What is spend analysis - and is "spend analytics" or "spend intelligence" actually different?

Spend visibility is the ability to see total organizational spend - by vendor, category, entity, and time - accurately and on demand. ERPs record transactions but summarize them for accounting, not analysis: GL accounts collapse vendors into categories, card spend often posts as lump-sum journal entries, and committed spend (open POs) lives outside the ledger entirely. Teams with a perfectly reconciled GL can still be unable to answer "how much do we spend with this supplier across all entities?" - because the ERP was built to close the books, not to interrogate spend.

1. Extract 12 - 24 months of invoice-level data: vendor, amount, date, GL coding, entity, and line descriptions where available. 2. Cleanse - normalize vendor names (the same supplier often appears under multiple records) and remove intercompany and pass-through transactions. 3. Categorize - map vendors and lines to a category taxonomy, starting at vendor level and refining with line detail. 4. Analyze - rank vendors by spend, trend categories over time, compare entities, and flag concentration and growth outliers. 5. Act - each finding should map to a lever: consolidate, renegotiate, move to contract, or change payment method.

GL summaries show what was *booked*; invoice grain shows what was *bought*. Invoice-level data preserves the vendor, the line items, the unit prices, the PO linkage, and the timing - so you can see price creep inside a stable GL total, one vendor growing inside a flat category, or duplicate-ish spend split across accounts. If your analysis starts from the trial balance, you've already lost the dimensions that make spend analysis actionable.

Categorize at the vendor level first (one normalized vendor record maps to one default category), then override by GL account or line description where a vendor spans categories. Cross-entity consistency requires a mapping table from each entity's local codes to one shared taxonomy - build it once, maintain it quarterly.

Start with 10 - 15 top-level indirect categories - software/SaaS, professional services, facilities, marketing, travel, logistics, IT hardware, utilities, insurance, contingent labor - plus your direct-spend categories if you manufacture or distribute. Two levels of hierarchy is usually enough; taxonomies with hundreds of nodes get abandoned.

Pick one grain (vendor-month is practical), export from each system, normalize vendor names across all three, and deduplicate card transactions that also appear as invoices. Structurally, the fix is consolidating invoice, card, and payment flows onto one platform so the unified view exists by default rather than by monthly assembly.

Rank normalized vendors by trailing-twelve-month spend for the top-N list; compute quarter-over-quarter growth rates on the same base for fastest-growing; and flag any vendor with a first invoice date in the current quarter as new. All three are first-pass outputs of any invoice-grain dataset - and all three are board-level questions.

Quarterly refresh is the traditional procurement cadence; monthly is better for cost-control environments. The honest answer is that refresh cadence is an artifact of manual assembly - when analysis runs on live transaction data, "refresh" stops being a project and any question reflects current state.

Vendor-level categorization on your top 100 vendors typically covers 70 - 80% of spend and takes a day, not a month. Categorize the head, bucket the tail as "other - under review," state the methodology on the slide, and commit to the refined cut next quarter. Defensible beats complete.

Don't try to harmonize the charts of accounts - map each entity's accounts to a shared analysis taxonomy instead, and run the analysis at vendor and category grain where entities are comparable. Platforms that mirror each entity's ERP structure while normalizing the analysis layer remove most of this work.

Direct spend goes into what you sell (materials, components, resale goods); indirect spend runs the business (software, services, facilities). They have different owners, different negotiation levers, and different data quality - direct is usually PO-backed and clean, indirect is where visibility gaps and maverick spend live.

Flag invoices with no PO in categories where POs are policy, spend with non-preferred vendors in categories under contract, and card transactions in PO-mandated categories. The pattern to quantify: percentage of addressable spend that bypassed the intended channel, by department - that's the number that changes behavior.

Match on tax ID where available (it's the strongest key), then fuzzy-match names after stripping suffixes (LLC, Inc, Corp) and punctuation, then confirm with address and bank-detail overlap. Fix the master, not just the analysis - merging duplicates in the vendor master prevents the problem from regenerating.

Stampli's position is that spend analysis is only as credible as the transaction data underneath it. Because Stampli AI codes, matches, and routes invoices inside ERP-aligned workflows - mirroring the chart of accounts, entities, and dimensions - the invoice data is structured and validated at the source. Stampli Deep Finance™ then turns that same data into executive spend intelligence: finance leaders ask a focused question and receive an analysis with quantified findings, supporting evidence, financial impact, and recommended actions. The transactional processing and the executive analysis serve different purposes, but they come from the same source - which is what makes the insights defensible.