Published 3 September 2026 · 5 min read
A fair question with a wide answer. The honest range for a first spend analysis is two days to three months, and which end you land on depends far more on data access than on analysis.
Where the time actually goes
- Getting the data: days to weeks. Almost always the longest phase, and almost never the one people budget for. Someone has to own the ERP export, and that person has other work.
- Cleaning it: hours to days. Supplier name normalisation, removing intercompany, deciding on credit notes, reconciling to finance.
- Classifying it: minutes to weeks. The widest variable. Manual classification of a long-tailed dataset is the multi-week case; automated is minutes plus review time.
- Analysing and writing up: one to three days.Genuinely quick once the data is categorised.
Realistic totals
- Small, one site, clean export, automated classification: two to three days.
- Mid-market, one ERP, some manual review: one to two weeks.
- Multi-entity, several systems, manual classification: six to twelve weeks.
- Consultancy engagement: four to eight weeks, much of it waiting for your data.
What makes it slow
- No single owner of the data request. The biggest cause by far. Name one person before you start.
- Multiple entities on different systems. Every additional source adds a reconciliation.
- Trying to be complete. Chasing the last 5% of unclassified rows can double the timeline and change no decision.
- Designing the taxonomy from scratch by committee. Anchor to a standard and adjust.
How to make it fast
Ask for twelve months of invoice data with supplier, description, amount and date. Reconcile the total. Classify. Stop at 95% coverage and label the rest as tail. Write five recommendations.
That sequence is achievable in a week for most mid-market datasets, and the result is not meaningfully worse than the three-month version — because the decisions it supports are the same ones.
The second one is much faster
Most of the effort above is one-off: finding the data owner, agreeing exclusions, designing the taxonomy. Repeat runs reuse all of it. If your second analysis takes as long as your first, something in the process was not written down.
Further reading
- Preparing an ERP export — how to shorten the longest phase.
- Software vs consultants vs in-house — timelines by route.
- How Structera works — the classification step in minutes.