The P2P data problem hiding in plain sight that could kill future efficiency

What looks like a simple data housekeeping issue – multiple records about the same supplier, in many systems – will have a big impact across your entire P2P operation. 

When we unearthed the same supplier listed over 100 times in one organisation’s ERP, AP automation, and procurement systems, it wasn’t just a data quality problem. It was a gateway to millions in lost negotiating power, overpayments, and operational inefficiency. The financial stakes are high, and the time to act is now.

This isn’t an isolated incident. It’s a pervasive problem plaguing every Procure to Pay system architecture, in every organisation. The sooner we acknowledge this, the sooner we can start working towards a solution.

When twenty suppliers are actually one

Consider this scenario: A major global supplier had over 20 different codes across different ERP instances within a single national buyer organisation. A pattern that repeated across multiple countries. The result?

• Invisible spend concentration: Twenty small ($250K) tactical vendors instead of one ($5M) strategic relationship
• Lost leverage: Twenty small contracts instead of one strategically negotiated procurement agreement
• P/O chaos: Purchase orders for the same supplier don’t match because they’re from different systems
• Payment term lottery: Different payment terms for the same goods, services and supplier across divisions, with no ability to optimise

The $1M overpayment that nobody noticed

In one particularly painful example, we found a single supplier with four different accounts across sites and ERP systems for the same buyer entity. Invoices were loaded on one account. Credit notes on another.

The damage over 12 months:

• $10M in payments processed
• $0.5-1M in credit notes sitting unused on different accounts
• Result: Up to 10% overpayment that went completely undetected

This wasn’t fraud. It wasn’t incompetence. It was the inevitable consequence of a fragmented data architecture.

Each acquisition makes it exponentially worse

If fragmentation is bad in a steady state organisation, it becomes catastrophic during M&A:

The day-one disaster:

• A newly acquired company introduces 10,000 suppliers
• 6,000 are duplicates you can’t identify
• Systems don’t consolidate
• Synergies evaporate into integration costs

The hidden integration costs: Post-merger integration often costs $10-50M, of which 60% is focused on “data harmonisation.” This problem is “solved” using armies of consultants to manually map the data instead of semantic matching. What should take months takes years. What should cost millions costs tens of millions.

The foundation that determines your future

To successfully deploy AI in the Procure to Pay space, we’ve built tools that produce unified, clean data. Without this foundation, the promise of AI-powered procurement remains just that – a promise.

The reality for most enterprises: AI initiatives stall not because the technology isn’t ready, but because the underlying data is too fragmented to support even basic automation. Models trained on duplicate suppliers and inconsistent data produce unreliable results that no CFO would trust for critical decisions.

The gap between organisations with harmonised P2P data and those without it isn’t just about efficiency anymore – it’s about competing in an AI-driven future. Every day spent manually reconciling data is another day lost for building the intelligent automation defining procurement leadership in the next decade.

It’s about having the fundamental capability to leverage AI when it matters. And increasingly, it matters for everything from risk prediction to payment optimisation to strategic sourcing.

The AI imperative: Why this matters more than ever

Traditional approaches to fixing P2P data fragmentation have failed because they treat the symptoms, not the cause. The fundamental issue is that P2P data is inherently about connections – between systems and between the contextual information about each supplier’s contracts, invoices, payments, entities, and hierarchies. These relationships need a new type of architecture designed specifically for connected data.

Step 1: A harmonised foundation – Creating consistency from chaos

Before any matching or analysis can begin, data from disparate systems should be mapped to a common schema by:

• Standardising field definitions (what “payment terms” means is consistent everywhere)
• Tagging data elements with metadata for lineage tracking and change detection
• Implementing continuous monitoring to identify updates, additions, or deletions in real-time
• Creating audit trails that maintain data integrity while enabling rapid processing

This harmonisation layer acts as your universal translator, ensuring that downstream processes work with consistent, well-structured data regardless of source system variations.

Step 2: Semantic matching – Teaching systems to think smarter

Traditional matching looks for exact text matches. Semantic matching understands meaning; for example, “IBM UK Ltd”, “International Business Machines (GB)”, and “IBM UK” should be recognised as the same entity. This is achieved by:

• Using context like addresses, tax IDs, and business relationships
• Applying ML and data enrichment to improve confidence scores
• Going beyond string matching to understand business reality

Step 3: Intelligent blocking – Making the impossible possible

Comparing every record to every other record would require billions of computations. Blocking makes it manageable by grouping records into smart “blocks” of candidates which share, for example, the same postcode or similar names. This helps because it:

• Reduces noise while preserving match potential
• Makes real-time matching computationally feasible
• Scales to millions of records without breaking

Step 4: The Unified Semantic Graph – The New P2P Brain

Once matched, entities live in a connected graph structure:

• Supplier nodes connect to addresses, contacts, siteIDs etc
• Relationships become queryable and actionable
• Complex questions get simple answers
• Hidden patterns become visible

Now you can instantly answer: “Which suppliers are high-risk, concentrated across business units, and tied to late payments?”

These steps represent a fundamental shift in how enterprises should approach P2P data, from managing records to understanding relationships. 

The choice is yours

Every day you delay addressing P2P data fragmentation, you’re:

• Leaving millions in working capital optimisation on the table
• Falling further behind AI-enabled competitors
• Burning out your best talent on mundane, manual reconciliation
• Making future acquisitions more difficult and costly
• Overpaying suppliers by 5-10%
• Increasing compliance risk

The question isn’t whether to fix your P2P data foundation. It’s whether you’ll do it before these hidden costs become visible crises.

That number you’re looking at? It’s not an IT problem. It’s your business case for transformation.