As part of our ‘What does good look like series,’ we turn to analytics. Analytics has always been present in procure to pay (P2P), but for most of the platform lifecycle, it has functioned more as a reporting layer than an execution layer. Historically, P2P analytics answered a single primary question: whathappened. Dashboards and reports showed spend by category, invoice cycle times, approval bottlenecks and compliance rates. This visibility mattered for audits and reviews, but it rarely influenced decisions while processes were still in motion.
The limitations were timing and coupling. In most P2P environments, analytics sat outside execution. Insights were produced after transactions were completed, when routing, approvals and payment behavior were already fixed.
As transaction volumes increased and process complexity grew, this gap became more visible. Knowing that invoices were paid late did not prevent late payments, that exceptions were high did not reduce them; that suppliers were risky did not necessarily change routing, approvals or payment behavior.
More mature P2P execution reframes analytics as an operational input, instead of a reporting layer.
The first step in that shift is moving from descriptive to diagnostic analytics. Instead of only reporting KPIs, platforms begin to surface why those KPIs look the way they do. For example, rather than showing average approval time, analytics identify which approval paths, roles or conditions consistently cause delays. Rather than listing exception rates, they show which combinations of suppliers, document types and rules generate rework.
This still looks like reporting, but it changes the conversation from ‘what is wrong’ to ‘where intervention matters.’ The next step is predictive awareness.
Predictive analytics in P2P does not mean forecasting the future in a general sense. Rather, it means anticipating likely outcomes within active processes, such as predicting which invoices are likely to miss payment terms, which requisitions are likely to stall in approval or which suppliers are likely to generate disputes based on current patterns. When these signals are surfaced early enough, teams can act before the outcome is locked in.
Prediction alone is not sufficient, though. The real inflection point comes when analytics begin to recommend actions and influence execution paths. This is where many platforms struggle.
Prescriptive analytics in P2P requires tight coupling between insight and control. If analytics identify a high-risk transaction, the system must be able to route it differently, apply additional validation or prompt a specific decision. If analytics detect a low-risk, repetitive transaction, the system must be able to reduce friction without manual intervention. What matters is making the default path smarter.
Another important shift is the scope of analytics. Traditional P2P reporting focuses on internal performance. More advanced execution includes external signals, such as supplier behavior trends, payment acceptance patterns, dispute frequency and network-level benchmarks. These signals help organizations understand whether an issue is local or systemic and whether performance is improving relative to peers or just internally.
Generative AI has added a new interface to analytics. Natural language queries and conversational reporting can reduce friction in accessing data. They help users explore questions without predefined dashboards. This is not a new foundation, though, nor do they replace the need for well-structured metrics, clear definitions and trusted data lineage. Gen AI is effective in P2P analyticsonly when it sits on top of a coherent analytical model. When it does not, it becomes just another way to retrieve the same static reports.
Organizations that extract real value from analytics ask different questions. They focus on where analytics should intervene rather than merely inform, which decisions recur often enough to benefitfrom guidance and which signals should automatically change routing, validation or prioritization. Analytics maturity is not measured by the number of dashboards or AI features. Analytics maturity ishow often insights change what happens next.
The final article of this series will synthesize these execution layers and describe what ‘good’ looks like today, grounded in observable behavior, not aspirational roadmaps, and what that level of maturity actually enables next.
