Articles / Analytics
Spend analytics: build the data foundation before you buy any dashboard
Most procurement savings start with knowing what you actually buy. A practical guide to cleaning, classifying and using spend data.
You cannot manage what you cannot see. Yet many organisations cannot answer a basic question: how much did we spend with this supplier, on this category, last year? The reason is rarely a lack of tools. It is messy data.
Why spend data is messy
- The same supplier appears under five names, such as "ABC d.o.o.", "ABC doo" and "A.B.C. Ljubljana".
- Invoice lines say "misc" or "services".
- Different departments use different ledgers and category codes.
- Subsidiaries buy from the same supplier without anyone noticing.
A dashboard built on this data produces confident-looking charts that are wrong.
The four steps
1. Collect. Pull transactions from every source: the ERP, the accounting system, purchasing cards, expenses and, if you are a public body, your published contract data. Do not wait for perfection. Start with a full year of invoices.
2. Clean. Merge duplicate suppliers, standardise names and currencies, remove test records and fix obvious date errors. Use the tax or registration number as the master key whenever you can, because names vary but identifiers do not.
3. Classify. Assign every line to a category taxonomy. Modern language models do a good first pass on free-text descriptions. Have a category owner review the lines the model is unsure about, and feed corrections back so accuracy improves.
4. Analyse. Only now do you build views and dashboards.
Questions that spend data answers
- Where is the money? Typically a small number of categories hold most of the spend. Those deserve strategic attention.
- How fragmented are we? Fifty suppliers for the same item is an opportunity to consolidate.
- Who is buying off-contract? Spend outside agreed frameworks is lost leverage.
- What are the price differences? The same product bought at different prices by different units is a quick win.
- How dependent are we? A single supplier holding a large share of a critical category is a risk.
Metrics worth tracking
- Spend under management (the share covered by a contract or framework)
- Number of suppliers per category
- Contract compliance
- Price variance for identical items
- Share of spend with small and local suppliers, if your policy sets targets
Keep the list short. Five metrics that are reviewed monthly beat thirty that nobody opens.
For public buyers
Public bodies have an extra benefit: published notices and award data, increasingly released as structured open data in many countries, give you a view of the market and of peer organisations. Compare your lead times, number of bids per tender and award values against similar buyers. It turns a vague feeling of "we could do better" into a specific target.
Pitfalls
- Buying a tool first. The software is the easy part. Plan time for data work.
- Chasing perfect classification. Ninety percent accuracy on the biggest categories is more useful than full accuracy on everything.
- No owner. Assign someone to maintain supplier master data, or the mess will return within a year.
- Insight without action. Every analysis should end with a decision: consolidate, renegotiate, re-tender or leave as is.
The bottom line
Spend analytics is the groundwork that makes every other technology work. AI, automation and supplier management all rely on clean, classified data. Invest in the foundation first and the dashboards almost build themselves.