How do you build a credible procurement savings pipeline?

Most pipelines are built the wrong way round. A target is set, usually as a percentage of addressable spend, and the pipeline is populated until it reaches the target. Opportunities are sized optimistically because the alternative is presenting a plan that visibly does not add up.

The result is arithmetically complete and commercially fictional. It survives until the first serious forecast challenge, which is whenever finance first reconciles claimed benefit against actual budget movement. In organisations where nobody does that reconciliation, it survives indefinitely, which is worse.

A pipeline that reaches the target on the day it is written was written to reach the target.

Building forwards produces something more useful than a number: a ranked list of what to do next, with the evidence behind each line and the name of the person who has to act.

Three things carry almost all of the analysis, and none of them needs a technology project.

1. 01Spend, cleaned to a usable levelTwelve months of accounts payable data with suppliers grouped to parent level and mapped to a category structure the organisation recognises. It does not need to be perfect. It needs the largest suppliers and the largest categories to be right. The hard parts are supplier name variants, a single ledger line covering several categories, and the spend that never touches a purchase order at all: cards, expenses and direct debits.
2. 02Contracts, expiry and noticeWhat is under contract, when it ends, what the notice period is, whether it renews automatically and who owns it. Expiry alone is misleading. A contract ending in March with a rolling extension and a six month notice window stopped being an opportunity the previous September, and a pipeline sequenced on expiry dates without notice periods will keep discovering that too late.
3. 03Demand and specification contextWhat is actually being bought and why. This comes from conversations rather than data, and the productive ones are with the people who raise the requisitions rather than the people who sign them off. Ask what the requirement would look like if they were specifying it today, and what they would drop if the budget fell.

Opportunities come from a small number of levers. Naming the lever on every line is what makes the plan honest about how hard it will be, and it stops the same spend being counted twice under two headings.

Price

Competition or renegotiation on the same requirement. Fastest to deliver, easiest to validate, and the first to run out.

Specification

Buying something different: a lower grade where the higher grade was never needed, a standard product instead of a bespoke one, a shorter service window. Larger than price in most categories, and it needs the budget holder to agree in writing before you go to market.

Demand

Buying less. Consumption controls, approval thresholds, duplicate services removed. Often the largest lever available and the one that depends most on support outside procurement, which is why it converts least.

Process

Reducing the cost of buying rather than the cost of what is bought: catalogue coverage, invoice consolidation, rework removed. Real, and it usually shows as capacity rather than cash, so report it as capacity.

Commercial model

Changing how the contract works: the indexation clause and its cap, volume bands, risk transfer, incentives, term. Slow, and it is the lever that determines whether the saving is still there at the end of the term.

Telling people to agree definitions with finance is the easy half. The half that decides whether it works is the record itself, because that is what finance is actually being asked to sign. A line that cannot answer these questions is not ready to be counted.

• The category, the supplier and the lever.
• The cost centres and budget lines affected. Without these nobody can check whether anything happened.
• The baseline: the figure, the method used to derive it, and the period it covers.
• Forecast annual benefit, and the month it starts.
• Recurrent or non-recurrent, and cash-releasing or not. Two flags, and the two that finance reads first.
• The procurement owner, the budget holder, and the finance contact who will validate it.
• The current stage, the date it entered that stage, and a link to the evidence.
• The implementation date, which is a different date from the award date and is the one that matters.

The sign-off artefact is usually a single page naming the budget line, the amount, the month the budget reduces, and a confirmation from finance, countersigned by the budget holder. Without something like it, saying that finance has confirmed a benefit describes a sentiment rather than a control. What counts as a saving and what counts as avoidance has to be settled before any of this is filled in.

A pipeline needs stages, and each stage needs an entry test somebody can fail. Without the test everything sits at the same optimistic stage until delivery.

1. 01IdentifiedA category, a lever and a rough size. No baseline yet. Weight it lightly and expect most of it not to convert.
2. 02ValidatedBaseline agreed, size estimated on evidence, procurement owner named, budget holder identified and willing. If nobody outside procurement will own the change, it does not pass this gate. Specification and demand lines fail here more than anywhere else.
3. 03In deliveryApproach agreed, activity under way, a date attached. Sizing becomes a range rather than a single figure.
4. 04AgreedContract awarded, price agreed or specification changed. The benefit is known and it is not yet anywhere.
5. 05RealisedFinance has confirmed the benefit against the agreed baseline and the budget has moved. This is the only stage that should appear externally as a saving.

Weighting the early stages makes the forecast usable, and the weightings themselves are worth arguing about rather than asserting, because they determine whether the forecast can be built into a budget. Set them, then review them against what actually converted, so that the second year's weighting is evidence rather than an opinion carried forward. The useful test of whether the gates are real is simple: has anything been rejected at one recently.

The gap between an agreed benefit and a lower cost base is where most pipelines lose their credibility, and it is almost always the same handful of mechanisms.

• The new price file was never loaded, or loaded against part of the catalogue, so a share of transactions still charges the old rate.
• Requisitioners keep raising free text orders against the old supplier because the new one is not set up.
• The discount was priced on a volume band the organisation never reaches.
• A rebate is contractually due and nobody owns the claim, so it is not invoiced.
• The specification change was agreed in a meeting and never written into the contract or the ordering system.
• The budget was never reduced, so the money was spent on something else and the saving is invisible in the accounts.
• The indexation clause quietly returns the price reduction over the term, which is why the commercial model lever matters more than it looks.

Tracking implementation as a separate step from agreement is what closes this, and a look back a year later is the only test of whether the cost base actually fell and stayed down. It is rarely done, and it is the difference between a function that reports benefit and one that delivers it.