Skip to content
Classification and Normalization

Your spend, structured once and kept current.

The permanent foundation under everything.

Messy spend from ERPs and files becomes one clean, governed record, classified into your own 4-level taxonomy. Every Caddie, Space, Scan and Scout runs on top of it.

INV-20931.pdf · Acme Industrial LLC

ACME IND,96000,,USD,,

Re: fastener quote v3 · ACME-INDUSTRIAL

{"vendor":"Acme Ind Sply"}

Acme Industrial Supply · 4 sources merged

L1 Indirect

L2 MRO

L3 Electrical supplies

L4 Wire and cable

Vendor aliases resolve to one canonical identity. Every line is classified through all four levels of your taxonomy, by proprietary AI plus rules.

Invoice Augmentation

Recovers what your ERP dropped.

Reads invoice PDFs, scans and attachments for line descriptions, units, quantities and contract references.

Taxonomy-Agnostic

Your taxonomy, or several.

Reclassify on demand, with no reimplementation and no consulting project.

Ingestion

Connect your systems, or upload files.

500+ prebuilt connectors, plus uploads of exports, spreadsheets, invoice PDFs and scans. Minimal day-to-day IT involvement.

Item Normalization

One description per item.

Inconsistent descriptions standardized for sharper category work.

Classification Review

A person stays in the loop.

Review, refine and accept. Every change keeps its provenance.

How Classification Works

You review a small set. It classifies the rest.

Proprietary AI learns from your decisions, then classifies every line through all four levels of your taxonomy.

Representative lines highlighted

Step 1 · Curate

A Representative Set

SpendCraft picks a concise sample that covers your spend.

Copper wire 12AWG

Cable ties, 100 pack

LED panel: moved to Lighting

Accept, or refine. Your call.

Step 2 · Review

Your Team Decides

Accept, or refine where it matters. Every change is recorded.

Classifying the full dataset

0%

L1

L2

L3

L4

Step 3 · Classify

Every Line, 4 Levels

Proprietary AI plus rules classifies the full dataset.

New data load

Auto-classified

98%

Waiting for review

24 lines

Classification Review keeps a person in the loop.

Step 4 · Keep Current

New Data, Same Standard

Each load is classified, with exceptions routed for review.

Running Today

Classification shouldn’t require a consultant.

Hours to first insight. Messy data is the starting point, not a blocker.

A global power equipment manufacturer

$0B

of spend classified.

See it on your own data.

A Proof of Value works one category or one moment end to end.