> ## Documentation Index
> Fetch the complete documentation index at: https://docs.amlok.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Embedded finance benchmarks

> Evaluate finance embedded in commerce and software platforms without duplicating underlying economic events.

Define the borrower, financing product and platform context separately. A platform's sales signal may support assessment, but a payment event is not itself a lending outcome.

## Financial-risk measures

| Measure | Definition and comparison rule |
| - | - |
| Revenue reconciliation | Supported sales linked to payment, settlement and bank receipt, with refunds and timing differences preserved. |
| Platform concentration | Share of verified activity or exposure dependent on one platform or channel. |
| Repayment burden | Contractual installments or receipt-linked remittances against a consistent income or receipt measure. |
| Reversal and dispute exposure | Refunds, disputes and chargebacks against an explicitly defined transaction cohort and observation window. |
| Credit performance | Product-specific payment and loss outcomes by platform, merchant and origination vintage. |

## Public benchmark sources

[Payments Canada's research reports](https://www.payments.ca/insights/corporate-reports) provide Canadian payment-market context. Payment-method adoption and transaction totals do not establish embedded-loan loss rates. Use product-specific lending sources alongside the [MCA](https://docs.amlok.io/research/industries/business/mca.md) and [working capital](https://docs.amlok.io/research/industries/business/working-capital.md) definitions where relevant.

An embedded-finance benchmark needs authorized platform evidence linked to a contractual credit outcome. Retain provider coverage, settlement delay and disputed or reversed transactions.

## AMLOK evaluation

Measure economic-event matching precision and recall across commerce, accounting, processors and banking. Evaluate whether refunds, fees and reserve movements explain settlement differences before classifying unexplained activity.

Hold out merchants and later periods; include a held-out platform where feasible. Report performance conditional on each source combination so improved coverage is not mistaken for a better model on identical evidence. For Signal, score persistent activity changes and distinguish an integration outage from lost sales.

Use the [record model](https://docs.amlok.io/research/record-model.md), [entity graph](https://docs.amlok.io/research/entity-graph.md) and [evaluation design](https://docs.amlok.io/research/industries/index.md#amlok-evaluation-design).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.