Published 14 August 2026 · 6 min read
Most spend analysis stalls at the first step: getting the data out of the ERP in a usable shape. This is what to ask for, and what to check before you start.
The columns you actually need
Only four are essential:
- Supplier name — the raw string, not a code.
- Line description — what was bought. The single most important field for classification.
- Amount — net of tax, in a single currency if possible.
- Date — posting or invoice date, consistently one or the other.
Useful if easy: supplier ID, PO number, cost centre, currency, quantity, unit of measure, and any existing category code (worth capturing to compare against, not to trust).
Line level, not header level
This is the request people get wrong. A header-level export gives one row per invoice with a total and often no description. A line-level export gives one row per item, with the description that makes classification possible.
If your finance system can only produce header level, the analysis is still possible but the granularity drops to supplier level — you will learn who you spend with, not what on.
How much history
Twelve months minimum, so seasonality is covered. Twenty-four to thirty-six is better because it lets you see trend and catch annual contracts. Beyond thirty-six the data usually predates a reorganisation or system change and adds noise.
Common system-specific traps
- SAP. The description field is often truncated at 40 characters. Check whether a long-text field exists.
- Oracle. Distribution lines can multiply rows — one invoice line split across cost centres appears several times. Deduplicate or accept that totals will overstate.
- Excel exports generally. Long numeric part codes get coerced into scientific notation. Export as CSV with text formatting, or check for values like “5.0044E+11”.
- Any system. Ask whether the export is net or gross of tax, and whether credit notes are included.
The reconciliation step
Before you analyse anything, sum the amount column and compare it to the total spend finance reports for the same period. If they do not broadly agree, find out why now. The usual causes are intercompany charges included or excluded, tax treatment, or a date-range mismatch.
Any conclusion built on an unreconciled file will be challenged the first time it reaches a CFO, and rightly.
What to do with an imperfect file
You will not get a clean export. That is normal. The usable minimum is a supplier name, a description with real words in it, and an amount that reconciles. Everything else can be worked around or noted as a limitation.
Further reading
- 7 spend data quality problems — what to check once you have the file.
- How Structera works — the upload template and workflow.
- Building a spend cube — what comes next.