Source to pay: process stages and owners

2026-09-22 · Tommaso Maria Ricci

Most companies that talk about source to pay are actually running two separate procurement organizations without realizing it. One negotiates: it runs the sourcing events, builds the business case, signs the contract and reports a saving to the finance team. The other buys: it raises requisitions, chases approvals, receives goods, processes invoices and pays. The first one is measured on savings identified. The second one is measured on invoices processed. Neither is measured on whether the money the first one promised ever showed up in the second one's spend.

That gap has a name in the industry, and the name is the reason source to pay exists as a discipline. Source to pay is the end to end chain from deciding what to buy through to paying for it, treated as one process with one owner, instead of two processes that meet at a contract nobody reads after signature.

The pressure to close that gap is not theoretical. The Hackett Group's 2026 Procurement Key Issues study projects procurement workloads rising roughly 8% in 2026 while headcount and operating budgets decline. More work, fewer people, flat budgets. The only structural answer is a process where value stops leaking between the stages, because there is no spare capacity left to chase it manually.

This guide is not a software comparison. It is the operating model: the seven stages with a named owner for each, the five handoffs where negotiated value disappears, the metrics in the order that they actually cause each other, the seven failure modes I see most often, what to automate and what to leave alone, and a 90 day sequence to fix the worst of it.

What source to pay actually covers, and where it ends

Source to pay covers everything from identifying a need or a category opportunity through to the supplier being paid and the outcome being measured. In practice it is two well known processes joined at the hip.

Source to contract is the upstream half: category strategy, market analysis, sourcing events, negotiation, supplier selection, contract execution. The output is a contract and a set of commercial terms.

Procure to pay is the downstream half: requisition, approval, purchase order, receipt, invoice matching, payment. The output is a paid supplier and a recorded transaction.

Source to pay is the assertion that these are not two processes but one, and that the seam between them is where most of the money goes missing. I have written the two halves separately, because each has its own controls and its own failure modes: the upstream half in the guide on the strategic sourcing process, and the downstream half in the guide on procure to pay process controls. This piece is about what happens between them.

Three things sit just outside the boundary and get dragged in anyway.

Supplier lifecycle management is adjacent, not inside. Onboarding, qualification, performance review and offboarding run on their own cycle, longer than any single transaction, and they feed source to pay rather than sit inside it.

Accounts payable operations overlap the tail end. Payment execution, cash discount capture and banking controls belong to treasury and finance, even though the invoice arrives through the procurement chain.

Demand management sits upstream of everything. Whether the organization needs the thing at all is not a procurement decision, though procurement is usually the last function with a chance to ask the question.

Getting these boundaries wrong produces the two most common organizational mistakes: a procurement team held accountable for supplier performance it cannot influence, and a finance team held accountable for savings it never controlled.

The seven stages, each with a named owner

The reason source to pay programs stall is rarely the stage list. Every consultancy has one and they differ only in wording. The programs stall because nobody can name the person accountable for each stage, which means every handoff is a negotiation instead of a process.

Here is the sequence with the accountability that makes it work. Adjust the titles to your organization, but do not leave a row blank.

|---|---|---|---|

Two rows carry more weight than the rest.

Stage 4 is the one most organizations skip. A contract that is signed but not loaded into the buying channel does not exist, operationally. Requesters cannot find it, so they buy the way they bought before. The saving stays on the slide.

Stage 7 is the one most organizations fake. Value realization means comparing what was actually paid against what the contract said, on the actual volumes bought, and reporting the difference. Most companies report the negotiated rate multiplied by forecast volume, which is a projection presented as a result.

What a good stage definition contains

For each stage, three things must be written down and agreed, or the stage is decorative.

The entry condition. What must be true before this stage starts. For stage 2, it is an approved category strategy with a named budget owner, not a request from someone who wants a quote.

The decision right. Who can say no, and whose no is final. If everyone can escalate every decision, the process has no authority and reverts to relationships.

The exit artifact. The document or system record that proves the stage completed. Without it, status reporting becomes opinion, and every steering meeting spends its first twenty minutes establishing where things actually stand.

The five handoffs where value disappears

This is the part that distinguishes source to pay from its two halves run separately. The stages usually work. The joints usually do not.

Handoff 1: from negotiation to contract

The commercial terms agreed in the negotiation room do not all survive into the contract. Volume tiers get simplified, rebate mechanics get softened, service credits get negotiated away by legal in exchange for something unrelated, indexation clauses get accepted in a form nobody models.

The control is simple and rarely applied: the person who negotiated signs off on the final contract against a term sheet, clause by clause, before execution. Not the redline, the term sheet. If a term moved, it is flagged and re-priced.

Handoff 2: from contract to buying channel

The contract exists and nobody can buy from it. Prices are not loaded, the supplier is not enabled, the catalog does not reflect the new items, the punchout was never configured. Meanwhile the requester has a job to do and buys from the incumbent at the old price.

The measurable symptom is contract utilization, and it is the single most diagnostic number in this whole domain. If you negotiated a contract for a category and less than three quarters of that category's spend flows through it, the negotiation was largely theatre. The discipline of keeping contracts alive after signature is a process of its own, which I covered in the guide on contract lifecycle management.

Handoff 3: from requisition to purchase order

The requisition is approved but the purchase order carries different terms, a different supplier entity, a different payment term, or no contract reference at all. Every one of those breaks the link between what was negotiated and what will be paid.

The control is a hard one: no purchase order without a contract reference for categories that have a contract, and no manual override of payment terms at the purchase order level. The exceptions will be loud in month one and silent by month four.

Handoff 4: from receipt to invoice

Goods arrive and nobody records receipt, or receipt is recorded in quantity but not in quality, so the invoice matches on paper while the business is still waiting for something usable. For services, this is worse: nobody can say whether the work was delivered, so the invoice is approved on trust and the milestone was never verified.

The control for services is a delivery confirmation tied to a named milestone, not a general approval by the budget owner. Budget owners approve because they want the supplier to keep working, not because they checked.

Handoff 5: from payment to measurement

The invoice is paid and the transaction closes. Nobody compares the price paid against the contracted price. Nobody aggregates that comparison by category. Nobody reports the difference to the person who negotiated. This is the handoff that almost nobody builds, and it is the only one that tells you whether the previous four are working.

The control is a monthly price variance report by category, automatically generated, sent to the category manager and the finance business partner together. Not a project. A standing report.

Metrics, in the order that they cause each other

Most procurement scorecards list metrics alphabetically or by data availability. That obscures causation, and a metric you cannot act on is a metric that generates meetings instead of decisions.

Here is the causal order. Each one moves the one below it.

|---|---|---|---|

Read that table upward when something is wrong. A poor realized savings ratio is never fixed by renegotiating harder. It is fixed by finding which row above it is broken. In most companies the broken row is contract utilization, and the cause is stage 4.

One warning on savings reporting. If procurement and finance do not agree in advance on what counts as a saving, the number becomes a quarterly argument rather than a measurement. The agreement should be written, should specify the baseline method for each category type, and should be signed by the finance business partner before the sourcing event starts, not after the result is known.

Seven failure modes, ranked by cost

First and most expensive: savings that are reported but never appear. A negotiated rate applied to a forecast volume, booked as value, never checked against actual invoices. This single failure mode does more damage to procurement's credibility than every other item on this list combined, because it is eventually discovered by finance rather than disclosed by procurement.

Second: maverick buying that is actually a usability problem. When people buy outside the channel, the reflex is a policy reminder. Usually the channel is slower than the alternative, the catalog is missing what they need, or approval takes four days for a fifty dollar item. Compliance is a design problem before it is a discipline problem.

Third: approval chains built for risk that no longer exists. Five approvers on a routine order, with three of them approving without reading because they have no basis to object. Long chains do not add control, they distribute the appearance of it. Control comes from thresholds, channel design and after the fact sampling.

Fourth: category strategies written once and never revisited. Market conditions move, the supplier base consolidates, the internal demand profile changes, and the strategy on file still reflects a market that existed three years ago.

Fifth: supplier data that nobody owns. Duplicate vendor records, stale bank details, entities that merged years ago, contacts who left. This is both an efficiency problem and a fraud exposure, and it is invisible until an invoice goes to the wrong account. The discipline that prevents it belongs to the supplier program, which I covered in the guide on building a vendor management program.

Sixth: payment terms used as a negotiation chip without treasury in the room. Procurement trades longer payment terms for a lower price, treasury discovers it later, and the supplier prices the financing cost back into the next renewal. The net effect is frequently negative and nobody computes it.

Seventh: tail spend ignored because it is small per transaction. It is small per transaction and large in aggregate, and it consumes a disproportionate share of the team's transactional capacity. Ignoring it is a decision that should be made explicitly, with a strategy for it, not by default.

What to automate, and what to leave alone

The Hackett Group's March 2026 release on procurement's AI agenda reports 43% of organizations actively pursuing AI deployment in procurement, nearly double the previous year, but only 12% operating at large scale, with 69% accessing AI through capabilities already embedded in their existing procurement platforms. That last number is the practical one: for most teams the question is not which AI vendor to buy, it is whether to turn on what the current platform already ships.

Deloitte's 2025 Global Chief Procurement Officer Survey, published on 19 August 2025 and covering more than 250 chief procurement officers across 40 countries, puts the top barrier at siloed operations, cited by 57%, followed by competing priorities at 46% and capability gaps at 40%. Note what is not at the top of that list: technology. The constraint is organizational.

Automate these

Invoice matching and exception routing. High volume, rule based, measurable. This is where touchless rates come from, and where the capacity to do anything else comes from. The detailed mechanics are in the guide on automating the accounts payable process.

Spend classification. Mapping transactions to categories is tedious, repetitive and improves with volume. It is also the prerequisite for every other analysis, which is why teams that skip it end up debating data rather than decisions.

Contract term extraction. Pulling payment terms, renewal dates, price mechanics and notice periods out of signed documents into a structured field. This is the enabler for the price variance control described above, and doing it by hand is why nobody does it.

Supplier risk monitoring. Continuous screening against external signals beats an annual questionnaire, because the questionnaire measures what the supplier chose to disclose at one point in time.

Requisition guidance. Steering the requester to the contracted supplier and the right item at the moment of buying. This is the highest leverage automation available, because it fixes compliance at the point where compliance is decided.

Do not automate these

Award decisions on strategic categories. A model can rank offers. It cannot weigh a supplier relationship that has absorbed three emergencies, or judge whether a bidder's aggressive price reflects efficiency or desperation.

Exception approvals above a material threshold. If an exception is worth money, a person signs it. Automated exception handling turns a control into a formality.

Supplier offboarding. The decision to stop using a supplier has legal, operational and sometimes human consequences that do not reduce to a score.

First time supplier onboarding for critical categories. The point of qualification is judgment about risk. Automating the paperwork is fine, automating the judgment is how single source dependencies get created without anyone noticing.

Four maturity stages, and how to tell which one you are in

Maturity models are usually vanity. This one is diagnostic: find your row by the symptom, not by the description you prefer.

Stage 1, transactional. Procurement is measured on processing volume. Nobody can state total spend by category without a data project. Savings are claimed but not defined. The symptom: the answer to "how much do we spend with this supplier" takes more than a day.

Stage 2, controlled. Spend is visible, purchase order coverage is decent, contracts exist and are largely followed for major categories. Savings are defined but measured at negotiation, not at payment. The symptom: procurement and finance disagree about the savings number every quarter.

Stage 3, integrated. The two halves are one process with one owner. Contract utilization is measured, price variance to contract is a standing report, category strategies are refreshed on a cycle. The symptom that you have arrived: the realized savings ratio is reported and it is uncomfortable.

Stage 4, strategic. Procurement participates in demand shaping and product decisions before the need is defined. Supplier capability informs what the company builds. The symptom: procurement is in the room before the specification is written, not after.

Most mid sized companies are somewhere between stage 1 and stage 2 and believe they are at stage 3. The test is a single question: can you state, for last quarter, the difference between the price you contracted and the price you paid, by category? If not, you are at stage 2 at best.

Where source to pay meets the rest of the business

Boundaries matter because unclear ones produce work done twice and decisions made nowhere.

Against order to cash. The mirror image on the sell side, with the same structure and the same leakage pattern at the joints. The symmetry is instructive: what your accounts payable team does to your supplier's days sales outstanding, your customer's team does to yours. The full treatment is in the guide on the order to cash process.

Against financial planning. Procurement commits money the budget holder owns. If the two processes do not share a category taxonomy, every variance conversation starts with a reconciliation.

Against legal. Legal owns the risk terms, procurement owns the commercial terms, and the failure mode is one function silently trading the other's position to close a deal.

Against treasury. Payment terms and early payment discounts sit at the intersection. Neither function should be able to change them unilaterally.

Eighteen questions to ask your own team

Not vendor questions. These are for an internal working session, and the discomfort they produce is the point.

What percentage of last quarter's spend went through a contract we negotiated?
For our top ten categories, what is the difference between contracted price and paid price?
How many active supplier records do we have, and how many transacted in the last twelve months?
How long does a routine fifty dollar purchase take from need to delivery?
How many approvers touch that purchase, and what is each one actually checking?
Who signs off that the terms in the contract match the terms we negotiated?
When a contract is signed, how many days until someone can buy from it?
Who is accountable for loading prices into the buying channel, by name?
What share of invoices post without human intervention?
What is our most common invoice exception, and what causes it?
For services, who confirms delivery, and what do they look at?
Do procurement and finance use the same definition of a saving? Where is it written?
When did we last refresh our category strategy for our largest spend category?
What percentage of spend is tail spend, and what is our strategy for it?
Have we traded payment terms for price without treasury in the room?
Which single supplier could stop us shipping tomorrow, and what is the plan?
What AI capability is already included in our current platform that we have not switched on?
If the person who runs our largest category left next month, what would break?

Question 8 is the one that produces the most silence in the room, and it is the cheapest to fix.

A case from another industry, the same principle

The pattern shows up far outside procurement. A sports distribution company I worked with grew sales by about 30% using AI driven marketing. The interesting part is not the technology, because the first version of that work produced demand that largely evaporated. Campaigns generated qualified interest, and the handoff to the people who could convert it was informal, undocumented and dependent on whoever happened to be available that week.

The gain arrived when the handoff was made explicit: a named owner at each stage, a defined artifact that proved the handoff had happened, and one report that measured the gap between what was generated and what was converted. The generation work barely changed between the two versions. The leak closed, and the results followed.

Source to pay has exactly this shape. The negotiation is usually competent. The buying is usually functional. The value disappears at the seams, and the seams are where nobody is measured. Before buying a platform, it is worth asking how much of your realized savings gap is a joint problem rather than a capability problem, because the answer changes the size of the project by an order of magnitude.

A 90 day sequence

Days 1 to 30: measure the gap

Pick your three largest categories. For each, pull the contracted prices and the actual invoice lines for the last two quarters and compute the price variance. This is tedious and it is the entire point: the number you produce is the business case for everything that follows, and it is almost always worse than anyone expects.

At the same time, compute contract utilization for those three categories: what share of category spend flowed through the contracted supplier at contracted terms. Then write the savings definition with your finance business partner and get it signed, before anyone has an incentive to argue about it.

Days 31 to 60: fix the worst joint

Do not launch a transformation. Take the single handoff with the largest measured loss, which in most organizations is contract to buying channel, and fix it for those three categories only: prices loaded, suppliers enabled, requisition guidance pointing at the right item, purchase orders carrying a contract reference.

Stand up the monthly price variance report as an automated standing report, sent to the category manager and the finance partner together. Not a dashboard nobody opens. A report with two named recipients.

Days 61 to 90: prove it and extend

Re-measure the same three categories. Contract utilization and price variance should both have moved, and if they have not, the constraint is somewhere you did not look, which is worth more than a success story at this stage.

Then extend the same pattern to the next three categories rather than broadening the scope of the first three. The discipline of fixing one joint across the whole spend base beats fixing every joint for a narrow slice, because the first approach compounds and the second one stalls the moment attention moves elsewhere.

If you are looking at this gap in your own organization and want to work out whether the constraint is the process, the platform or the ownership model, describe how your last large negotiation ended up in the buying channel and I will tell you where I would look first. In most cases the answer is not what the company expected when it started shopping for software.

Three operating models, and what each one breaks

The stage list and the metrics assume someone is organized to run them. How you organize changes which failure modes you get, and choosing the model deliberately is cheaper than discovering it by accident.

Centralized. All sourcing and buying authority sits in one team. Compliance is high, leverage is high, and the business complains about speed. The characteristic failure is that procurement becomes a queue: requesters route around it for anything urgent, and the maverick spend shows up in exactly the categories where the business feels time pressure. If you run this model, the number to watch is cycle time for routine purchases, because that is what determines whether people comply.

Decentralized. Business units buy for themselves, with procurement offering support. Speed is high, business alignment is high, and leverage disappears. The characteristic failure is the same supplier charging three different prices to three units, none of whom know about the others. If you run this model, the number to watch is price dispersion for the same item across units, and it is usually larger than anyone believes.

Center led. A small central team owns category strategy, contracts and the buying channel, while execution sits with the business. This is where most mid sized companies should land, and it is the model source to pay assumes. The characteristic failure is ambiguity: the center believes it owns the decision, the unit believes it owns the decision, and both are partly right. The fix is not a better organization chart, it is the decision rights written per stage, as described earlier.

One additional consideration cuts across all three. Whatever the model, the person accountable for a category's realized savings should be the same person accountable for its supplier relationship. Splitting those two creates a predictable behavior: the relationship owner protects the supplier from the savings owner, and the negotiation stops being adversarial in the way it needs to be.

The capacity trap behind all three

Every one of these models runs into the same wall, which is why the workload projection at the top of this guide matters. When workload rises and headcount does not, teams protect the visible work: the large sourcing events, the escalations, the executive reporting. What gets dropped is the invisible work, and the invisible work is exactly stages 4 and 7, enablement and measurement.

That is the mechanism behind most savings leakage. It is not that anyone decided enablement was unimportant. It is that enablement has no deadline, no complainant and no meeting, so it loses every week to work that has all three. The only durable fix is to give those two stages an owner, an artifact and a recurring date, which converts them from good intentions into work that someone is measured on.

FAQ

What is the source to pay process, and what stages and owners does it include?

Source to pay is the end to end chain from deciding what to buy through to paying for it and measuring the result, treated as one process rather than two. It has seven stages, each with an accountable owner: spend analysis and category planning owned by the category manager, sourcing and negotiation owned by the category manager, contract execution owned by legal with procurement, supplier and catalog enablement owned by procurement operations, requisition and approval owned by the requester and budget owner, receipt and invoice matching owned by the receiver and accounts payable, and payment and value realization owned by treasury and the finance business partner. Stages four and seven are the ones organizations most often skip, and they are the ones that determine whether negotiated savings ever materialize.

What is the difference between source to pay and procure to pay?

Procure to pay is the downstream half only: requisition, approval, purchase order, receipt, invoice matching and payment. It starts when someone decides to buy something specific and ends when the supplier is paid. Source to pay includes all of that plus the upstream half, known as source to contract: category strategy, market analysis, sourcing events, negotiation and contract execution. The practical difference is accountability. A procure to pay program optimizes transaction efficiency. A source to pay program is accountable for whether the value negotiated upstream survives into the money actually spent downstream.

Why do negotiated savings not show up in the financial results?

Usually because of one of three breaks. The contract terms differ from what was negotiated, because the redline process traded something away without the negotiator re-pricing it. The contract was never loaded into the buying channel, so requesters kept buying from the incumbent at old prices. Or the saving was calculated as the negotiated rate multiplied by a forecast volume and never compared to actual invoice lines. The diagnostic is a price variance report: for each category, the difference between contracted price and price actually paid, on actual volumes. If nobody produces that report monthly, reported savings are a projection.

What metrics should a source to pay program be measured on?

Seven, in causal order, because each one moves the next: spend under management, contract utilization, catalog and channel adoption, purchase order coverage, touchless invoice rate, price variance to contract, and realized savings ratio. Realized savings ratio, meaning realized value divided by negotiated value, is the only one finance ultimately cares about, but it is never fixed directly. When it is poor, read the list upward to find which row is broken. In most organizations the broken row is contract utilization, caused by contracts that were signed but never enabled in the buying channel.

What should be automated in source to pay, and what should not?

Automate invoice matching and exception routing, spend classification, contract term extraction, supplier risk monitoring and requisition guidance at the point of buying. These are high volume, rule based or measurable, and they create the capacity to do everything else. Do not automate award decisions on strategic categories, exception approvals above a material threshold, supplier offboarding, or first time qualification for critical suppliers. The dividing line is judgment: automate the work that produces a recommendation, keep a person on the decision that carries consequences. Before buying anything new, check what the existing platform already includes, since most organizations access AI through embedded capabilities rather than separate tools.

How long does it take to fix a source to pay process?

Measuring the gap takes about 30 days for three categories: pull contracted prices against actual invoice lines and compute both price variance and contract utilization. Fixing the single worst handoff for those categories takes another 30 to 60 days. A full program across all categories typically runs 12 to 18 months, but that is the wrong way to start, because a broad program delays the first measurable result past the point where attention holds. Fix one joint across three categories, prove the movement in the numbers, then extend the same fix to the next three.

Who should own the source to pay process end to end?

One person, senior enough to hold both halves accountable, typically the procurement leader with a formal joint mandate from finance. The mandate matters more than the title: without finance co-owning the savings definition and the measurement, the realized savings ratio becomes a quarterly dispute rather than a shared number. The common failure is splitting ownership between a sourcing leader and an operations leader with no one accountable for the seam between them. That structure guarantees that every handoff problem is somebody else's, which is exactly why the value leaks there.