Classifying spend in 2026: five options compared

Quick answer: Sievo is an enterprise platform built for large, multinational spend volumes, with classification as one module inside a much broader suite. Pearstop is an AI-native classification engine built specifically for hard FM, construction, and manufacturing spend, priced per line rather than per seat. Coupa is built for the same scale, a full suite in which classification is again one module among many. Simfoni and Spendkey are narrower spend analytics tools aimed at mid-market teams. Manual in-house classification is slow and inconsistent, typically 60 to 75 percent accurate.

On this page: Spend classification options compared at a glance · Cost models behind each classification approach · Speed from raw spend to usable data · Industry depth versus generic spend platforms · FAQ

Spend classification options compared at a glance

A Head of Procurement asked to fix spend visibility this year is usually looking at some version of the same shortlist: buy an enterprise platform, find something built for the specific spend they actually have, buy a mid-market analytics tool, or keep doing it by hand. Each of those paths has a genuinely different shape, not just a different price tag.

We've compared six different options here: five platforms and one manual in-house approach.

Sievo, founded in Helsinki in 2003, is an analytics-first platform used by large multinationals such as Carlsberg and Deutsche Telekom. Its product suite covers spend analysis, savings tracking, contract analytics, and ESG reporting, with classification acting as the input layer to all of it. Coupa, founded in California in 2006 and taken private by Thoma Bravo in 2023, is built the other way around: it is a full business spend management suite covering procurement, invoicing, sourcing, and payments, and spend classification is one module inside that much larger system rather than the product itself. Simfoni, backed by Kearney and PeakSpan Capital, grew through acquiring EC Sourcing Group and Xeeva to add source-to-contract tools and data enrichment on top of its original spend analytics core, serving mid-market to enterprise teams. Spendkey, founded in London in 2020, is the newest and smallest of the four, offering spend analytics as a lighter, self-serve service. Pearstop, founded in Dublin in 2024, is the newest of the five, built as an AI-native classification engine from the start rather than an AI layer added to an older platform. Its team worked the underlying classification problem directly before building the product, which is why it competes with far more established suites on accuracy and speed. That same lean, AI-native build also keeps its cost base lower than a legacy platform's, reflected in pricing per line processed rather than per seat or module.

Manual classification is not a product at all. It is a person, or a small team, coding spend lines by hand in Excel or directly in ERP fields, category by category.

Cost models behind each classification approach

None of Sievo, Coupa, Simfoni, or Spendkey publish list pricing, which is normal for enterprise procurement software and not a criticism of any of them. Pearstop does not publish a rate card either, though its pricing mechanic is simpler to state outright: a cost per line processed rather than a seat or module fee, covered in full below. For the other four, what is public is the shape of the cost, not the number. Coupa's is the best documented of the four: independent estimates commonly cited for a 500 to 5,000 employee organisation deploying three or more Coupa modules put the annual subscription in the 200,000 to 800,000 dollar range, with implementation adding a further 100,000 to 400,000 dollars. That figure covers the whole suite, procurement, invoicing, sourcing, and payments, not classification on its own, which is the structural point: classification is a fraction of what the licence is paying for.

Sievo and Simfoni are priced the same way in principle, an annual licence sized to the organisation and the modules in scope, sold through an enterprise sales process. Spendkey, being smaller and newer, is positioned as a lighter subscription, though it too does not publish rates.

Pearstop's cost structure is different in kind, not just in size. It is priced per line processed rather than per seat, per module, or as a flat annual fee. Once the taxonomy and a client's own classification rules are built, the marginal cost of running one more line through the pipeline sits close to zero, so a heavy month of spend costs proportionately more and a quiet month costs less. That is a mechanical consequence of building the product around classification specifically, rather than allocating cost across a much broader source-to-pay suite that classification happens to sit inside.

Manual classification, by contrast, has no licence fee at all, which is exactly why its real cost is easy to miss. Classification consumes 60 to 70 percent of total spend analysis effort in traditional, hand-coded approaches, and manual accuracy typically lands between 60 and 75 percent. That is not a software cost. It is the fully loaded cost of the analyst hours going into work that, done by hand, still gets a quarter or more of the lines wrong.

Speed from raw spend to usable data

Enterprise suites take time to stand up because they are not doing one thing. Coupa, Sievo, and Simfoni implementations typically involve integrating multiple systems, configuring workflows well beyond classification, and training users across procurement, finance, and sourcing teams, because classification is one piece of a much wider rollout. None of that is a flaw in the products. It is the cost of buying breadth.

Pearstop is built to answer one question fast: what did we actually buy. Because it is a bolt-on to the CAFM or ERP a client already runs rather than a replacement platform, there is no source-to-pay rollout to wait through. The classification pipeline auto-assigns 90 to 95 percent of spend lines without human input for hard FM and construction clients, with the remainder routed to a short human review queue rather than left unclassified. That review feeds back into the model, so accuracy improves on the client's own data over time instead of resetting every reporting period the way manual coding does.

Manual classification has a speed problem of its own: it does not compound. A team that hand-codes this month's spend has to hand-code next month's spend the same way, at the same rate, with no learning loop carried forward unless someone writes the decisions down. Worst case, several people research the same ambiguous line independently, land on different answers, and nobody notices until a report does not reconcile.

Industry depth versus generic spend platforms

Sievo, Coupa, Simfoni, and Spendkey are built to classify spend across any industry, from retail to pharmaceuticals to manufacturing to services. That breadth is a genuine strength for a multinational buying office supplies, IT hardware, and travel across dozens of categories with no single dominant spend pattern. It is also, structurally, the reason none of them can go as deep on any one vertical as a platform built for just a few.

Hard FM, construction, infrastructure, and manufacturing spend look different from that kind of general spend, and they look similar to each other. An FM contract's invoices for reactive callouts, planned maintenance, and consumables share the same ambiguous line-item habits a construction subcontractor's invoices do: vague descriptions, inconsistent units, the same supplier billing the same service five different ways across five sites. Because Pearstop works only in these industries, the same taxonomy patterns, the same supplier naming conventions, and the same edge cases show up project after project, so classification logic built for one hard FM client carries over to the next rather than starting from a blank taxonomy each time.

That focus is the actual difference. It is not that a generalist platform classifies spend badly. It is that a platform built around one shape of messy data gets faster at that shape of messy data every time it sees it again, in a way a platform designed to also handle marketing spend and travel expense structurally cannot.

Stephanie Wiechers

CEO & Co-founder, Pearstop

Stephanie leads Pearstop's go-to-market and strategic direction. She works directly with procurement and FM leaders across Europe to understand how data quality affects margins, contracts, and AI readiness.

LinkedIn →