A spend taxonomy is a hierarchical classification system that organizes your procurement transactions into meaningful categories. It sounds simple. It's actually the foundation that determines whether your spend analysis dashboards deliver insight or confidently wrong answers.
The first step of data unification is building a spend taxonomy that works. Without one, you're flying blind. With a poor one, you're worse off—you're making decisions based on illusions. With a good one, you unlock the ability to see where your money actually goes, identify maverick spend, negotiate better contracts, and build a data culture that enables your whole team to drive procurement performance with data-driven insights.
What a spend taxonomy actually does
A spend taxonomy serves multiple purposes, and understanding each one helps explain why it matters so much.
Standardization. Your organization probably sources the same category of products and services through multiple suppliers, multiple cost centers, and multiple business units. Without standardization, you have no visibility into the total spend for any meaningful category. You can't see that you're buying the same software license from four different vendors, or that your total "Temporary Labor" spend—which might be classified as consulting in one division, staffing in another, and contract labor in a third—is actually your largest controllable expense. Standardization lets you see the elephant in the room.
Aggregation. This flows from standardization: once you classify spend consistently, you can aggregate it. You can answer questions like "What are we spending on cloud services across the whole company?" or "How much goes to women-owned suppliers?" or "What's our logistics spend trend over the last 18 months?" These questions are impossible without a taxonomy.
Supplier Intelligence. The data challenge gets harder when you zoom out to the supplier level. Take IBM. As Parekh notes in Suplari interviews: "Just like IBM can be International Business Machines or it can be IBM or it has subsidiaries, but you're dealing with one supplier. And just because that supplier is IBM doesn't mean you know if you're buying software or services from that supplier." A spend taxonomy lets you see the portfolio of products and services you're buying from a single vendor, which is essential for negotiation, relationship management, and risk assessment.
Compliance and Governance. Many industries require spend reporting by category—government contractors track small-business spending, healthcare systems track pharmaceutical purchases, and regulated industries track third-party risk across supplier segments. A taxonomy gives you the structure to meet these requirements.
AI Readiness. This is where Suplari's approach diverges from legacy platforms. Most procurement systems assume clean data as a precondition for AI. Suplari is built to work with the messy reality: fragmented, inconsistent data across ERPs, P2P systems, and spreadsheets. But even a platform designed to handle messy data performs better—more accurately, more quickly—when there's an underlying taxonomy structure that the AI can learn from and improve upon. The taxonomy becomes the foundation on which machine learning models classify transactions, and it gets better over time as the model learns from human corrections.
How to build a spend taxonomy that actually works
Here's a framework that avoids the pitfalls above.
Step 1: understand your current state before you design anything
Don't design in a vacuum. Spend time in the data and understanding the key needs of your stakeholders.
Run a data exctract on your last 12-24 months of transactions. Pull vendor names, amount, department, and—critically—whatever category codes or descriptions are already attached to the transactions. Look for patterns. What's already being classified? What's a mess? Are there clear natural groupings, or is classification truly random?
Talk to your procurement team and finance team. Not in a design meeting—ask them: "What questions do you need to answer about spend? What transactions confuse you? What would be useful to track that you can't right now?" A category that looks logical to the CFO but that no one in procurement actually cares about is a category that won't get maintained.
Also audit your systems. How does your ERP structure category codes? Does your P2P platform have category fields? What does your supplier master record? This step prevents you from designing a taxonomy that can't actually map to your operational systems.
Step 2: design a reasonable hierarchy
Most organizations find a three-to-four-level hierarchy works best:
Keep it simple. Your Level 1 segments should be 8-15 categories. Level 2 should expand to maybe 40-60 categories total. Level 3 is where you get to your operational granularity. Aim for 200-500 leaf categories—categories you'll actually use and maintain.
One critical design choice: how to handle services. Services are harder to classify than products because the same supplier often sells radically different things. A consulting firm might sell strategy, implementation, staffing, and training — all bundled into a single invoice. The traditional approach is to force a choice: pick one category and move on. A better approach is to classify by primary spend type and build a governance process for reviewing transactions that span multiple categories. This reduces the "forced choice" problem that makes classification feel arbitrary and ensures your highest-spend services categories get the attention they deserve.
Step 3: populate it iteratively, not perfectly
This is where Nikesh Parekh's insight applies directly: "The concept of getting your data right or data perfect is a mirage."
Start by manually classifying a representative sample of 500-1,000 transactions. This serves two purposes. First, it teaches you where your design breaks down—which is valuable feedback to refine the taxonomy. Second, it creates training data.
If you're using an AI-enabled platform like Suplari's Spend Classification product, that training data powers the model. The AI learns from your manual classifications and begins to classify the rest of your transaction history automatically. Unusual transactions — vendors with inconsistent naming, invoices with ambiguous descriptions, edge cases where the historical data is unclear — still require human review. That review isn't failure. It's the system learning. Each correction improves the model for future classifications, and the accuracy compounds over time.
The goal is to get to "good enough" in 90 days, not "perfect" in 12 months. A taxonomy that's 85% accurate and in use is infinitely more valuable than a perfect taxonomy that's still being debated.
Step 4: assign governance and stewardship
Your taxonomy will decay without governance. As new suppliers onboard, as products and services evolve, as the business changes, the taxonomy drifts.
Assign a data steward—typically someone in procurement or finance with strong operational knowledge and attention to detail. This isn't a full-time role (unless your organization is very large), but it should be a clear responsibility. The data steward's job is to:
- Review and approve category assignments for new suppliers
- Monitor classification accuracy over time (sample spend transactions quarterly)
- Update the taxonomy when it breaks down or becomes outdated
- Train new team members on proper classification
- Maintain the mapping between the taxonomy and your ERP and P2P systems
This governance prevents the "day two decay" problem—where a system works great on launch day, but by month two it's fallen apart because no one's maintaining it.
Why many procurement spend taxonomies fail
The first mistake is perfectionism. Too many organizations try to design their spend taxonomy in a conference room. Finance leadership gathers with procurement and IT, they spend weeks debating whether "software licenses" should be Level 2 or Level 3, and they emerge with a beautiful 47-page taxonomy document. Then they try to retrofit five years of messy transaction history into this idealized structure, the project stalls, and the taxonomy gets shelved.
The second mistake is over-complication. The United Nations Standard Products and Services Code (UNSPSC), managed by the United Nations Development Programme (UNDP), contains over 157,000 codes in its latest release (UNv24.0301). It's a four-level hierarchy (Segment, Family, Class, Commodity) with an optional fifth level, and it's the global standard for procurement classification. But here's the thing: no procurement team actually uses all 157,000 codes. In reality, most organizations overlay a custom taxonomy of 200-500 active leaf categories for their daily operations. You don't need perfection. You need utility.
The third mistake is treating taxonomy as a one-time project. Data, as Nikesh Parekh, Suplari's founder and advisor, points out, is "always going to be in a state where you have to keep optimizing it. The CPO who says 'I have bad data' is probably the CPO that will be fired." A spend taxonomy isn't a destination. It's a living system that evolves as your business changes, as you acquire new suppliers, and as you discover gaps in your original design.
The fourth—and most costly—mistake is building it in isolation from the systems that will actually use it. A taxonomy that looks good on paper but doesn't align with how your ERP structures data, or that doesn't map cleanly to your supplier master, or that your team finds too granular to apply consistently, is worse than no taxonomy at all.
The services spend classification challenge
One specific area deserves deeper attention: services.
Services spend is the hardest to classify consistently, and it's also the largest controllable expense for most organizations. You can buy the same software license from the same vendor every month and know it will be classified correctly. But services are heterogeneous. A single statement of work from an IT consulting firm might include strategy, implementation staffing, and training—each of which belongs in a different category.
Suplari's approach here reflects the practical reality. The Spend Classification product uses AI to categorize transactions, but services require more nuance than products. The key is designing your taxonomy to accommodate that complexity — specific services subcategories rather than catch-all buckets — and pairing it with a governance process that handles the ambiguous cases consistently rather than ignoring them.
The lesson: your taxonomy should account for this. Design your services categories to be as specific as possible (not just "Professional Services" as a catch-all), and build in a process for transactions that span multiple categories. A governance structure that acknowledges the complexity is stronger than a perfect hierarchy that ignores it.
From taxonomy to action
A well-designed spend taxonomy isn't the end goal—it's the entry point.
Once you have a solid taxonomy in place, it becomes the foundation for:
- Spend Analytics: You can actually see where your money goes, identify trends, and spot opportunities to consolidate or renegotiate.
- Supplier Intelligence: You understand the full portfolio of products and services you source from each vendor, which is essential for negotiation and relationship management.
- Compliance and Risk Management: You can report spend by category, track small-business or diverse-supplier spending, and monitor third-party risk across supplier segments.
- AI and Automation: Most importantly, you create the conditions for AI tools to actually deliver value. Clean, consistent data unlocked by a solid taxonomy is what allows machine learning to improve procurement decisions at scale.
Suplari's AI Data Platform is built on this principle. Rather than requiring customers to have perfect, pre-cleaned data before implementation, it starts with messy reality and unifies data across ERPs, P2P systems, and spreadsheets. It continuously cleans, classifies, and enriches transactions. And critically, the Spend Classification component learns from your specific taxonomy and corrections, getting smarter over time. Organizations can see value in 90 days instead of the 6-12 months typical with legacy tools that require data to be perfect before you start.
The real ROI of a spend taxonomy
The financial payoff is real, but it's not always dramatic on day one.
The immediate benefits are visibility and compliance: you can answer basic questions about where money goes, and you can generate the reports your stakeholders require.
The medium-term benefits come from aggregation: once you see total spend by category and supplier, you can negotiate better contracts, consolidate redundant vendors, and manage tail spend more actively.
The long-term benefits come from embedding this into your culture. A CPO who can articulate the company's spend by category, who knows the top suppliers in each category, and who has the data to support negotiation has a fundamentally different conversation with procurement teams and finance. They stop reacting and start strategizing.
And the AI benefit—the benefit that becomes available once your data is unified and your taxonomy is sound—is the ability to continuously monitor and optimize spend, identify risks before they become crises, and make better sourcing decisions based on data rather than legacy relationships or habit.
Get started, improve over time
You don't need permission to start. You don't need perfect data or flawless systems. You need three things:
- Clarity about why you're doing this (what questions do you need to answer?)
- Access to transaction-level spend data
- Commitment to iterate rather than perfect
Map your current state. Talk to your team about what they need to see. Design a reasonable three-to-four-level hierarchy with 200-500 leaf categories. Pick a representative sample of transactions and manually classify them. Use that as training data to automate the rest.
By month three, you'll have a working taxonomy. By month six, it will be informing decisions. By month twelve, it will be the foundation of your procurement intelligence practice.
That's a spend taxonomy that works.
