Less Than One Percent

Procurement's data problem was always a funding problem

Procurement controls more money than any other function in most companies and is funded with less than one percent of it. Thirty years of that has left a function that spends its days reacting to events it was never equipped to anticipate, not for lack of skill but for lack of information. The expensive part of fixing it was always the same unglamorous task: making messy transactional data trustworthy. That task recently got cheap.

Troy, Ohio

At 8:01 on the morning of June 26, 1974, a man named Clyde Dawson took a pack of Wrigley's Juicy Fruit out of a shopping basket and slid it across a glass window set into the checkout counter of a Marsh supermarket in Troy, Ohio. A machine under the glass read the stripes printed on the wrapper. The register produced a price. Dawson was the head of research and development for the grocery chain and the basket was a prop; one of the original scanners from that store is now in the Smithsonian.

What makes the moment worth remembering is not the technology. It is what the grocery business had been doing until 8:00 that morning.

A supermarket in 1973 did not know what it sold. It knew what it had bought, and it knew what the registers took in by closing time, and in between there was fog. Which of the nine brands of dish soap actually moved. Whether the end-cap display did anything at all. What time the store was busy. Whether last week's promotion brought in new shoppers or just sold the same soap, cheaper, to the same people. Grocers had views on all of it. Some of those views were excellent, held by people who had spent thirty years standing in stores watching customers. What they did not have was information.

Here is the part that gets left out of the story. The scanner did not sweep the industry. Installing one cost more than $200,000 in 1977 dollars, and by the end of that decade roughly one percent of American grocery stores had one. When the economist Emek Basker went back through census records decades later, she found that early scanners raised store labor productivity by about 4.5 percent in the first few years — real, but not enough. Her conclusion was blunt: the first generation of machines probably did not provide a positive return on investment. For years, item-level knowledge of your own business was a luxury good that didn't work especially well even for the people who could afford it.

Then the economics changed. By the end of the 1980s scanning was the norm rather than the exception, and the fog lifted for everybody at once.

And what the grocers found on the other side was not a weapon. In 1995, two researchers went looking for the obvious consequence — proof that all this new data had let retailers squeeze profit out of the manufacturers who sold to them. Paul Messinger and Chakravarthi Narasimhan examined two decades of accounting and stock-market measures and reported that it hadn't happened: retailers had not gained profitability at manufacturers' expense. What the scanner delivered was less cinematic and considerably more useful. It ended the guessing. In 1986 a grocery executive described to the CBC what he was about to be able to see: "With the scanning data, we'll know exactly how our customers shop: what time they come, how often they come." Decisions that had been made by instinct were about to be made by looking.

Which brings us to a function sitting inside almost every company in the world that is still on the wrong side of 8:01 a.m.

Less than one percent

A company will spend a billion dollars with suppliers and allocate something in the neighborhood of six or seven million to the people and systems responsible for that billion.

Procurement manages the largest controllable cost base most companies have. For a typical manufacturer or distributor, what gets paid to outside suppliers dwarfs payroll, dwarfs capex, and is the only major line on the income statement that a single department can materially move without asking a customer for anything.

Hackett Group benchmarking has put the cost of running that department at 0.72% of the spend it manages, and 0.59% among the organizations Hackett classes as world class. CAPS Research, cutting the same question by industry, found an average of 1.2%: 0.4% in financial services, 0.7% in utilities. Call it a range rather than a figure: depending on who is counting and which industries sit in the sample, the answer lands somewhere between half a percent and a little over one.

Sit with those numbers for a second, because they are the whole argument. Then the company will ask those people why the savings number isn't bigger.

The comparison that makes it vivid is what happened to every function standing next to procurement. Sales got a system of record, and then a second one, and then an entire industry of tools built on top. Finance got the ERP, and then planning software on top of the ERP. HR got a human capital platform. Marketing got a stack so elaborate it needs its own operations team to run it. Each of those investments was justified the same way: this function makes decisions worth a great deal of money, and it should not be making them blind.

Procurement got a spreadsheet. As of a 2019 study, half of procurement organizations were still using legacy tools such as Excel to store and analyze their data. In a 2020 Harvard Business Review Analytic Services survey of 779 executives, 60% named the lack of transparency between finance, procurement and their suppliers as a business risk, and 24% said flatly that they do not effectively evaluate their own suppliers' business practices, citing manual and incomplete data entry as the barrier.

The gap is visible from the other direction too. Deloitte's most recent global survey finds that the procurement organizations it classes as Digital Masters already put about 24% of their budgets into technology, rising to 26% next year. That is what it looks like when a procurement function is properly equipped. It is also, by implication, a measure of the gap everyone else has to close.

None of this was a scandal, just a series of individually defensible decisions. Every CFO who declined to fund a spend analytics program was choosing between a concrete request for money and an abstract promise of savings, and taking the abstract promise on faith is not what CFOs are paid for. The trouble is what accumulates when you make that same reasonable call for thirty years.

Because underinvestment does not show up as a line item. It shows up as a missing foundation.

The second job

The fires are not the alternative to strategy. They are the interest payment on thirty years of deferred investment.

Ask a procurement leader what their team spent last quarter doing, and you will not hear about category strategy. You will hear about the expedite request that came in on a Friday. The supplier who went sideways. The invoice that didn't match the purchase order. The requisition that had to be approved today because a plant was waiting. The price increase letter that arrived as a complete surprise, because nothing anywhere in the company was watching for it.

This is usually described as procurement being too tactical, as though it were a matter of discipline — as though the team could simply decide to spend more time on strategy. That gets the causation backwards, and it is the most important thing in this paper to get right.

The tactical work is not a distraction from the strategic work. It is what strategic work looks like when you have no information.

A team that cannot see a price trend forming has no option but to react when the increase lands. A team that cannot see how much of its spend runs through a single supplier has no option but to scramble when that supplier fails. Reactivity is a symptom of systematic blindness, not a failure of ambition or skill. You cannot get ahead of something you cannot see coming, and every hour spent on the emergency is an hour not spent on the thing that would have prevented it.

And then, layered on top of the firefighting, there is the second job.

Before anyone can analyze spend, somebody has to make the data usable — pull it out of one system or eleven, standardize the units, resolve the fact that the same supplier appears as ACME INDUSTRIAL, Acme Ind., and ACME-IND/TX, and assign several hundred thousand transactions to categories that mean something. In the industry this is called block-and-tackle work, which is a polite way of saying it consumes the capacity the fires left behind. The analyst who was hired for their judgment about markets spends three weeks a quarter mapping vendor names.

That is what underinvestment bought. Not just fewer people. A function whose most expensive talent is occupied with data janitorial work, in service of a foundation that was never funded, so that it can react to events it was never equipped to anticipate.

It was never a size problem

Money bought a snapshot, and snapshots age.

The obvious inference is that this is a problem of scale — that big companies solved it and everyone else is waiting their turn. The obvious inference is wrong, and this is where the story rhymes with Troy, Ohio.

The traditional fix was priced as a luxury good: an enterprise implementation, a taxonomy built by consultants, a data warehouse, a small standing army. Most companies could not buy it. In one survey of mid-market firms, 53 percent called spend analytics a high or essential priority (rising to 61 percent among those between $500 million and a billion in revenue), and the obstacles they named were exactly the block-and-tackle ones: data extraction, cleansing, classification. They knew what they were missing. That was the whole of their relationship with it.

But the companies that could pay did not escape either. They bought a project when what they needed was a capability. A one-time cleanse is stale the quarter after it ships. A hand-built taxonomy survives until the next acquisition, or the next ERP migration, or the arrival of a supplier nobody wrote a rule for. That same survey turned up a small but telling detail: of the dozen companies that had gone ahead and adopted newer autoclassification and dashboard tools, most reported mixed results, because the data underneath them was still bad. Buying the tool did not fix the foundation.

So the failure takes two shapes, and they look nothing alike from the inside.

The large enterprise drowns. Eleven ERPs from nine acquisitions, forty legal entities, three competing taxonomies, the same supplier under nine different names, and more reports than anyone can read. Every question becomes a reconciliation project, and the answer arrives late enough not to matter.

The mid-market starves. One ERP, no classification layer, no analyst who owns it, and a procurement director who could tell you precisely which two categories they'd attack if they could see them.

Different mechanisms. Identical outcome: data that exists and cannot be acted on. A thirty-billion-dollar manufacturer with eleven ERPs and a three-hundred-million-dollar distributor with one are, functionally, equally blind. Both have been told their problem is their size. It isn't. It's maturity — and until recently, maturity was not something you could buy.

What changed

It has become tiresome to be told that AI changes everything, so let me make the narrowest claim we can, because the narrow version is the strong one.

Artificial intelligence did not make procurement strategic. It did not invent spend analysis, which is not remotely a new idea. What it did was collapse the cost of the single task that gated everything else: making messy transactional data trustworthy at scale.

Classification used to mean writing rules. A human being decided that anything from a vendor whose name contained "STEEL" belonged in metals, and then someone else discovered Steele Consulting in the metals bucket, and the rules multiplied until the rulebook itself became the thing nobody could maintain. Learned classification works differently — it infers category from patterns across the whole record, description, amount, cadence, counterparty, and it can be re-run continuously rather than rebuilt periodically. Enrichment that once required a research team can be pulled from external sources automatically. The economics of the block-and-tackle work fell through the floor.

The consequence is bigger than spend analytics getting cheap enough for small companies. The approach finally matches the shape of the problem, at any size. Continuous instead of one-time. Learned instead of hand-built. Which means it survives the acquisition, the migration, and the supplier who shows up next Tuesday under a name nobody has seen before.

For the mid-market, that is access to something they were previously priced out of. For the enterprise, it is delivery of something they have paid for repeatedly and never actually received: a foundation that stays true. Cost is no longer the gate for the small. Complexity is no longer the gate for the large.

This is what we mean by uplift, and it is worth being precise about it, because you have been sold a dashboard before and were not impressed. The uplift is not a better report on what you already knew. It is access to information you have never had — the ability to ask questions that were previously unaskable, about your own money. The Uplift Engine™ is the part that does the unglamorous work: continuously cleansing, classifying and enriching procurement data so that it stays trustworthy instead of decaying. That is the product claim in this paper, and it is the reason we started there rather than somewhere more exciting. There is nowhere else to start. Everything else in procurement technology is built on top of an assumption about data quality that, in most companies, is not true.

Eyes open

Information asymmetries close when somebody turns on a light.

What follows from a trustworthy foundation is a change in operating mode rather than a report, and the difference is easiest to see in pairs.

That last one deserves a moment. In most supplier negotiations the supplier knows more than the buyer does: what the buyer paid last time, what the buyer's sister division pays, how much volume is actually at stake, what the cost drivers have done. This is not a fair fight, and it has nothing to do with the negotiator's skill. It is an information asymmetry.

All of which belongs in front of a board rather than inside a procurement review.

EBITDA. A dollar taken out of third-party spend arrives at EBITDA at close to full value. It requires no new customer, no new headcount, no capital. At a ten percent operating margin, that dollar does the work of about ten dollars of new revenue — and then it gets multiplied by whatever the market pays for the company's earnings. This is arithmetic, not a promise. It is also, for most companies, the cheapest enterprise value available anywhere on the income statement.

Risk. Where is our spend concentrated. What breaks if one supplier stops shipping. How much of our cost base sits in a single country. Which contracts renew automatically in the next ninety days, and what are they worth. These are board questions, and procurement is the only function positioned to answer them, which it has historically been unable to do, because the answers live in transaction data nobody could read. Value leaks quietly here too: World Commerce & Contracting, with Ironclad, puts post-signature contract value leakage at around 11 percent, and notes that the real figure is probably higher because most organizations do not track it at all. McKinsey estimates that unfulfilled supplier obligations alone leak roughly 2 percent of spend at large enterprises ($40 million a year on $2 billion), and that as many as 80 percent of procurement functions are not fully aware of their own competitive terms and contract structure.

Competitive advantage. A company that understands its cost structure better than its competitors understand theirs can price differently, bid differently, and commit differently. Cost intelligence is a strategic asset rather than a back-office report, and it happens to be one of the few that competitors cannot buy off the shelf.

The evidence that this is real rather than aspirational is now reasonably good. Deloitte's 2025 Global Chief Procurement Officer Survey compared the organizations it calls Digital Masters against those it calls followers.

Hackett's separate 2025 research on digital world-class procurement organizations found they lose 60% less savings to noncompliance and maverick buying, and deliver 2.6 times the return on investment with 31% fewer people.

Notice which measure separates the two groups most dramatically. Not savings. Innovation enablement — the most forward-looking item on the list, the one furthest from firefighting. That is not a coincidence. That is the entire thesis in a single statistic: what visibility buys you is not a better answer to this quarter's emergency. It is the capacity to be somewhere other than the emergency.

Why we keep score in ROI

A word about our own posture, since a paper like this is not neutral.

Every vendor says they focus on ROI, and the phrase has been worn smooth. Our reason for it is mechanical. A function that has been underinvested in for thirty years does not get its budget back by asking for it. Underinvestment is a loop: thin resourcing produces soft numbers, soft numbers weaken the investment case, and a weak case produces thinner resourcing. The only thing that reverses the direction of that loop is a return legible enough that the next investment becomes obvious to the person holding the budget.

So ROI is the mechanism by which an uplift becomes permanent rather than a one-time tooling refresh that gets cut in the next downturn. That is why we lead with it.

Which means the standard applies to us. A platform sold on the promise of ending three decades of underinvestment should pay for itself quickly, visibly, and in numbers a CFO can tie to something. We would rather be judged that way than on adjectives.

Troy, Ohio, again

The grocers of 1973 were not bad at their jobs. They were extremely good at a job that had to be done without information, and the people who were best at it had developed instincts that were genuinely valuable. What the scanner did was not replace their judgment. It aimed it.

There is a pleasing coincidence in the numbers. By the end of the 1970s, about one percent of American grocery stores could see what they sold. And grocery ran on margins that the CBC, reporting in 1986, put at around one percent, which is precisely why nobody in that business could afford to let a competitor see something they couldn't.

Procurement today runs the largest controllable cost base in the enterprise on less than one percent of it. The difference between 1974 and now is that the fog is no longer expensive to lift. It is no longer a question of whether a company is large enough, or whether the project can be justified across three budget cycles, or whether there is an analyst free to maintain the taxonomy. That constraint is gone.

What remains is a decision about whether a function that has spent thirty years fighting fires in the dark should be handed a light.

Sources

• The Hackett Group (2016) and CAPS Research (2014) benchmarks on the cost of procurement as a percentage of spend, via the Institute for Supply Management. These remain the most recent public benchmarks of their kind that we could locate.
• The Hackett Group, Digital World Class Procurement research, July 2025.
• Deloitte, 2025 Global Chief Procurement Officer Survey.
• Harvard Business Review Analytic Services for Basware, survey of 779 executives, 2020; LevaData, Cognitive Sourcing Study, 2019, on spreadsheet use in procurement.
• Mid-market spend analytics survey (419 usable responses, companies of $100M–$1B revenue), reported in Supply & Demand Chain Executive, 2018.
• World Commerce & Contracting, with Ironclad, Closing the Procurement Value Gap, on post-signature contract value leakage; McKinsey & Company, Mitigating procurement value leakage with generative AI.
• Emek Basker, Raising the Barcode Scanner: Technology and Productivity in the Retail Sector, US Census Bureau Center for Economic Studies working paper, 2011.
• Paul R. Messinger and Chakravarthi Narasimhan, "Has Power Shifted in the Grocery Channel?", Marketing Science, 1995.
• History.com and the Smithsonian National Museum of American History on the first UPC scan, June 26, 1974; CBC Archives, on grocery scanner adoption, 1986.