Procurement Data Foundations: The Minimum Data Model That Supports Better Decisions
Procurement data foundations are the core records, definitions, ownership and quality controls that let a team understand suppliers, contracts, spend, commitments and performance. The goal is not a perfect data lake; it is sufficiently trusted data for the decisions procurement must make.
TL;DR
Start with the decisions that need better data, not a generic data-cleanup project.
Clarify ownership of key records: supplier, contract, category, PO, invoice, receipt and budget.
Improve data at the point it is created, then monitor the exceptions that matter.
The core procurement data objects
Record | Why it matters |
|---|---|
Supplier | Supports onboarding, risk, payment, spend visibility and relationship management |
Contract | Connects obligations, pricing, term, owner and approved buying route |
Category | Groups demand and spend for strategy, sourcing and analysis |
Request and PO | Records demand, authority, commitment and transaction control |
Receipt and invoice | Supports payment, matching and operational visibility |
Cost centre, entity and project | Connects a purchase to financial ownership and planning |
How to improve procurement data
Define the decision or workflow that poor data is preventing.
Name the system of record and accountable owner for each key object.
Set a practical minimum standard for the fields and classifications that matter.
Improve data capture in the working process, not only through periodic clean-up.
Measure exceptions, duplicates and missing fields that affect the intended decision.
Review ownership when systems, suppliers or processes change.
Common data problems
Duplicate suppliers and confusion between legal entity, trading brand and service provider.
Categories that do not reflect the market, demand or commercial ownership.
Contracts that are signed but not connected to purchasing behaviour.
Data copied between systems with no agreed source of truth.
Teams measuring data completeness without knowing whether the data is usable.
Data governance vs data quality
Data quality describes whether the records are accurate, complete, current and usable for a purpose. Data governance describes how the organisation assigns ownership, standards, decisions and controls to keep them that way. Both are needed.
Frequently asked questions
Does procurement need its own data platform?
Not necessarily. Procurement may use ERP, P2P, CLM, supplier-management and analytics systems. The priority is clear system ownership and reliable hand-offs, not a separate platform for its own sake.
Which data should be fixed first?
Fix the records that affect a material decision or workflow: supplier identity, contract linkage, category, financial ownership or PO/invoice matching are common starting points.
Who owns procurement data?
Ownership is distributed. Procurement owns its requirements and many business definitions, while finance, IT, data teams and other functions may own source systems and controls.
A practical exercise from the data-quality conversation.
The archive interview with Susan Walsh, The Classification Guru, centres on spend classification, taxonomies and accuracy. Start your own data improvement with a representative sample rather than promising to clean every record at once. Include routine suppliers, ambiguous descriptions and records the team finds difficult to classify.
For each record, check the supplier entity, amount, currency, period, category and source. Distinguish a confirmed fact from an inference. “Services” may not provide enough information to classify a purchase. Resolve the uncertainty using the relevant document or business owner, and record why the final classification was chosen.
Write the rule so someone else can apply it.
Create a small classification guide with examples, exclusions and an escalation route. Explain what a category includes and how to handle purchases spanning several categories. Consistent definitions matter more than forcing every uncertain record into a category quickly.
Ask a second person to apply the guide to the same sample. Compare disagreements and find whether they arise from unclear rules, missing source material or inconsistent judgement. Update the guide and repeat the check. This gives you practical evidence of consistency without pretending the entire dataset is already reliable.
Connect the repair to a procurement decision.
Use the cleaned sample to test a question the business cares about. It might be supplier concentration, fragmented purchasing or a category opportunity. Explain the coverage and remaining limitations before presenting the result. An accurate sample can support a bounded conclusion while still requiring further work before a portfolio-wide claim.
Assign ownership for the data after the exercise. Decide where corrections are made, who approves category changes and how new records enter the process. A one-off cleanup decays if the original intake continues to create the same errors.
Keep a log of recurring issues and their causes. When incomplete purchase descriptions create most of the uncertainty, improve the request or transaction process. When identifiers are inconsistent across systems, work with the relevant data owner. Cleaning at the reporting stage can help, but preventing the error can remove repeated work.
The useful outcome is a decision the team can defend and a process that maintains the supporting evidence. A larger dataset or more elaborate dashboard does not compensate for an unclear source or a classification rule nobody can explain.