Pricing
ROI Analysis · Computed from your live data

See what OmniPATH saves you. In hard numbers.

Processing cost reduction. Time saved. Duplicates caught. Match rates improved. Every ROI metric computed from your live platform data - not a spreadsheet estimate, not a sales-deck projection.

£1 vs £17.50
Cost per document
12 min
Manual baseline
Live
Every metric
ROI · May 2026
Live
Saved vs manual
£13,893
+14% vs Apr
Time saved
148h
+12% vs Apr
Duplicates caught
6 grps
£8,420 at risk
3-way match rate
94.7%
+0.4pp vs Apr
Cost per document · 842 docs this month£13,893 saved
Manual
£17.50
OmniPATH
£1.00
Touch-free rate · 6m
72%+4.2pp
The numbers behind the platform

Cost. Time. Risk. Quantified.

The four numbers every finance leader asks about - answered with platform data, not estimates. Each metric ships with period-over-period comparison built in.

£1 vs £17.50
Cost per document
OmniPATH cost vs the manual-processing benchmark from Ardent Partners research
[methodology]
12 min
Manual baseline
Per-document processing time assumed for the manual workflow - receipt to approval
Live
Every metric
All ROI figures computed from your actual processing data - not modelled or estimated
Δ%
Period-over-period
Every stat ships with previous-period delta - measure improvement, not just totals
Eight ROI metrics

Eight angles on "is this worth it?"

Cost savings, time saved, accuracy, automation rate, duplicate prevention, match rate, pipeline throughput, anomaly detection. Each one a clear number with a clear source - drawn from the processing analytics and financial analytics modules already running.

Metric 01
Processing cost savings

Total documents processed × £17.50 manual benchmark, less the platform processing cost at £1.00 per doc. Net savings converted to your org currency at the posting-date FX rate.

500 docs → £8,750 manual · £500 OmniPATH · £8,250 saved
Metric 02
Time saved

Wall-clock processing time vs 12 minutes per doc manual baseline. Returned as both a time_saved_seconds integer and a formatted HH:MM:SS string. Wall-clock accounts for parallel processing - doesn't double-count.

500 docs × 12 min = 100h manual · 8h actual · 92h saved
Metric 03
Model accuracy

3-signal composite - verified (human-reviewed gold standard) → confidence (AI self-assessment) → quality (pipeline assessment). Uses the highest-priority signal available, returns method so you can see which.

94.2% accuracy · verified on 128 documents
Metric 04
Touch-free rate

Percentage of invoices that processed without any human edits - touch-free vs single-touch vs multi-touch. Every touch-free document is 12 minutes of finance time eliminated.

72% touch-free · +4.2pp vs last month
Metric 05
Duplicate invoice exposure

Number of duplicate invoice groups detected, plus the total_potential_value at risk. Per-cluster detail with invoice number, supplier, occurrence count and group total - actionable, not just a count.

6 groups · £8,420 in prevented double-payments
Metric 06
Three-way match rate

PO vs GRN vs Invoice match rate. Returns fully_matched, partial_matched, unmatched counts plus the match_rate_pct. Every unmatched invoice is manual investigation work avoided.

94.7% fully matched · 12 unmatched need review
Metric 07
Pipeline performance

Operational throughput of the AI processing pipeline - throughput_per_hour, latency_p50 / p95, processing_p50, avg_queue_time, worker_utilisation. The "is the platform keeping up?" answer.

2.1s p50 · 4.8s p95 · 72% worker utilisation
Metric 08
Anomaly & fraud detection

Clean-vs-flagged document split, plus separate trend endpoints for anomaly rate and fraud rate over time. The flagged-document count - the things a human reviewer would have missed at scale.

3.2% anomaly rate · 27 flagged this month
Metric 01 deep-dive

From £17.50 to under £1. Per document. [methodology]

The £17.50 manual benchmark is the industry-standard figure for end-to-end AP invoice processing - receipt, data entry, coding, matching, approval routing - sourced from Ardent Partners and IOFM research. The £1.00 OmniPATH cost is the actual platform processing cost per document. The maths is straightforward; the savings compound fast.

  • Hardcoded benchmarks · £17.50 manual / £1.00 platform · industry-standard, defensible, consistent across orgs
  • Multi-currency · ECB FX rate at processing date · totals reflect what the cost was on the day, not today
  • Computed per-query · pulled from /v0/analytics/performance on every call · always reflects the live data
  • Filterable · scope by site, area, team, supplier, category, date range · "what's our ROI on FS Foodservice specifically?"
Cost savings · 842 documents · May 2026
Documents processed
842
Manual benchmark · £17.50 × 842
£14,735.00
OmniPATH cost · £1.00 × 842
−£842.00
Net savings
£13,893
Source · Ardent Partners / IOFM research on manual AP processing cost. £17.50 figure includes receipt, data entry, GL coding, three-way matching and approval routing labour.
Metric 03 deep-dive

Three signals. One accuracy number.

Model accuracy is hard to express with a single number - every signal has a different cost and confidence. The cascade resolves this: verified (human-reviewed) is the gold standard and used wherever available; confidence falls back when verification is sparse; quality is the last-resort pipeline assessment. The response always tells you which signal was used.

  • Verified score · accuracy on human-verified docs only · highest priority · most expensive to produce
  • Confidence score · average AI self-assessment across all processed docs · medium priority · always available
  • Quality score · pipeline's own quality assessment · lowest priority · safety-net for when no review has happened
  • Method declared · response includes method: "verified" | "confidence" | "quality" · so you know what's behind the number
Accuracy cascade · priority order 94.2%
1
Verified score
Human-reviewed gold standard · 128 docs scored
94.2%In use
2
Confidence score
AI self-assessment · 842 docs available
91.8%Fallback
3
Quality score
Pipeline assessment · 842 docs available
88.4%Last resort
method: "verified" · 128 of 842 docs reviewed
Period-over-period

Every stat. With its delta.

Six trend fields ship on every processing-stats response - totals matter, but the direction-of-travel matters more. Each card surfaces the _diff_pct against the previous matching period.

Total processed
total_processed_diff_pct
842
+18% vs Apr (713)
Touch-free rate
touch_free_diff_pct
72%
+4.2pp vs Apr (68%)
Avg processing time
avg_processing_time_diff_pct
2.1s
−8% faster than Apr
Cost saved
cost_saved_diff_pct
£13,893
+14% vs Apr
Time saved
time_saved_diff_pct
148h
+12% vs Apr (132h)
Model accuracy
accuracy_diff_pct
94.2%
+1.1pp vs Apr

Plus six trend-chart endpoints for time-series visualisation: touches, anomalies, doc-types, model performance, approvals, fraud.

Before vs after

From "we think it's working" to "here are the numbers".

Most procurement and AP automation tools never get measured. The team feels faster, the spreadsheet shows fewer late payments, but nobody can quantify the actual ROI. OmniPATH ships with the answers built in.

×
Without OmniPATH ROI Analysis
The unmeasured platform
  • Cost per invoice unknown - no measurement of manual processing effort
  • Time saved is a guess - "we think we're faster" with no data
  • Duplicate invoices discovered months later - or never
  • Match rates unmeasured - three-way matching done manually or not at all
  • Anomaly detection relies on human vigilance - things slip through
  • No period-over-period comparison - no way to measure improvement
  • Model accuracy unverifiable - "the AI is doing its job" with no proof
With OmniPATH ROI Analysis
Quantified, dated, defensible
  • Cost per document tracked: £1 OmniPATH vs £17.50 manual benchmark
  • Time saved computed from real wall-clock processing vs 12-min manual baseline
  • Duplicate exposure quantified - group count plus £ value at risk
  • Three-way match rate measured: fully matched, partial, unmatched with %
  • Touch-free rate tracked - % of invoices processed without human intervention
  • Model accuracy with 3-signal breakdown: verified, confidence, quality
  • Every metric ships with period-over-period _diff_pct for trend tracking
  • Filterable by site, team, supplier, category, date range
LEDGE × ROI

Ask LEDGE about your savings.

Every ROI metric is a LEDGE-accessible endpoint. "How much have we saved this month?" returns the figure, the chart and the period-over-period comparison - without you opening a single dashboard.

"How much have we saved this month?"
"What's our touch-free rate this quarter vs last?"
"How many duplicate invoices were caught this year?"
"What's the total £ value of duplicates prevented?"
"Show me processing time trend over 6 months."
"What's our three-way match rate?"
"How has model accuracy changed since we started?"
"Compare processing cost per site."
Frequently asked

The practical details.

Where does the £17.50 manual benchmark come from?
It's an industry-standard figure for end-to-end manual AP invoice processing - receipt, data entry, GL coding, three-way matching and approval routing - sourced from Ardent Partners and IOFM research. The number is a labour-cost estimate per invoice across mid-market organisations. Different research firms quote slightly different ranges (£12–£25 is common); we've taken the £17.50 mid-point as a defensible default. The savings figure is genuinely conservative for organisations with above-median AP labour costs and genuinely aggressive for organisations running highly streamlined manual workflows.
Can I change the £17.50 / £1.00 benchmarks to match my own labour cost?
Not currently - both benchmarks are hardcoded so the savings figure is comparable across orgs running the same platform. Per-org configurable cost benchmarks would let each org dial in their own labour rate, which is appealing on the surface, but it also makes the savings number unverifiable in any external context. We've taken the position that hardcoded defensible benchmarks are more useful than per-org tuned ones. If your internal labour cost is meaningfully different, you can recompute on top of the raw docs_processed figure for an internal report.
Is there a single ROI dashboard that puts all eight metrics on one page?
Honest answer - no, not as a single bundled dashboard endpoint today. The eight metrics live across the processing-analytics module (cost, time, accuracy, touch-free, pipeline, anomaly) and the financial-analytics module (duplicates, three-way match). All accessible via the API, all queryable from LEDGE, but composing them into one view is something the consuming application does - there's no single /v0/analytics/roi endpoint today. A dedicated ROI dashboard is on the wish list; flag if it'd matter to you.
Can I track cumulative savings since we started?
Cumulative-since-go-live tracking isn't built - savings figures are computed per-query for whatever date range you pass. You can approximate by querying with date_from set to your platform start date, but there's no persistent running total being incremented behind the scenes. Lifetime totals, year-over-year accumulators and "how much have we saved since launch" counters would be reasonable additions if there's demand.
Can OmniPATH project our ROI for next year?
No forward-looking projection or payback-period calculation is built. The ROI metrics report what has happened, not what will happen. If you want a forward projection, the typical pattern is to extrapolate from the period-over-period trend on your own - month-on-month savings × 12 gets you close, but it's not something the platform produces as a built-in feature. Forecasting would require some kind of trend model and we've intentionally kept the platform's claims to "what the data shows" rather than "what the data predicts".
Can I get the ROI metrics as a PDF report?
JSON via the API today - no formatted PDF report card. CSV / XLSX export is supported on the wider analytics endpoints if you want to bring the raw data into another tool to format. A board-ready ROI report-card PDF would be a reasonable feature; not in today.
Can I see ROI per supplier?
All ROI endpoints accept a supplier_id filter - so you can ask "what's the touch-free rate for invoices from FS Foodservice?" or "how many duplicates from Brakes?" individually. What's not built is a per-supplier savings calculation that allocates the £13,893 total down to each supplier's share - that's an aggregation rather than a per-supplier metric. Practically the same answer falls out of dividing total savings by share of docs_processed per supplier, but as a built-in number per supplier, it isn't there.
How does multi-currency work for the savings figure?
Cost calculations use the currency_conversion_rate stored on each purchase at the time it was posted, sourced from ECB rates. So the £13,893 saved figure reflects what the savings were worth on each invoice's posting date, not today's spot rate. Historical numbers don't drift when FX rates move - matching standard accounting practice and avoiding the "your reports change every time the rate moves" problem.
What's the difference between verified, confidence and quality accuracy scores?
Verified is accuracy on documents that a human reviewed and either confirmed or corrected - the gold standard, but only available for the subset of docs that were actually reviewed. Confidence is the AI's self-assessment of how confident it was in each extraction - averaged across all processed docs. Quality is the pipeline's own quality assessment, used as a fallback when neither verification nor confidence data is meaningful. The platform always uses the highest-priority signal that has enough sample size and tells you which one it used in the response under method.
Ready when you are

See your ROI from day one.

Bring 50 of your invoices to the demo. We'll process them, return the cost savings, time saved, accuracy and touch-free rate - live, with the underlying data - so you can see exactly what the platform would mean for your AP team.

30-min demo · Bring 50 invoices