In short: A competitive sourcing event has a fixed cost of roughly twenty hours of buyer time, which sets a break-even category spend below which running one destroys value even when it works, and most of the tail sits below that line. The way to move the line is to change the unit of work through aggregation, catalogues and framework rates rather than to push buyers at more events. The tail's real exposure is the small number of suppliers whose loss stops production, which spend rankings cannot see because the join from supplier to bill of materials lives outside the accounts payable file. Verifying tail savings is genuinely hard, and process cost only falls when a headcount or a system contract actually leaves.
The spend cube for the board pack comes back with 11,400 suppliers paid in the last twelve months. Of those, 8,900 took less than 25,000 each. Together they account for around six per cent of the 840 million the business spent, and they hold no contracts, no scorecards and no named owner. The category managers have all committed their savings targets against the other ninety-four per cent, which is a rational allocation of their time, and every year the tail gets a slide and no plan.
The next thing that usually happens is a directive to reduce supplier count. It fails for a reason worth understanding, because the buyers ignoring the tail have been doing correct arithmetic.
The break-even that keeps the tail unsourced
A competitive event on a small category is not a cheap thing. Scoping the requirement, scanning the market, writing and issuing the request, chasing responses, normalising quotes, running the evaluation, getting the award signed and putting a contract in place lands somewhere around twenty hours of buyer time end to end. At a loaded sixty an hour that is 1,200 before you count the requester's time in specification and evaluation, and before onboarding a supplier who has never been paid by you.
Set that against what an event returns. Assume a well-run first-time competition on an unmanaged category delivers seven per cent, and assume the agreement holds for two years. The break-even annual spend is 1,200 divided by 0.07 times two, which is about 8,600.
Now compare that with the tail. Those 8,900 suppliers average around 5,700 each. The median is far below the mean, because tail distributions always have a long tail of their own. On the arithmetic above, the majority of the tail is spend where a competitive event costs more than it can possibly return, even assuming it works perfectly and the saving is real.
That is the finding to take into the room. The buyers who skip the tail are declining a negative-return activity, and any programme that asks them to run more events on smaller categories is asking them to do worse work.
Two numbers in that break-even are worth interrogating on your own data rather than accepting mine. The hours per event, which you can get from a handful of buyers with a stopwatch and honest recall, and the contract term, which changes the answer more than anything else because it multiplies the return without changing the cost. A five-year framework on a stable category has a break-even under 3,500.
What the tail actually costs to run
The price paid is the smaller half of the tail's cost. The larger half is process.
Do this on your own numbers. Take the fully loaded cost of the transactional part of the purchase-to-pay function, which means requisition handling, purchase order creation, supplier master setup, invoice matching and exception resolution, and leave out strategic sourcing and category management. Divide it by the number of purchase orders raised. In a business with 94,000 orders a year and 5.6 million of transactional cost, that is about 60 an order.
Then split the orders the way you split the spend. If 61,000 of those 94,000 orders sit with tail suppliers, the tail consumes about 3.7 million of processing to manage 50 million of spend, which is 7.3 per cent of the value passing through it. The managed spend runs 33,000 orders against 790 million, which is about 0.25 per cent. The tail is roughly thirty times more process-intensive per unit of value, and that ratio is stable enough across businesses that it is worth calculating before any other analysis.
That comparison reframes the objective. The prize in the tail is mostly the process cost and the risk surface, with unit price a distant third. Lisa Ellram's total cost of ownership work in purchasing, published through the mid-1990s, made the general version of this argument: acquisition price is a minority of what a purchase costs over its life, and the minority share is largest exactly where transaction volumes are high and values are low.
Kraljic's 1983 grid reaches the same place from a different direction, putting most tail items in the routine quadrant where the prescription is process efficiency rather than negotiation. DD1 works through how to pick a category strategy, so I will leave the grid there.
Mechanisms that move the break-even
The way to make the tail worth working is to change what a unit of work is, so that the fixed cost of buying gets spread over more spend or removed from the path entirely.
Aggregation into bundled events. Take forty small categories that share a supply base, and run one event covering all of them. The fixed cost is paid once rather than forty times, and the break-even drops by roughly the same factor. This works where the categories genuinely share suppliers, and it produces a bad outcome when someone bundles unrelated categories to hit a target and ends up with a supplier who is competitive in three of them and expensive in the rest.
Catalogues and punchout, so the requisition never becomes an event. Once a price is loaded and a requester can order against it, the marginal cost of the tenth purchase collapses. The work moves from per-transaction to per-catalogue, which is where you want it. The maintenance burden is real and it is the reason catalogues rot: a catalogue nobody updates becomes a channel for buying last year's prices.
Framework agreements and rate cards on categories that repeat. Pre-priced labour rates, standard freight lanes, laboratory consumables. The negotiation happens once for a term, and every requisition afterwards is an administrative act.
A low-value channel with the threshold set at the break-even. Below the number your own arithmetic produces, the correct policy is a purchasing card or a light-touch order against a preferred supplier with no sourcing step at all. Setting that threshold too low is the common error, because it forces buyers to run value-destroying events to stay compliant with a policy.
None of these reduces supplier count directly, which is deliberate. Supplier count falls as a consequence of catalogues and frameworks rather than as a target in its own right. Consolidation pursued for its own sake concentrates supply in ways that create a different problem, which DD3 covers.
The tail is ranked on the wrong variable
Spend is a poor proxy for consequence, and in the tail it is close to uncorrelated with it.
A calibration gas supplier billing 9,000 a year can stop a filling line inside a week. A two-person machine shop producing one bespoke gasket appears in the accounts payable file below every threshold anyone has set. A software licence renewal at 4,000 gates a system that the planning team runs on. A translation vendor holding the only approved version of your regulated label artwork sits nowhere on a spend report. None of these show up in a spend-ranked review, and all of them are single points of failure in the sense HH5 uses the term.
Ranking the tail on consequence rather than on value means answering a different question for each supplier: if this one stopped delivering tomorrow, how many days of production or how much revenue is at risk, and how long would a replacement take to qualify. Sort on that and a set of somewhere between fifty and three hundred suppliers separates out of the tail and belongs in a managed segment with contracts, contact details that are current, and a named alternative.
The reason this is rarely done is that the join required does not exist in the finance system. Accounts payable knows the supplier and the amount. Getting to consequence means linking supplier to part number, part number to bill of materials, bill of materials to finished goods, and finished goods to revenue or to a production line. That chain crosses three or four systems and usually breaks at the first step, because the purchase order line references a free-text description rather than a part.
Building that link is most of the work in a tail programme, and it is the part that produces something durable. It also has to survive the state of the supplier master, where the same company routinely appears four or five times under different spellings and legal entities, which is why the 8,900 in the opening paragraph might really be 6,200. T3 covers the resolution problem.
Where this stops
Tail savings are unusually hard to verify, and anyone running a programme should say so before the first number is reported.
The baseline problem is the core of it. On a managed category you have a contract price, a volume and a clean before-and-after. In the tail, the baseline is often a single historical invoice for a slightly different specification, bought under time pressure, at a price that may or may not have been representative. A saving calculated against that baseline is an estimate resting on an estimate. Van Weele made this argument about purchasing performance measurement generally as far back as 1984, that the absence of an agreed standard makes purchasing savings claims difficult to audit, and the tail is the extreme case of it.
Process cost is where the claims get softest. Removing 20,000 transactions a year from the purchase-to-pay flow is a real reduction in effort, and it turns into money only when a headcount, a contractor line or a system contract actually leaves. Otherwise it is capacity released, which is worth having but never reaches the accounts, and reporting it as a saving is how tail programmes lose credibility in their second year.
There is also a floor on how far this can go. Some fraction of the tail is genuinely irreducible: one-off engineering purchases, emergency maintenance, local services in places where you have one site, regulated purchases with a specified supplier. Chasing that fraction produces policy that people work around, and maverick spend created by an unworkable policy costs more than the policy saves.
Pull last year's accounts payable extract, filter to suppliers below your own break-even figure, and check how many of them appear on a bill of materials or a maintenance schedule. That count is the part of the tail that needs managing, and it is normally small enough to work through by hand this quarter.