Goscore
EN / NO
Run a backtest ↗
Evidence

Every number on this site, and where it came from.

Goscore publishes a small number of results from historical lending cases. This page says what each one measured, what it does not establish, and what a bank receives before it is asked to rely on any of them.

§ 01 — Goscore results

Historical Goscore cases.

These come from lending cases Goscore ran with partners, recorded in its 2023 bank materials. They are separate cases with different products, periods and denominators. This site does not add them together, and none of them is presented as a benchmark you should expect to repeat.

01
Case 01

Second look on declined applications

17.5 → 29.8%
acceptance rate, before and after
11.7 → 9.7%
default rate across the same lending

The lender’s existing scorecard ran first and was left unchanged. Goscore re-read the applications it had declined, using verified income, obligations and cash flow, and returned the cases where the evidence pointed the other way.

basis
Historical Goscore partner case, recorded in Goscore’s 2023 bank materials.
read it as
Both figures describe the same lending. The homepage renders them per 1,000 applications so the two rates can be read against each other.
02
Case 02

Real-time lending case

+21%
new customers
−32%
default rate

A separate lending case in which Goscore’s real-time financial assessment was used in the decision, measured against the lender’s prior performance.

basis
Historical Goscore case, recorded in the same 2023 materials.
read it as
A different case from 01, with a different portfolio and period. The two are not two views of one result.
03
Case 03

Recovery among rejected applicants

26.9%
of rejected applicants identified for recovery
3–6 mo
typical re-engagement window in this case

Applicants the existing process had rejected were re-assessed on current financial evidence, and just over a quarter were identified as recoverable business.

basis
Historical Goscore second-look case.
read it as
Identified for recovery is not the same as funded and repaid. What each lender then approved depended on its own policy.
04
Case 04

Consent flow in the lender’s own channel

+14%
more full applications
Up to 50%
more completed applications, standalone flow

Testing of a consent-based data-sharing flow inside the lender’s own application journey, measured on application completion rather than on credit outcomes.

basis
Historical Goscore consent-flow testing. The 50% figure comes from a standalone flow rather than an embedded one.
read it as
This measures the funnel, not risk. It says nothing about approval quality or realised loss.
05
Case 05

Operating scale

30k+
applications processed by 2022

Cumulative applications processed through Goscore’s financial analysis by 2022.

basis
Historical Goscore material.
read it as
A volume figure. It is evidence the system has run at scale, not evidence of a lending outcome.
§ 02 — How to read them

What these figures do and do not establish.

  • 01

    Each case is one lender, one product mix, one policy and one period. A rate that moved for one book does not transfer to another.

  • 02

    The lender’s own scorecard remained in place in the cases where that is stated. Goscore was additional evidence, not a replacement decision system.

  • 03

    Different cases use different denominators. Adding them together, or quoting the best figure from each, would produce a number no case supports.

  • 04

    Nothing here is audited or independently verified. It is Goscore’s own record of its own work, and it is offered as a reason to run a test, not as a reason to skip one.

  • 05

    The only figure that should decide anything at your bank is the one produced on your own portfolio. That is what a backtest is for.

Provided per case

What we send you, per case.

A published figure is a summary. Before a credit or model-governance function relies on one, it needs the detail underneath it. We provide the following per case during scoping, under a mutual NDA:

Sample
How the sample was defined and how large it was.
Period
The period covered, and the outcome window observed after it.
Baseline
The decision process in place before Goscore, and what it produced.
Configuration
Whether Goscore ran as a replacement, a challenger, or a second look after a decline.
Definitions
What counted as an acceptance and what counted as a default.
Measurement
Whether the result was observed in production or reconstructed retrospectively.
Run a portfolio backtest →

The number that matters is the one from your book.

Send a historical sample of applications with the outcomes you already know. We replay them through Goscore and compare against what your process actually did. Nothing touches a live credit policy.

Run a portfolio backtest →