Sprzedajemy.pl: personalization that got the visitors of a classifieds site browsing, clicking and coming back

The challenge

Traffic was there, but many visits ended after a page or two: people looked, did not click, and left without contacting a seller.

Every visitor browses differently, one spends an hour on Alfa Romeo after a glance at BMW, another only checks the newest listings, so one message for all would not move them.

What we built

• Firecrux on the site with a single tag, working as a layer over the site’s code, so the client’s developers kept their own roadmap.
• Personas built from behaviour and from each visitor’s similarity to others, instead of hand-made segments.
• Recommendations of similar listings and reminders of the ones a visitor looked at before, on the listing page and on return visits.
• Web push to the people most engaged in a category, or to those browsing it right now: for example to users who spent the most time on Alfa Romeo in Motoring, even if they glanced at BMW.
• Exit intent handled per persona, with an offer that fits what the visitor was looking at.
• A/B tests for any element, per persona, segment, device or dataLayer value, without extra tracking code.
• Global rules against over-personalization, such as at most five banners per session and two minutes between chat greetings.

How it works

Firecrux collects events from the site, groups visitors by what they do, and decides in real time what each of them sees. Each change runs as a test first, and only what makes people click stays.

Results

• Visitors who came for one listing were shown the next ones worth their time, so a visit turned into a session.
• People came back to listings they had looked at, and to new ones in the categories they follow.
• The site’s team got one place to run personalization and tests, without waiting for a developer.