Skip to content
Free guide: How to use AI in compliance

Buyer evaluation tool

What does an accepted review file really cost?

Compare a KYC, KYB or EDD review pilot with your current process. Use the same accounting rules, inspect the quality measures and keep unresolved work in view.

Build your scorecard Read the methodology

Agree the test before comparing the numbers

Use this scorecard for your internal process or any service provider. It reports your inputs, sets no universal pass rate and makes no claim about Zyphe performance.

  1. 1. Define equivalent cohorts

    Record the period, case mix, evidence available, policy version, complexity and acceptance owner. Agree how setup costs are allocated. Keep the underlying case log.

  2. 2. Review quality independently

    Check completeness, ownership traceability and escalation decisions against an expert-reviewed reference set. The reference sample can differ from the operating cohort; record how it was selected.

  3. 3. Record jurisdiction and scope

    US and EU teams can use the same operational arithmetic. Agree the institution-specific requirements, access, oversight and approval boundaries separately. A scorecard does not establish regulatory suitability.

Enter the work, cost and outcomes

Start with one cohort or enter both. Blank fields are not treated as zero. Optional escalation counts use an expert-reviewed reference set.

Use one currency and one measurement cutoff. No currency conversion is performed. Enter aggregate figures only.

Current process

Fill every main field, using 0 where none applies. Leave the whole cohort blank if you only have one set of results.

Files

Distinct files in this cohort. Count each file once, including unresolved files.

Files accepted without correction. File acceptance is not customer approval.

Additional files accepted after correction. Exclude first-pass acceptances.

Files awaiting review, evidence or correction at the measurement cutoff.

Work and cost

All internal preparation, review and rework time for this cohort, including unsuccessful files.

Use a blended, fully loaded hourly rate for the internal time above.

Fees attributable to this cohort. Enter 0 if none.

Data charges not already included in the service cost. Enter 0 if bundled.

The setup or integration cost assigned to this cohort. Use the same allocation method for both.

Costs not included above. Avoid counting internal time or fees twice.

Optional: escalation quality

Use a separate, expert-reviewed reference set. Count each reference case in exactly one category. Complete all four fields, including zeros, or leave all four blank. Record false clearances separately in your review log.

Candidate pilot

Fill every main field, using 0 where none applies. Leave the whole cohort blank if you only have one set of results.

Files

Distinct files in this cohort. Count each file once, including unresolved files.

Files accepted without correction. File acceptance is not customer approval.

Additional files accepted after correction. Exclude first-pass acceptances.

Files awaiting review, evidence or correction at the measurement cutoff.

Work and cost

All internal preparation, review and rework time for this cohort, including unsuccessful files.

Use a blended, fully loaded hourly rate for the internal time above.

Fees attributable to this cohort. Enter 0 if none.

Data charges not already included in the service cost. Enter 0 if bundled.

The setup or integration cost assigned to this cohort. Use the same allocation method for both.

Costs not included above. Avoid counting internal time or fees twice.

Optional: escalation quality

Use a separate, expert-reviewed reference set. Count each reference case in exactly one category. Complete all four fields, including zeros, or leave all four blank. Record false clearances separately in your review log.

Calculations run in this browser. This calculator does not save or submit your inputs.

Enter your figures or load the illustrative example.

Every measure has a visible denominator

Methodology version 1 · 17 October 2026. These are operational definitions for this tool, not regulatory thresholds or independently verified benchmarks.

Accepted files

First-pass acceptances + additional acceptances after rework.

Count distinct files once. Keep unresolved and other unaccepted files visible.

Cost per accepted file

Total attributable cost ÷ accepted files.

Total cost includes external service, additional data, allocated setup, internal work and other costs. Internal work cost = minutes ÷ 60 × hourly rate. Avoid double counting bundled fees or internal effort.

First-pass acceptance

Files accepted without correction ÷ all files submitted.

Unresolved files stay in the denominator, so interim rates are provisional. This is not a rate among only completed reviews.

Accepted after rework

Additional files accepted after correction ÷ all files submitted.

This is not the total rework rate: files still in correction or never accepted may also have required rework.

Internal effort

All internal preparation, review and rework minutes ÷ all files submitted.

Include time spent on unaccepted and unresolved files. This is an aggregate mean; record case-level medians and ranges separately.

Missed-trigger rate

Reference cases requiring escalation that were not escalated ÷ all reference cases requiring escalation.

An expert-reviewed reference set determines the denominator. Record false clearances separately from other missed triggers.

Unnecessary-escalation rate

Reference cases not requiring escalation that were escalated ÷ all reference cases not requiring escalation.

Use a separate denominator from missed triggers. Count each reference case in exactly one of the four reference categories.

Zero, pending and missing results

A zero denominator produces “Not available”, never a zero rate. Reference counts left blank mean quality was not assessed here.

Cost differences are shown only for two complete, resolved cohorts with accepted files and confirmed comparability. A zero baseline cost has no percentage change. Small samples remain uncertain; the tool does not estimate statistical significance.

Assess evidence completeness and ownership traceability alongside these figures. A cheap file with unsupported findings may fail your acceptance criteria. A file with well-documented gaps can be ready for human review without being complete or ready for customer approval.

See the full evaluation checklist and fictional EDD file →

Bring your queue and your acceptance criteria

Use the scorecard to discuss workload, evidence, cost allocation and the reviewer handoff before agreeing a pilot.

Discuss a review pilot Compare providersUBO & EDD reviewPeriodic KYC review