ivi AI personalisation product design case study

B2C SaaS streaming service with 65k titles and 60M monthly active users.

During my work at ivi, I was responsible for two strategically important areas: personalisation and onboarding. The key metrics were retention, LTV, and user activation, particularly conversion to viewing content.

Role: Product Designer
Company: ivi Ltd
Period: May 2021 – May 2022
Design team: 6 Designers
Revenue: $227M

Case study presentation

A recorded presentation that combines the ivi case study, product decisions, and supporting conference slides in one storyline.

Certificate

UI UX Global Summit 24 Speaker

Issued by Geekle on 8 August 2024 for Svetlana Prodius.

View verified certificate

Presentation slides

Opening slide from Svetlana Prodius speaking about adaptive user experiences

Conference slide describing UX and UI design skill with AI and product design branches — A framework for combining UX, UI, product design, and AI/ML thinking.

Conference slide about an AI ML recommendation system — How recommendation systems sit at the intersection of AI and ML.

Conference slide about selecting the right input data for the system — The most important question: which signals should be fed into the system.

Conference slide explaining how training data is collected, models are evaluated, and predictions are generated — From collected training data to evaluated models and generated predictions.

Conference slide showing recommendation examples based on liked artists and films — Examples of how preference signals can be translated into recommendations.

Conference slide about insufficient data and large-scale data for adaptive systems — A comparison between cold-start conditions and large-scale historical data.

My role

As Product Designer for Activation and Retention, I owned the onboarding and engagement experience at ivi. I was responsible for designing the new onboarding flow.

I worked alongside a researcher, product manager, and UX copywriter. The first version was launched in June 2021.

Personalization

What should I watch tonight? A large catalogue of films and TV shows can still make choosing frustrating. ivi uses personalisation to adapt its catalogue and recommend relevant content based on each viewer’s interests.

The problem. The algorithm worked well for only 10% of users who had enough data about watched, skipped, saved, liked and disliked content. The same group had excellent retention and LTV.

The task. Collect enough information about a viewer’s interests during their first session to immediately adapt the catalogue and recommendations.

I explored ten question formats and design directions

I proposed around ten question formats and multiple UI directions that could help understand user interests before meaningful watch history existed.

Final versions

The final onboarding variants turned preference collection into a smoother, more confident first-time experience and helped adapt the catalogue earlier.

Improve retention

The task was to increase TV-show viewing. Research showed that series viewers watched for longer and returned more often than audiences focused on standalone and children’s content.

The solution was to add a personalised TV-show block to the main page, using recommendation mechanics and behavioural data to show a relevant selection.