How to Control Maverick Spend: Why It's a Data Problem, Not a Discipline Problem

Maverick spend is not a discipline problem

Maverick spend — the practice of purchasing goods or services outside of established, negotiated contracts and procurement frameworks — is traditionally viewed as a compliance failure. Site managers buy what they need from whoever answers the phone. Project managers use their preferred subcontractors. Engineers order parts at retail because the approved vendor takes three days and the line is down now.

The instinct is to fix this with policy: mandate the approved vendor list, restrict purchase card use, require additional sign-offs. Some of that is appropriate. But policy enforcement without data visibility is friction without insight. You are creating obstacles to a behaviour you cannot accurately measure.

The organisations that reduce maverick spend most effectively do something different first. They classify their spend at commodity level. And then they look at what that classification reveals.

Why FM and construction are structurally exposed

In most industries, purchasing is reasonably centralised. A procurement function sits between demand and supply, and most transactions flow through it. In FM and construction, that model breaks down for structural reasons.

In hard FM, a reactive maintenance call arrives at 11pm. The on-call engineer needs a replacement part. The approved distributor's branch is closed. The nearest trade counter has the part at list price. The part gets bought, the job gets done, and the transaction is booked to the site maintenance budget as "parts." No one considers whether there was a framework agreement covering that part. No one checks whether the critical spares stock at that site should have contained it. The data goes into the ERP as a free-text line item with no commodity code.

In construction, every project manager builds a trusted network of subcontractors over years of working together. That network is real, valuable, and entirely outside any procurement framework. When a PM needs a groundworks contractor quickly, they call the person they know. The result is that a €500M construction portfolio might have 200 groundworks transactions with 60 different suppliers, with price variance of 20–40% on comparable work, hidden inside individual project budgets.

These are not discipline failures. They are the predictable outputs of operational environments where speed matters and procurement data is too poor to support centralised control.

What gets exposed when you classify your spend

Classification does not prevent maverick spend. It makes it visible. And visibility is what makes the subsequent conversation about policy and frameworks factual rather than anecdotal.

Here is what typically surfaces when an FM or infrastructure organisation runs a commodity-level classification exercise on 12–18 months of transaction history:

• Duplicate vendor spend on identical items. In MRO environments, the same bolt — identical manufacturer, identical part number — appears purchased from three different distributors at three different prices across different sites. Before classification, those transactions are three separate line items with different descriptions ("M12 bolt 50mm," "hex bolt stainless M12," "fastener stainless 50mm"), different GL codes, and no apparent connection. After classification, they resolve to a single UNSPSC commodity code. The price variance — typically 25–50% between the cheapest and most expensive source — is immediately visible. At volume, that variance is recoverable money.

• Contract leakage on negotiated categories. An organisation may have a framework agreement for electrical consumables with a preferred distributor at agreed pricing. Classification of invoice data frequently reveals that 30–40% of electrical consumable spend is going to out-of-framework vendors. This is not visible from the framework vendor's invoices alone — you only see it when you classify the total population of electrical consumable spend across all sites.

• Reactive vs. planned maintenance spend ratio by asset class. In hard FM, emergency reactive maintenance typically costs 3–5 times the equivalent planned maintenance for the same asset work. Classification allows you to separate reactive from planned spend by asset type. When that analysis shows that 70% of HVAC maintenance spend is reactive, it is an asset management problem with a cost implication — not just an operational nuisance. You cannot run that analysis without commodity-level classification of maintenance spend.

• Invisible categories with significant volume. The most common finding in a first-pass classification exercise is a category that procurement had no visibility of. Grounds maintenance distributed across project codes. Temporary power generation scattered across site variation orders. Waste disposal booked to a dozen different cost centres. Each of these is a real spend category with leverage potential — but only if you can see it as a category. Classification makes that possible.

The MRO case: €200k from parts no one was managing

A UK-based MRO manufacturer — 40,000 active parts in the catalogue, multi-site production environment — ran a parts enrichment and classification exercise as a prerequisite to a sourcing review. The immediate objective was to clean up the asset and parts master data, which had accumulated significant duplication and inconsistency over years of organic growth.

The classification output identified a subset of approximately 800 parts where the current purchasing route was a distributor, but where OEM direct-sourcing was viable at significantly lower unit cost. The distributor margin on these parts ranged from 25–50%. At the purchase volumes involved, switching those 800 parts to OEM direct represented approximately €200k in annual savings.

None of that was visible before classification. The parts existed in the system under inconsistent descriptions across sites. There was no way to aggregate purchase volume by part, because the descriptions did not match. Classification resolved that. Once every part had a consistent commodity code and manufacturer reference, the volume aggregation was straightforward. The sourcing opportunity was obvious from the data.

This is the pattern. The savings do not require novel procurement strategy. They require visibility of what is already happening.

The Golden 500 approach

One of the most effective interventions for MRO and maintenance spend control is the virtual catalogue — a mandated list of pre-priced items that site and maintenance teams must use for common purchases. The concept is simple. The implementation challenge is always the same: which 500 items should be on it?

Without classification, that question cannot be answered reliably. You can ask site managers what they buy most frequently, and they will give you a list that reflects their individual site rather than the portfolio. You can look at distributor invoices, but descriptions vary by supplier and by purchaser. You cannot aggregate across sites without a common classification scheme.

With classification, the answer is a query. Rank all commodity codes by transaction frequency and total spend across the portfolio. The top 500 items by transaction volume, across all sites, are your catalogue candidates. Every one of them is already being purchased repeatedly. The catalogue does not change the demand — it changes the purchasing route, the pricing, and the control.

The classification exercise is what makes the catalogue buildable. Without it, you are asking site managers to agree on a list through a committee process that takes months and produces a catalogue shaped by whoever talks loudest rather than by actual purchase data.

For a deeper look at how classification enables category management strategy — Kraljic positioning, framework agreements, volume consolidation — see the post on why UNSPSC classification is the foundation of category management in FM and infrastructure.

Price variance across sites: the number that moves CFOs

The conversation with a CFO about maverick spend is difficult when the evidence is anecdotal. "Site managers are buying outside the framework" is a complaint. "We have 35% price variance on M&E consumables across sites, representing approximately €340k in recoverable spend annually" is a business case.

Classification produces the second type of evidence. When you can show — at commodity level — that the same item is being purchased at materially different prices across your site portfolio, the conversation about catalogue compliance and framework enforcement becomes specific and quantifiable.

Typical price variance findings in FM and construction MRO: 20–40% variance on the same SKU across sites, concentrated in reactive maintenance categories. For an organisation spending €5M annually on MRO, the recoverable portion of that variance — even assuming partial compliance improvement — is a material number. Classification is what surfaces it.

What this requires in practice

Exposing maverick spend through classification is a data exercise before it is a procurement exercise. The steps are sequential:

First, extract 12–18 months of transaction data from your ERP — SAP, Oracle, or equivalent. Supplier name, description, quantity, price, GL code, cost centre, and date are the minimum fields required.

Second, run commodity-level UNSPSC classification on that dataset. This is where automated classification tools earn their value. A dataset of 100,000 transactions classified manually would take months. Automated classification at 90–95% accuracy processes the same dataset in days.

Third, run the spend analysis on the classified output. Price variance by commodity across sites, vendor count by commodity, spend inside vs. outside framework, reactive vs. planned maintenance ratio by asset class. These analyses are queries once the data is classified. They are impossible before.

Fourth, take findings to the business. Not to enforce compliance — yet — but to make the current state visible to the people who can change it. Operations directors and CFOs respond to specific numbers. Category managers need the data to make the case internally.

The policy conversation comes after the data conversation. Enforcement without visibility creates resistance. Enforcement after visibility creates understanding.

Frequently asked questions

How do we know whether our maverick spend is a data visibility problem or a genuine compliance problem?

The test is whether procurement can currently show, at commodity level, what is being purchased outside negotiated frameworks. If that analysis does not exist — if the answer is "we think it is significant but we cannot quantify it" — you have a visibility problem. Compliance problems can only be assessed after visibility problems are solved. Classification is the first step.

We have an approved vendor list. Does that not already control maverick spend?

An approved vendor list controls which suppliers are authorised. It does not control whether purchases from those suppliers are going through negotiated contracts at agreed pricing, versus spot purchases at list price from the same supplier. Classification of transaction data distinguishes between these — you can see, for each commodity, whether the unit price paid aligns with the framework rate. Approved vendor lists create the policy; spend classification measures compliance with it.

At what spend volume does a classification and maverick spend analysis become worthwhile?

As a rough guide: if you have more than 20 sites or projects, or more than €5M in annual MRO and maintenance spend, or more than 500 active suppliers, the analysis will almost certainly surface recoverable spend that exceeds the cost of running it. The MRO manufacturer case above involved 40,000 parts and produced €200k in annual savings. Smaller datasets produce proportionally smaller findings, but the cost of a classification exercise scales with data volume — the analysis itself is not expensive relative to what it typically finds.

Related reading: Why UNSPSC Classification Is the Foundation of Category Management in FM and Infrastructure | UNSPSC Classification Accuracy: What 90–95% Actually Means

Pearstop Team

Pearstop

Pearstop helps procurement and operations teams in hard services, FM, construction, and manufacturing turn messy data into a reliable foundation for decisions, AI, and category management.

LinkedIn →