Spend classification software vs consultants vs in-house

Published 25 August 2026 · 6 min read

There are three ways to get your spend classified: pay a consultancy, do it internally, or use spend classification software. They differ less in quality than in cost, speed, and what happens six months later.

How much does spend classification cost?

Roughly, for a mid-size indirect dataset:

Consultancy: GBP 15,000–50,000 for a four-to-eight week engagement, repeated when the data ages.
In-house: two to four weeks of analyst time. No invoice, but it is the most expensive option if that analyst would otherwise be negotiating.
Software: priced per volume rather than per project, typically tens to low hundreds of pounds for a dataset in the tens of thousands of rows.

The numbers matter less than the shape. Consultancy is a large one-off; in-house is a recurring internal cost; software is a small recurring external one. Which suits you depends on how often you need to do this.

Consultancy

A spend analysis engagement typically runs four to eight weeks and costs somewhere between fifteen and fifty thousand pounds, depending on data volume and how much strategy work is bundled in.

What you get: classified data, a category structure, and — the real product — a set of recommendations from people who have seen the same categories at thirty other companies. That benchmarking is genuinely hard to replicate.

The catch: it is a snapshot. The classification describes the data you handed over, and starts ageing the day it is delivered. Twelve months on, most teams are choosing between paying again and working from a stale file. Ask upfront what a refresh costs — the answer shapes the whole decision.

In-house

A capable analyst with Excel can classify a mid-size dataset. The marginal cost looks like zero because the salary is already committed.

What you get: institutional knowledge that stays. Someone who understands why intercompany recharges are excluded and which supplier is really three legal entities is worth a great deal, and no external party will acquire that.

The catch: it is slow, and the cost is opportunity cost. Three weeks of an analyst is three weeks not spent on negotiation. It also concentrates risk in one person — when they leave, the logic leaves with them unless it was written down, and it usually was not.

Automated tools

Classification runs in minutes rather than weeks, priced per volume rather than per project.

What you get: speed, and repeatability. Re-running next quarter costs the same as the first time, which changes spend analysis from a project into a routine.

The catch: a tool classifies, it does not advise. It will not tell you that consolidating two suppliers breaks a single-source risk you accepted deliberately. Output still needs a procurement brain applied to it, and any vendor implying otherwise is overselling.

Judge one on three things: how it handles the long tail rather than the top 100 suppliers, whether you can review a sample before committing, and whether re-running produces the same taxonomy or a slightly different one each time. Instability there is a real problem — a category that moves between runs makes year-on-year comparison impossible.

Choosing

One-off, board-level, needs external credibility — consultancy. You are buying the recommendations and the signature, not the classification.
Small dataset, analyst available, no deadline — in-house. Write down the rules as you go.
Recurring, large, or long-tail heavy — a tool. This is the case where manual effort scales worst and automation scales best.

The combination that works well in practice: automate the classification, keep the interpretation in-house, and bring in a consultancy only for the categories where you genuinely lack market knowledge.