> ## 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.

# Digital lender benchmarks

> Evaluate digitally originated lending across risk, data coverage and operational quality.

Digital origination is a distribution and operating model. Use the relevant [consumer](https://docs.amlok.io/research/consumer-underwriting.md) or [business](https://docs.amlok.io/research/business-underwriting.md) product cohort for credit outcomes, then evaluate the digital workflow separately.

## Financial-risk and operating measures

| Measure | Definition and comparison rule |
| - | - |
| Credit performance | Arrears, calibration and mature losses by product, channel and vintage. |
| Application coverage | Share of eligible applications with sufficient authorized evidence for the defined assessment. |
| Evidence error | Incorrect or unsupported extracted fields, reconciliations and calculations per reviewed application. |
| Assessment time | End-to-end time including data retrieval, retries, review and correction. |
| Abstention | Share requiring more evidence or human review, with reasons and later outcomes reported. |

## Public benchmark sources

Use the [Bank of Canada credit indicators](https://www.bankofcanada.ca/rates/indicators/financial-stability-indicators/) for matching consumer-product context and [ISED's financing survey](https://www.ised-isde.canada.ca/site/sme-research-statistics/en/survey-data-and-analysis/credit-conditions-survey) for Canadian business-financing context. Neither establishes a universal digital-lending approval rate or model-performance target.

Application completion and conversion depend on acquisition channel, eligibility rules and workflow design. Report those populations before comparing operational rates.

## AMLOK evaluation

Use a temporal holdout and, where available, a held-out acquisition channel. Evaluate fraud or identity outcomes separately from repayment risk. A rejected application without observed repayment cannot be assigned a favorable or unfavorable repayment label from the decision alone.

Compare the same evidence snapshot across systems and include end-to-end review effort. Test duplicated submissions, missing-source histories and delayed provider updates. For Signal, evaluate changed evidence and recurring monitoring quality, not outbound communication or servicing automation.

See the [document benchmarks](https://docs.amlok.io/research/document-benchmarks.md), [frontier comparison protocol](https://docs.amlok.io/research/frontier-benchmark.md) and [robustness framework](https://docs.amlok.io/research/robustness.md).


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