# Behavioural signals in our scoring: what we look at and what we don’t

*Source: https://credicorp.co.uk/behavioural-signals-in-scoring-what-we-do-and-dont/*

When you apply for a loan with us, our decisioning engine reads a handful of stable signals about you and your company — affordability, business credit profile, identity confirmation, prior arrears — and gives a score. That score never decides on its own; a person reviews every borderline. In 2026 we added a small extra layer to help those reviews: a set of *behavioural signals* drawn from how your account is used. Because anything that influences a credit decision deserves a plain-English explanation, this article sets out exactly what we do and do not use, what weight it carries, and how you can opt out.

## What the behavioural signals are

There are seven, and they are all **soft positives** with one exception. “Soft” means each one can only move the score by a small amount — typically half a point to two points each — and the combined ceiling is about nine points up or one and a half down. None can override a hard underwriting decision. They exist to help borderlines, not to flip clear-cut cases. The signals are: whether you have installed our app and use it as an app (not a tab), how engaged you are with the site over the last 30 days, the shape of your application funnel (a submitted application is a small positive; serial abandons over months without ever submitting is a small negative), whether you have used our on-site money tools (budgeting, repayment calculator), how long the relationship between you and us has run, whether you have ever used Web Share to share a Credicorp page (a tiny positive — advocates tend to perform well in our data), and whether you re-engaged after a previous decline (a small positive on the second application because it shows considered re-application rather than a fresh attempt).

## What we deliberately do not use

Several pieces of information we could read are **kept out of scoring on purpose**: night-time usage patterns, low-connectivity / Save-Data mode, and use of Simple View. These can correlate with protected characteristics — shift work, financial pressure, accessibility needs — and a fair model has no business letting them shape outcomes. We do let our human reviewers see them as *informational only*, with a clear label, so context is available without affecting the score.

## How a reviewer sees this

When a person reviews a borderline, they see a single signal table on your record with every signal we used, its raw value, its weight cap, and the actual contribution it made. They can also see, alongside, the informational items we explicitly kept out of scoring. Everything is itemised so the person making the decision can spot a signal that looks wrong and override it. We do not show your raw behaviour to anyone — only the aggregated signals.

## Your right to opt out

You can opt out at any time. If you visit our site with Do-Not-Track on, we never capture anything. If you would rather make the choice from the page itself, the cookie banner has a one-click opt-out which both blocks new capture *and* deletes the history we already have. Behavioural signals will simply not appear in your record, and the rest of decisioning will run as usual. Anything you have already authorised — your loan agreement, your payments — is unaffected.

## Why we tell you this

We could keep this layer invisible. We do not, because a lending decision is too important to take with information the borrower cannot see. UK GDPR Article 22 gives you the right to ask a human to look at any automated decision; we go a step further and explain the inputs in plain language so you can decide whether to engage with them. If a behavioural signal is unfair in your specific case, tell us, and a reviewer will look at it.

---

Credicorp Limited — UK lender to limited companies (Company No. 16093826). credicorp.co.uk