# Approved public metrics (single source of truth)

**Purpose:** Marketing, case studies, `/docs`, and homepage copy must align to this table.  
**Process:** Change numbers only after customer sign-off or documented measurement update.  
**Last reviewed:** April 2026

When updating the site, grep for old values and replace from here.

---

## Customer-attributed metrics

| Customer | Metric | Approved value | How measured / definition | Primary URL on site |
|----------|--------|----------------|---------------------------|---------------------|
| Pransform Inc. | Monthly document volume | 50,000+ documents/month | Operational throughput (BPO) | https://number7ai.com/case-studies/pransform-bpo-invoice-automation |
| Pransform Inc. | Client companies served | 300+ companies | Relationship count (stated by customer) | same |
| Pransform Inc. | AP staffing | 12 → 3 FTE | Headcount in AP capture/exception roles | same |
| Pransform Inc. | Staffing reduction | 75% | (12−3)/12 | same |
| Pransform Inc. | Straight-through processing (STP) | 90%+ | Documents completed extract → validate → export without human correction on that document, under customer approval rules | same + https://number7ai.com/docs/benchmarks |
| Pransform Inc. | STP prior baseline (for narrative) | ~60% | Industry-style mixed-AP benchmark for contrast; not Pransform’s measured “before” unless customer confirms | https://number7ai.com/docs/stp-self-healing-and-fair-benchmarks |
| Pransform Inc. | Field-level accuracy (production AP) | 99.5% | Per-field correctness on Indian AP invoices (not document-only pass rate) | https://number7ai.com/docs/benchmarks |
| Pransform Inc. | Post-extraction corrections (errors) | ~2,500/month → <250/month | “Error” = required correction after initial extraction (wrong/missing field, failed validation) | same |
| Pransform Inc. | Error volume reduction | ~90% | Relative change on same volume | same |
| Pransform Inc. | Payback period | Under 1 month | Customer-reported internal economics vs platform cost | same |
| JSK Automation | Pilot document count | 2,400+ documents | Pilot corpus | https://number7ai.com/case-studies/jsk-automation-purchase-order-processing |
| JSK Automation | Field-level accuracy (pilot) | 98.7% | Mixed channels: WhatsApp, email, scanned PDF | same + benchmarks |
| JSK Automation | Pilot turnaround | <72 hours | First document to full pilot report (stated narrative) | https://number7ai.com/docs/benchmarks |
| JSK Automation | Operating model | Zero dedicated data-entry headcount on automated path | Qualitative outcome; do not invent a % without sign-off | case study |
| Fairlorry | LR–invoice matching (before) | 2–3 days manual batch cycle | Operational SLA before automation | https://number7ai.com/case-studies/fairlorry-real-time-reconciliation |
| Fairlorry | LR–invoice matching (after) | Real-time / continuous | Automated matching as LRs arrive | same |
| Fairlorry | Billing leakage recovered | 3–5% of freight revenue | Reconciliation of weight, damage, rate errors previously caught late or missed | same + benchmarks |
| Fairlorry | Matching time reduction (case narrative) | 80% | From benchmarks / case narrative—keep wording “matching time” not generic “efficiency” unless customer approves broader claim | https://number7ai.com/docs/benchmarks |
| Fairlorry | Deployment duration | <2 weeks | Calendar time for 4-module custom layer (stated deployment) | case study + benchmarks |
| Fairlorry | Build shape | 4-module intelligence layer on existing TMS | Qualitative | case study |

---

## Aggregate / non-customer benchmarks (use with attribution)

| Topic | Value | Attribution | URL |
|-------|-------|-------------|-----|
| Mixed AP STP (industry ballpark) | ~60% | Third-party analyst research (e.g. AYR, Everest Group)—we do not control the estimate | https://number7ai.com/docs/benchmarks |
| IDP year-one ROI range (analyst) | 30–200% | AYR-style survey language | https://number7ai.com/docs/benchmarks |
| Document-type accuracy table | See benchmarks page | Derived from production + pilots; each row has basis column | https://number7ai.com/docs/benchmarks |

---

## Do not use on public pages without separate approval

- **“99.9%” as a default product claim** — replace with documented **99.5%** production AP (Pransform) or pilot **98.7%** (JSK), or neutral language (“high field accuracy”) unless a new measurement is approved and added here.

---

## PDFs (download CTAs)

| Customer | File path |
|----------|-----------|
| Pransform | `/case-study/pransform-bpo-invoice-automation-case-study.pdf` |
| Fairlorry | `/case-study/fairlorry-freight-reconciliation-automation-case-study.pdf` |
| JSK | _(add path when PDF is published)_ |