# How do you stop personalization from reinforcing bad choices? | PrefEngine

Source: https://prefengine.com/blog/personalization-feedback-loop/

PrefEngine answers


Add exploration, cap repeated exposure, and review outcomes by segment. An engine trained only on yesterday’s clicks can bury good products before shoppers see them.

## Use the smallest rule that works

Start with a useful default. Change it only when a fresh signal is strong enough to improve the decision. Keep inventory and availability as hard constraints, then measure conversion, profit, and returns against a control.

## What to check before launch

Write down the rule, its data inputs, the fallback, and the condition that turns it off. Review small segments carefully. A dramatic result from a few sessions is not evidence.

## Bottom line

Add exploration, cap repeated exposure, and review outcomes by segment. An engine trained only on yesterday’s clicks can bury good products before shoppers see them. Test the result against a control before expanding it.

Source: [FTC, Bringing Dark Patterns to Light](https://www.ftc.gov/business-guidance/resources/bringing-dark-patterns-light). Reviewed Sep 22, 2026.