What is Straight-Through Processing (STP) Rate?

Last reviewed: 2026-08-12 · Applies to: document automation / AP workflows / payments / insurance claims

Scope: This definition covers Straight-Through Processing (STP) Rate as used in document data extraction and back-office automation — measuring what share of invoices, forms, and other documents complete their full processing cycle without any human touch. It does NOT cover the broader concept of process automation software, robotic process automation (RPA) platforms, or STP in the securities-trading sense beyond a brief historical note.
Also known as: Sometimes called "touchless processing rate," "touchless rate," or "automation rate." In healthcare claims, the equivalent concept is called the "auto-adjudication rate."

Straight-Through Processing (STP) Rate is the percentage of documents or transactions that complete their entire processing cycle — ingestion, classification, extraction, validation, and posting — without any human intervention. It is the single metric that most directly maps to labor cost, because every document that does not pass straight through requires paid human time.

How STP Rate Works

STP rate is calculated with one formula, consistent across every industry that uses it — IBM defines it simply as "the percentage of transactions that are processed automatically" (IBM Think):

STP rate = documents processed touch-free ÷ total documents received × 100%

The numerator counts only documents that passed through every stage — ingestion, classification, extraction, validation, and posting — without a human clicking "approve," correcting a field, or resolving a mismatch. The denominator is every document that entered the system, including the ones that failed and got routed to a human. OpsDog's definition of the healthcare equivalent, auto-adjudication, adds an important edge case: a claim that was auto-processed but later found to have errors requiring manual resolution must be excluded from the numerator (OpsDog). STP counts documents that stayed straight-through all the way — not documents that appeared automated and then needed fixing.

What counts as "touch-free" is deliberately strict. A document that is automatically extracted but then manually reviewed is not STP. A document that passes extraction and validation but sits in an approval queue awaiting a manager's click is not STP. Only zero human touchpoints anywhere in the cycle counts. This is why the STP rate is always lower — often dramatically lower — than the "accuracy" numbers vendors quote: accuracy measures how often characters or fields are read correctly, while STP measures whether the entire document flowed through end-to-end with no human involvement at all.

Why STP Rate Matters

STP rate is the only widely-benchmarked metric that directly converts into labor hours. Ardent Partners' 2025 AP metrics report (n=212) puts the industry-average touchless rate at 32.6% — meaning roughly two-thirds of all invoices still touch a human somewhere in their journey (Ardent Partners, 2025). Every one of those touched invoices costs real time: IOFM's benchmarks put the manual touch on a typical invoice at 12.5 minutes, with exceptions adding another 15–45 minutes on top (IOFM, 2024).

The spread between teams is the real story. Ardent's Best-in-Class organizations hit 49.2% touchless — a 2.1x advantage over the 23.4% achieved by all other firms, driving invoice processing costs of $2.78 per invoice vs $9.40 for the average organization (Ardent Partners, 2025). The Hackett Group finds AP-solution adopters average 60% touchless rates, and teams above 30% touchless are 3.5x more productive (The Hackett Group, 2025). AI-era deployments push higher still — analyst figures describe high performers in the 60–80% range (Levvel Research, via Ascend, 2025), while vendor-reported production numbers reach 85–92% for agentic platforms (Hypatos, vendor-reported).

STP rate is also where marketing claims get fuzzy. Vendors routinely quote "99% accuracy" without specifying what was measured — characters, fields, or documents — and none of those numbers is the STP rate. A system can have 99% character accuracy and still route documents to a human, because a single misread character corrupts an entire field (see the field accuracy versus character accuracy reference). The STP rate is the number that shows up in your department's headcount — which is why it matters and why vendors so rarely publish it.

History and Evolution of STP

STP was born in financial settlement, not document processing. The first efforts took shape in the 1970s, when the Federal Reserve Bank of San Francisco introduced the first Automated Clearing House (ACH) for payroll direct deposit and SWIFT (founded 1973) built the global messaging network for cross-border bank transfers — both eliminating the manual re-entry of payment instructions that had previously been telegraphed or keyed by hand (Investopedia; Routable). In the 1980s–1990s, securities trading and insurance claims adopted the same logic: automate settlement and adjudication so transactions flow without manual touchpoints.

The 2010s brought STP to documents. Ardent Partners has tracked "touchless" invoice processing since its earliest AP research, defining it as "an invoice being received, approved, and payment scheduled without any manual intervention on the part of the AP team" (Ardent Partners, 2020). By the 2020s, AI-era extraction platforms made much higher rates reachable: where template-based OCR topped out around 10–20% STP, modern AI document processing reports 60–75%, and agentic platforms claim 85–92% in production (Hypatos, vendor-reported, 2019/2026). The metric itself is unchanged since the 1970s; what changed is how much of the document journey can actually be automated.

Where STP Rate Is Used

STP rate is measured differently in different industries, though the formula is the same. The definition narrows depending on what "complete processing" means in that domain:

  • Accounts Payable (the dominant use today): STP means an invoice is received, extracted, validated against purchase orders and receipts, approved, and posted — all without human action. This is where the best benchmark data lives (Ardent 32.6% average, Hackett 60% among adopters).
  • Payments and financial settlement (the origin): STP means a transaction flows from initiation through validation, enrichment, routing, and settlement without manual intervention — the basis of ACH, SWIFT, and modern payment rails (IBM Think). In cash application — matching incoming payments to open invoices — industry-average STP runs 60–80% on rule-based platforms, rising to 95%+ with AI-native matching (Transformance).
  • Insurance and healthcare claims: STP (called auto-adjudication) means a claim is validated, adjudicated, and paid by rules and automated systems. Baseline STP sits at 5–10%, rising to 20–40% with automation, and 40–60% for specific low-severity claim types at leading carriers (Peakflo, 2026). The industry benchmark target hovers around 80% auto-adjudication, though actual first-pass rates range from 10% to 70% (HCIM).

Common Misconceptions

  • Misconception: "STP rate is the same as OCR accuracy."
  • Reality: OCR measures how many characters were read correctly; STP measures whether an entire document completed end-to-end without human involvement. Tungsten Automation frames the division cleanly: "OCR determines the 'what' when it comes to data capture, while STP addresses the 'what next'" (Tungsten Automation, 2025). A system can read 99% of characters correctly and still have a low STP rate, because extraction is only one of the five stages that must all pass touch-free.
  • Misconception: "High extraction accuracy guarantees high STP."
  • Reality: Extraction accuracy is a necessary but not sufficient condition. STP additionally requires correct document classification, successful validation against reference data (purchase orders, vendor masters), matching, and ERP integration — any of which can route a document to a human even when every field was extracted correctly. Finofo's implementation guide is blunt: "Achieving it at scale requires more than good OCR" — it also depends on vendor master hygiene, PO coverage, and tolerance settings (Finofo, 2026). A 99% extraction rate on a 15-field invoice still leaves more than one in three documents with at least one error — and each such document needs human review.
  • Misconception: "100% STP is achievable — automation means no humans."
  • Reality: True 100% STP is rare in production. LlamaIndex's glossary notes most organizations achieve between 26% (international transactions) and 67% (best-in-class), because "exceptions, edge cases, and scenarios the rules engine can't handle" always remain (LlamaIndex). Poor-quality scans, novel document formats, ambiguous data, and fraud flags will always exist; the honest target is pushing STP as high as economically sensible and routing the remainder to well-designed human review — not eliminating humans entirely.

Frequently Asked Questions

What is a good straight-through processing rate?

A "good" STP rate depends on your baseline: the industry average in AP automation is 32.6%, best-in-class teams reach 49.2%, AP-solution adopters average 60%, and AI-era high performers report 60–80% (Ardent Partners 2025; Hackett 2025; Levvel). Vendor-published figures of 80–92% are real but reflect controlled, well-prepared document populations — treat them as upper bounds, not typical performance.

How do you calculate STP rate?

STP rate = documents processed touch-free ÷ total documents received × 100%. Only documents that pass through every stage — ingestion, classification, extraction, validation, and posting — with zero human intervention count in the numerator; everything that enters the system counts in the denominator (IBM Think). Documents auto-processed but later corrected by a human must be excluded from the numerator (OpsDog).

What is the difference between STP rate and exception rate?

They are near-inverses with different denominators. Exception rate measures the share of documents that fail validation or matching and need special human handling; STP rate measures the share that complete entirely touch-free. A document can be touched by a human (approval workflow) without being an exception, which is why STP and exception rates do not add to 100% (Precoro). The two together describe the automation boundary: exceptions are the documents that break the flow, STP is the flow that never broke.

Is STP rate the same as automation rate?

No. Automation rate describes how much of the process is automated; STP rate measures whether a document actually completed end-to-end with no human touch. A workflow can be 100% automated and still route every document to a human for approval or exception handling — yielding an STP rate near zero. Hyperscience makes the point directly: "automation rate doesn't account for time spent handling exceptions," and an STP rate of 100% offers no value if total handling time isn't faster (Hyperscience).

How do you improve your STP rate?

Attack the stages that break the flow, in order. (1) Fix data quality upstream — clean vendor master data and high PO coverage raise the ceiling of every downstream stage (Finofo's implementation guide calls vendor master hygiene "the precondition for STP"). (2) Upgrade extraction from template-based OCR to AI that handles layout variation, since template-era systems cap STP around 10–20% while AI platforms report 60–75% (Hypatos, vendor-reported). (3) Tighten validation and matching rules so that valid documents stop getting flagged. (4) Measure STP by document category — PO-backed vs non-PO, new vs returning vendors — because each category has a different ceiling (Finofo).

What is straight-through processing in document extraction?

In document extraction specifically, STP means the extracted data flows directly from the AI/OCR engine into your downstream system — ERP, accounting, or workflow — and is posted without a human reviewing or correcting it. It is the end-to-end completion rate of extraction, the metric that determines how many processing FTEs you actually need. It sits one level above document-level accuracy: a document can be 100% accurately extracted and still not be STP if a human must review it before posting.

Sources

  1. Ardent Partners — "AP Metrics That Matter in 2025". n=212 AP professionals; the definitive source for touchless/STP benchmarks: 32.6% average, 49.2% Best-in-Class, 23.4% all others, $9.40 vs $2.78 cost per invoice. Primary source for all AP STP benchmark figures.
  2. IBM Think — "What Is Straight-Through Processing?". IBM's official definition: "The STP rate refers to the percentage of transactions that are processed automatically," plus the full stage breakdown (initiation, validation, enrichment, routing, settlement). Primary source for the definition and formula.
  3. Investopedia — "Understanding Straight-Through Processing (STP)". Historical source: 1970s ACH and SWIFT origins, payroll direct deposit, and the evolution from telegraphic transfers. Primary source for the history section.
  4. The Hackett Group — AP Solutions Research (2025). Consulting benchmarks: 60% average touchless among AP-solution adopters; teams at 30%+ touchless are 3.5x more productive. Corroborates the 60% AI-era anchor.
  5. IOFM — AP Benchmarking / Ask the Expert (2024). Industry association data: 12.5-minute average manual touch per invoice, 15–45 minutes additional per exception. Source for the labor-cost conversion.
  6. Levvel Research (now Deloitte), via Ascend (2025). Analyst data: industry touchless ~30% vs high performers 60–80%. Source for the 60–80% high-performer range.
  7. Hypatos — Document AI (2019, updated). Vendor-reported generational comparison: template-era 10–20% STP, modern AI 60–75%, agentic platforms 85–92%. Vendor-reported; used only for the automation-generation framework.
  8. Finofo — "Straight-Through Processing in AP: A Practical Guide". Implementation experience: realistic STP progression (30–45% go-live to 75–85% best-in-class) and the dependency on vendor master hygiene, PO coverage, and tolerance settings. Source for the "more than good OCR" argument.
  9. Tungsten Automation — "How Advanced OCR Powers STP" (2025). Framework: OCR captures the "what," STP handles the "what next" (validation, business logic, routing). Source for the OCR-vs-STP distinction.
  10. Transformance — "What is Straight-Through Processing?" (2026). Glossary: STP in cash application, formula (auto-matched & posted / total), industry-average 60–80% STP on rule-based platforms. Source for the cash-application variant.
  11. Peakflo — "Automated Insurance Claims Processing". Industry KPI table: claims STP baseline 5–10% → 20–40% automated, leading carriers 40–60% for specific claim types, manual 23 days vs STP-eligible under 24 hours. Source for the insurance-claims variant.
  12. OpsDog — "Claims Auto-Adjudication Rate" KPI. Benchmarking firm: definition of auto-adjudication and the strict rule that auto-processed-then-corrected claims are excluded from the numerator. Source for the touch-free definition edge case.
  13. LlamaIndex — "How Straight Through Processing Works in Practice". Glossary: real-world STP range 26% (international) to 67% (best-in-class); true 100% STP is rare. Source for the 100% STP misconception correction.

Related reading: How to test document extraction accuracy · Field-level vs straight-through accuracy · When to switch from OCR to AI extraction

📮 contact email: [email protected]