend split by plant and supplier, not a company-wide vendor count.

Mill-direct linerboard. The formula contract is in force. Invoices have drifted above the formula while the named kraftliner series the formula points at has not moved in a way that explains the invoices. Proposed action: renegotiate the formula application, timed to the notice period. Evidence is invoice versus contract, with the series named. It is not a switch.

What they almost did. They mixed folding carton and RSC into one packaging consolidation because both sit under the same family code. Unlike SKUs: split or kill. They also proposed a switch from the qualified mill to a trader whose spot delivered price looked cheaper on a few loads. The mill is on the food-contact file. The trader is a spot desk with no mill qualification. Cheaper on a week is not a qualified alternate. They then loaded every row into the quarterly savings pipeline. Nothing had been accepted. Nothing had invoiced differently. That is booking the pipeline.

The owner accepts the RSC consolidate as a sourcing event, accepts the linerboard row as a commercial conversation with a date, kills the mixed packaging cluster, and kills the trader switch with a reason so it does not return next month.

Every live row needs a named owner and a decision. Unowned rows expire. Do not auto-accept, and do not keep a silent backlog that finance can screenshot as a target.

• Accept. The row becomes work: an RFQ, a commercial meeting, a catalog or blanket change. Write the work, the date, and who owns it. Savings tracking starts only after invoices reflect the change.

• Kill. Write the reason: unlike items, qualification gap, coverage hole rather than a panel, market signal missing, or no capacity this quarter. A kill with no reason is how the same cluster resurfaces.

• Split. The cluster mixed unlike SKUs or mixed a qualified mill with a trader. Split, then score each piece again. Do not average them into one verb.

Match the queue to the events the team can actually run. When an accepted consolidate keeps repeating at intake, send it back to demand aggregation so the next requisitions hit the catalog. When an accepted renegotiate approaches expiry, join it to the renewal brief.

Success is a short list the owner still trusts, with reasons you can audit. Failure is a board pack of identified opportunities, a mill swapped for a trader, and a savings figure that cannot be found in the ledger.

Sievo, SpendHQ, Coupa, and SAP Ariba are the class of spend analytics and source-to-pay platforms where transactions, contracts, and category cubes already live. Use them as the system of record for what you spent and what coverage you had. Opportunity identification is a scoring and decision layer on those cubes, not a second gold file and not a substitute for a category owner.

Do not rank those tools here. None of them will stop you from mixing unlike SKUs, treating a trader print as a mill switch, or booking the pipeline, if you let the export become the savings report.

Keep accept and kill inside procurement. A slide of ranked opportunities with estimated value and no owner decision is a savings claim by another name.