Procurement process improvement: how do you fix an existing process?
Short answer
Procurement process improvement starts with measurement: find where requests wait and where spend escapes the process, fix the largest cause first, then measure again.
Key takeaways
- Improve one part of the process at a time, from data: map what really happens, measure a baseline, find the cause, change one thing, then measure again.
- A few measures with published definitions are enough to start: requisition-to-order cycle time, maverick spend, retrospective orders, catalogue use and invoices blocked at matching.
- Most delays sit in waiting time between steps, such as approvals that add no information, not in the work itself. Cut steps only where segregation of duties is kept.
- Fix the cause, not the symptom: an invoice without a purchase order usually means ordering was too slow or too hard, not that requesters ignore the policy.
- An improvement holds only when someone owns the measure, reviews it every month and writes the new rule into the procurement policy.
What is procurement process improvement?
Procurement process improvement is the work of making an existing procurement process faster, cheaper or better controlled, based on measured facts. You map how requests, orders and invoices really move, measure where time and money are lost, fix the causes one by one, and check the result in the same measures.
A simple case. Requesters complain that ordering a laptop takes three weeks. Mapping the process shows that the request waits nine days for three approvals, though the laptop is on a catalogue at a contract price. Removing two approvals for catalogue items under a set amount cuts the wait, and the next month’s figures show whether it worked.
This page is about improving a process that already exists. The steps themselves, from need to payment, are described in the procurement process. Long-term choices about categories, suppliers and goals belong to a procurement strategy. How the function is organised and governed is covered by procurement management, and tools that automate steps are compared in procurement automation software.
How do you know the procurement process needs improving?
The process needs improving when people work around it. Each workaround is a sign of where it is too slow, too hard or unclear. The usual signs are easy to spot in daily work:
- Invoices arrive with no purchase order number, and a buyer raises the order afterwards so that the invoice can be paid.
- Requesters buy from a supplier they know rather than from the contracted one, because they cannot find the contract.
- The same item is bought at three different prices in the same month.
- Requests wait days for approvals that are always given.
- Buyers retype the same data from a request into the ERP, then into an email to the supplier.
- Invoices are blocked for small price differences that nobody had agreed to accept.
The procurement process guide lists the common challenges at each hand-off with their usual fix. This page shows how to measure which of them costs most in your organisation, and how to fix it so the fix holds.
What are the steps of a procurement process improvement?
A procurement process improvement follows the same cycle as any process improvement. Quality practitioners call it DMAIC: define, measure, analyse, improve and control. ASQ describes it as a data-driven strategy that improves a process by finding and removing defects. Applied to procurement, it gives seven steps.
- Choose one scope. One route, such as low-value indirect purchases, or one problem, such as invoices without orders. A whole-process project rarely finishes.
- Map what really happens. Follow ten recent purchases from request to payment, using system dates, not the policy. Note each wait, each re-keying and each approval.
- Measure a baseline. Calculate two or three measures over the last six or twelve months, with a written definition for each.
- Find the cause. Ask why the measure is poor until you reach something you can change: a missing catalogue, an approval rule, a field nobody fills in.
- Change one thing and test it on one site, one category or one team for a few weeks.
- Measure again with the same definition and the same data. Keep the change if the measure moved; undo it if not.
- Hold the gain. Write the new rule into the procurement policy, name an owner for the measure and review it every month.
Public bodies often have this duty in writing. The UK central government commercial standard, GovS 008, lists continuous improvement among its commercial management practices, alongside systems and management information.
Which measures show where the procurement process loses time or money?
Use measures that have a published definition, so you can explain them and compare your own results from one period to the next. The table gives the measures that best show where a procurement process loses time or money, with the source of each definition.
| Measure | Definition | Data needed | Definition from |
|---|---|---|---|
| Requisition-to-PO cycle time (days) | Days, weekends included, from receipt of the purchase requisition to release of the purchase order to the supplier | Requisition and PO dates | APQC benchmark measure |
| Procure-to-pay cycle time (days) | Average days for the full procure-to-pay cycle, measured separately for goods and for services | Requisition, PO, receipt, invoice and payment dates | APQC benchmark measure |
| Maverick spend (%) | Purchase value bought outside the process ÷ total purchase value; CIPS defines maverick spend as “unauthorised spend often with non-approved suppliers” | Spend by supplier and channel | APQC measure; CIPS glossary |
| Retrospective orders (%) | Orders whose purchase order date comes after the invoice date ÷ all orders invoiced | PO and invoice dates | City of York Council audit report |
| Catalogue use (%) | Purchase order lines placed with e-procurement catalogue suppliers ÷ all PO lines | PO lines by channel | APQC benchmark measure |
| POs approved electronically (%) | Purchase orders approved in the system ÷ all purchase orders | Approval records | APQC benchmark measure |
| Invoices blocked at matching (%) | Invoices blocked for payment because a price or quantity is outside tolerance ÷ invoices received | Invoice blocks and reasons | Matching tolerances in ERP documentation (SAP, Microsoft) |
Pick two or three measures linked to the scope you chose; tracking twenty measures at once usually means none of them is acted on.
Two notes on these measures. First, write down the denominator: a share of orders and a share of spend can tell different stories. Second, external benchmarks for these measures exist, from APQC for example, but they depend on each organisation’s definitions and are often available only to members. Use your own trend as the main reference.
The US federal government gives an example of a spend measure used to drive improvement. A 2019 OMB memorandum on category management, M-19-13, asked agencies to reduce “unaligned” spending. It sorted spend into tiers: tier 0 for purchases made in a decentralised way, up to tier 3 for government-wide contracts designated “Best In Class”. Spend in tiers 1 to 3 counted as spend under management.
Which fixes work for the most common problems?
Each common problem has a measure that reveals it and a fix that removes its cause. The table maps them, and the guides linked below explain each fix in depth.
| Problem | Measure that shows it | Fix that removes the cause |
|---|---|---|
| Slow approvals | Requisition-to-PO cycle time; waiting time per approval step | Approval levels by amount and risk, one approver for catalogue items under a limit, named deputies during absences |
| Invoices without a purchase order | Retrospective orders (%) | A faster ordering route for small purchases, then a No PO No Pay rule with a short list of exempt categories |
| Off-contract buying | Maverick spend (%) | Contracts loaded as catalogues; the request form proposes the contracted supplier first |
| Low catalogue use | Catalogue use (%) | Catalogues for the items bought most often, kept up to date; punchout to supplier websites where it fits |
| Re-keying between systems | Share of POs sent to suppliers manually | Requisition converted to PO in the system; orders sent by portal, email from the system or EDI |
| Invoices blocked at matching | Invoices blocked (%) and their reasons | Prices on the PO taken from the contract; receipts recorded on delivery; agreed tolerances |
| Wrong supplier data | Payments returned; duplicate suppliers | One supplier record with controlled changes |
Start with the problem that has the largest measured cost and the simplest fix; leave system changes for when rule changes are not enough.
Each fix has its own guide: approval rules in purchase requisitions, catalogues in punchout catalogues, small purchases in P-cards, tolerances in three-way matching and supplier records in supplier master data management.
One limit applies to every fix: do not remove a control to save time. Fewer approvals is fine when the remaining approval still separates the person who asks, the person who approves and the person who pays. US federal internal control standards call this segregation of duties: no one person controls all key parts of a transaction.
What does a procurement process improvement look like in practice?
A worked example, with illustrative figures that are not a benchmark. A public hospital group receives about 6,000 invoices a year for indirect purchases. Accounts payable complains that too many arrive without an order.
| Step | What the team found or did | Figure |
|---|---|---|
| Baseline | Orders raised after the invoice date, last 12 months | 1,500 of 6,000 invoices (25%) |
| Map | Ten cases followed: requesters order by phone because the requisition takes 6 days to approve | Median requisition-to-PO time: 6 days |
| Cause | Every request, even €50 of catering, needs three approvals | 3 approval steps for all amounts |
| Change | One approver for catalogue items under €1,000; P-cards for under €250; No PO No Pay announced three months ahead | Pilot on two sites, 8 weeks |
| Measure again | Same definition, pilot sites, next 3 months | Retrospective orders down from 25% to 12%; median time 2 days |
| Hold | Rule written into the policy; accounts payable reports the measure monthly to the procurement lead | Owner named |
The fix was not a reminder to follow the rules: it made the official route faster than the workaround, and only then enforced it.
The order matters. Announcing No PO No Pay before making ordering faster would have blocked invoices and angered suppliers, while requesters would still have had no quick way to order. This matches what public bodies report: the City of York Council, for example, tracks retrospective orders by department and keeps exemptions, assessed case by case, as part of its No PO No Pay approach.
How do you make procurement process improvements last?
An improvement lasts when the new way is the easiest way and someone checks it every month. Without that, the old workarounds return within a few months, often when the person who led the project moves on.
- One owner per measure, named in the policy, who reports it monthly and explains any change.
- The rule in writing: approval limits, exempt categories and tolerances in the procurement policy and the system, not only in a slide.
- The system enforces the rule where it can: approval workflow by amount, matching tolerances, a payment block on invoices without an order.
- A quarterly review of the measures with the procurement lead and finance, which chooses the next scope to improve.
- Feedback from requesters and suppliers, collected after each change, because a faster process on paper can still feel slower to the person using it.
Accounts payable measures complete the picture after the order. APQC tracks, among others, cost per invoice and the days from receipt of an invoice to payment, and reports that top performers spend far less per invoice than bottom performers. Improve ordering first: many invoice problems start with an order that came too late or with the wrong price.
Questions
What are examples of procurement process improvements?
Common examples are approval levels set by amount instead of one route for everything, contracts loaded as catalogues, P-cards for very small purchases, a No PO No Pay rule with clear exemptions, purchase orders created from approved requisitions without retyping, and agreed matching tolerances so that small differences do not block invoices.
Do you need new software to improve the procurement process?
Not to start. Many improvements are rule changes: fewer approval steps, a faster route for small purchases, catalogues for frequent items. Software helps when the rule is clear but people cannot follow it easily, or when the data to measure the process is missing. Change the rule first, then automate it.
How long does a procurement process improvement take?
A focused change on one route, such as approvals for low-value purchases, can be mapped, tested and measured in about three months, because you need a few weeks of data before and after. A change that needs new catalogues, supplier onboarding or a system change takes longer.
What is a good level of maverick spend?
There is no single public figure that applies to every organisation, because definitions differ: some count only spend with non-approved suppliers, others any purchase outside the process. Define it once, measure it on the same basis each quarter, and judge success by the trend in your own data.
What is the difference between process improvement and procurement transformation?
Process improvement fixes one part of an existing process at a time, measured before and after. Procurement transformation changes the operating model as a whole: the organisation of the function, its systems and its role in the business. A transformation usually contains many process improvements, run in sequence.
New guides on eProcurement Toolkit
One short email when a guide on this line is added or revised. Nothing else.
Coming soon: sign-up opens with the first email.