Procurement KPIs: 15 Metrics Worth Tracking, and How to Define Them

Most procurement dashboards have two problems. They track too many things, and they define none of them.

Forty metrics on a screen means nobody knows which one matters. And a metric without a written definition means two people can report different numbers for the same thing and both be right. Both problems make the dashboard easy to argue with and hard to act on.

So this guide keeps to fifteen, in five groups. For each: what it is in plain English, how to calculate it, what good looks like where there is a published benchmark worth trusting, and the trap to avoid.

#Before you measure anything: three rules

Rule 1: Define it in writing first

Take something that sounds simple: did the supplier deliver on time? McKinsey studied this for the consumer goods industry and found there is no standard definition. Does on time mean the date you asked for, or the date they promised? Within a delivery slot, or any time that day? Is "in full" measured by order, by line, or by case? Different answers produce very different scores from the same deliveries.

Write every definition down, agree it with whoever you report to, and do not change it without saying so.

Rule 2: Agree the baseline with finance before you start

Savings figures cause more arguments than any other procurement number. The Procurist, a practitioner newsletter, puts the problem plainly: "The real challenge is agreeing the baseline upfront with Finance. Otherwise, procurement says it protected millions while Finance simply sees that the budget was respected."

Rule 3: Pair every speed metric with a quality metric

APQC, the benchmarking organisation, makes this point about purchase orders per person: since those measures encourage faster throughput, "it is important to have counterbalancing quality measures to ensure that the drive for speed does not result in increased errors". The same is true of every speed metric below.

#Group 1: Money saved

1. Spend under management

What it is: the share of total company spending that procurement actually influences, through a contract, a preferred supplier or a sourcing process.

Formula: managed spend ÷ total addressable spend × 100.

Benchmark: Ardent Partners' 2025 research puts the average at about 71% — the first time it has passed 70% in two decades of their tracking. Each extra dollar brought under management produced 6% to 12% savings in the first contract period.

The trap: defining "managed" generously. A supplier someone once emailed is not managed. Write down exactly what counts.

2. Realised savings, not just negotiated savings

What it is: savings that actually showed up in what you paid, as opposed to savings agreed at the negotiating table.

Formula: (baseline unit price − actual unit price paid) × actual volume bought, summed across contracts.

Why the distinction matters: McKinsey's June 2025 research, covering more than 340,000 transformation initiatives, found the average procurement savings pipeline loses one-third of its value during planning and another 20% during execution. Negotiated savings is the promise. Realised savings is what arrived.

The trap: reporting negotiated savings and calling them savings. Track both, and report the gap.

3. Cost avoidance

What it is: money that would have been spent but was not — a price rise blocked, an unnecessary purchase stopped, a better specification found.

Formula: (price or cost that would have applied − price or cost that did apply) × volume, against a baseline agreed in advance.

Why it matters: The Procurist argues that "cost reduction has a ceiling" while cost avoidance has none, because you are catching bad spend before it is committed. It is often more valuable than savings.

The trap: this is the easiest metric to inflate, so finance distrusts it by default. Report it separately from savings, never added to it, and only against a baseline finance agreed beforehand.

4. Off-contract spend

What it is: spend that went outside agreed contracts or preferred suppliers. Sometimes called maverick spend.

Formula: spend not on contract or with a preferred supplier ÷ total spend in the categories you have contracted × 100.

Benchmark: we do not quote a single percentage, because published figures vary enormously by industry and definition. What is well evidenced is the impact: the Hackett Group's leading procurement teams lose 60% less of their negotiated savings to off-contract buying and non-compliance than their peers. APQC lists the share of purchases made through maverick buying among its most downloaded procurement measures.

The trap: treating it as a discipline problem. High off-contract spend usually means the proper route is too slow or the contract is wrong for what people need.

#Group 2: Money kept

5. Payment terms achieved

What it is: how long, on average, you take to pay suppliers, and whether that matches the terms you agreed. At company level this is called days payable outstanding, or DPO.

Formula (company level): accounts payable ÷ cost of goods sold × number of days in the period. For procurement's purposes, also track the share of spend on your standard terms.

Benchmark: The Hackett Group's 2025 working capital survey of the top 1,000 US listed non-financial companies put DPO at 59 days. The same study found $1.7 trillion tied up unnecessarily in working capital — 35% of the total.

The trap: longer is not automatically better. Stretching payment terms on small or fragile suppliers can push them into trouble, which becomes your supply problem. APQC raises exactly this concern about paying smaller suppliers more slowly.

#Group 3: Speed

6. Requisition-to-purchase-order time

What it is: how long it takes from someone asking to buy something to a purchase order going to the supplier. This is the number the rest of the business actually feels.

Formula: median time from requisition submitted to purchase order issued. Use the median, not the average, so a few stuck requests do not distort it.

Benchmark: Procurify's data across more than 250 mid-market organisations shows about 58 hours for companies of 100 to 500 people and 65 hours for 500 to 2,000. Procurify's guidance is that beyond 72 hours, the waiting itself becomes the bottleneck. The Hackett Group's leading teams run this step 58% faster than their peers.

The trap: measuring only approved requests. Include the ones that were rejected, sent back or abandoned.

7. Invoice processing time

What it is: how long from receiving an invoice to paying or approving it for payment.

Benchmark: Ardent Partners' 2025 research puts the average at 9.2 days. The best 20% of companies take 3.1 days; everyone else takes 17.4.

The trap: faster is not always the aim. The goal is predictable processing, so you can pay on your agreed terms and capture early payment discounts where they are worth having.

8. Sourcing cycle time

What it is: how long a sourcing project takes, from agreeing the need to signing the contract.

Benchmark: the Hackett Group's leading teams run sourcing cycles 24% shorter than their peers. There is no reliable absolute benchmark in days, because a stationery tender and a manufacturing contract are not comparable.

The trap: comparing across categories. Track it by category type, and compare each against its own history.

#Group 4: Efficiency

9. Cost per invoice

What it is: the full cost of processing one supplier invoice — people, systems and overhead — not just the software.

Formula: total cost of the invoice-processing activity ÷ number of invoices processed in the period.

Benchmark: Ardent Partners' 2025 research puts the average at $9.40, with the best 20% at $2.78 and everyone else at $12.88. APQC's data, from 1,485 organisations and reported by CFO.com in 2018, gives $2.07 or less for the top quarter, $5.83 at the median and $10 or more for the bottom quarter.

The trap: the two studies define cost slightly differently, so compare yourself against one, not a mix of both.

10. Touchless rate

What it is: the share of invoices processed with no human intervention at all — they match the order and receipt automatically and go straight through.

Benchmark: Ardent Partners puts the average at 32.6%, and 49.2% for the best 20% of companies.

The trap: a high touchless rate with a high error rate is worse than a lower touchless rate done right. Pair it with an exception or error measure.

11. Cost of the procurement function

What it is: what it costs to run procurement, as a share of the spend it manages.

Formula: total procurement operating cost ÷ total spend × 100. APQC also normalises per $1,000 of revenue or per $1 billion of purchases.

Benchmark: the Hackett Group's leading procurement teams run at 19% lower cost as a share of spend than their peers, with 31% fewer staff, while delivering 2.03 times the savings.

The trap: cutting this number by cutting capability. The leaders are cheaper because they automated and redesigned, not because they did less.

12. Catalogue share of routine orders

What it is: the share of routine, low-value orders placed through an approved catalogue rather than raised by hand.

Why it matters: catalogue orders carry the approved supplier and agreed price automatically, so they are fast and compliant at once. Hackett's advisers noted on their podcast that leading teams put 81% more indirect spend through electronic catalogues. Procurify found catalogue adoption of 73.8% among smaller mid-market companies.

The trap: a catalogue full of outdated prices. Measure catalogue accuracy alongside catalogue usage.

#Group 5: Suppliers and risk

13. Supplier on-time, in-full delivery (OTIF)

What it is: the share of deliveries that arrived on the agreed date with the full agreed quantity.

Formula: deliveries both on time and complete ÷ total deliveries × 100 — using a definition you have written down, for the reasons in Rule 1.

Benchmark: requirements vary by industry. As a reference point, Walmart has required 98% on-time, in-full delivery from suppliers since September 2020, with a fine of 3% of the cost of goods on shipments that miss.

The trap: penalties change behaviour in odd ways. McKinsey found about 25% of consumer goods deliveries arrived more than two hours early for their slot — suppliers avoiding late penalties by creating a different problem at the unloading dock.

14. Supplier concentration

What it is: how much of your spend sits with your largest few suppliers.

Formula: spend with top five suppliers ÷ total spend × 100.

Benchmark: Procurify found 57% to 59% across mid-market segments and suggests 55% to 65% as a healthy range.

The trap: this is a spend view, not a risk view. A small supplier that is your only source for a critical part matters more than the percentage.

15. Critical suppliers with a current risk check

What it is: of the suppliers you have identified as critical, the share whose financial health and continuity plans have been checked in the last twelve months.

Why it matters: RapidRatings found only 15% of enterprises fully use supplier financial health data in setting payment terms, and 30% do not use it at all. Meanwhile 82% had a material supplier disruption in the previous year.

The trap: counting a questionnaire returned as a risk checked. The check should involve evidence, not just the supplier's own answers.

#The dashboard on one page

#The number that is not on the list

There is one more measure worth having, and it is the simplest: ask the business what it thinks. In Deloitte's 2025 survey of more than 250 chief procurement officers, 84% of the leading organisations met or beat their plan for internal stakeholder satisfaction. Among the rest it was 59%.

Joseph Richardson, a long-time procurement leader, made the point on the Art of Procurement podcast that value "is a relative term" — what matters to one stakeholder may not matter to another. A short survey twice a year, asking the people who use procurement whether it helps them, tells you things no system metric can.

#How to report it

1. Monthly, internally: the speed and efficiency metrics, so problems get fixed quickly.
2. Quarterly, to leadership: one page covering all five groups, with the trend, not just the latest number.
3. Once a year: re-check every definition and baseline with finance, and retire any metric nobody acted on.

That last step matters. A metric nobody acted on for a year is not a metric. It is decoration.