Predictive web personalisation

Content personalisation adapts text, images, banners and entire web sections based on the visitor’s profile and behaviour. A new visitor from organic search sees different content than a returning customer arriving from an email.

Predictive models determine which content has the highest engagement probability for each visitor. This includes the homepage hero banner, recommended blog articles, the most relevant testimonials or the CTA that generates the most conversions for that segment.

In ecommerce, predictive product personalisation goes beyond "related products". Advanced models predict which products each user is most likely to buy, even if they’ve never seen them. This relies on collaborative filtering (similar users bought X) combined with current context data.

Catalogue personalisation includes search result ordering, featured products in categories, cart recommendations and suggested products in post-purchase emails. Every touchpoint is an opportunity to serve the most relevant recommendation.

Pricing personalisation uses propensity models to optimise offers at the individual level. It’s not about price discrimination (charging more to those who can pay more), but offering the minimum incentive needed to convert: a user with high purchase intent doesn’t need a discount; one with low intent may need a 10% incentive.

Transparency is essential: users shouldn’t perceive that they’re paying more than others for the same product. The most accepted practices are personalised coupons, limited-time offers and profile-adapted bundles, not different base prices.

Real-time personalisation adapts the experience during the user’s session, not just before they arrive. Every click, scroll and second on page generates signals the system uses to refine recommendations and the experience.

This requires specific technical infrastructure: a decision engine processing events in milliseconds, a feature store maintaining updated profiles and a rendering layer capable of serving dynamic content without penalising page performance.

Predictive personalisation must be validated with rigorous A/B testing. The most important test is: does the personalised experience generate more conversions than the generic one? Without this validation, you could be investing resources in personalisation that doesn’t move business metrics.

Tests should measure business metrics (revenue, conversion, retention), not just engagement metrics (clicks, time on page). Personalisation might increase clicks but not sales, or improve conversion but reduce average order value. Net impact is what matters.

Predictive personalisation requires user data, which raises privacy and consent obligations. GDPR requires explicit consent for cookie tracking and transparency about data usage. The trend towards a cookieless world compels a shift towards first-party data.

Companies treating privacy as a constraint lose out to those treating it as an opportunity. A personalisation system built on first-party data (data users voluntarily share) is more precise, more sustainable and more respectful. Transparency about data usage increases trust and willingness to share.