36% reduction in cost per flight search through AI-powered bidding

The Challenge

Continuously seeking innovative ways to engage its audience through its marketing efforts, Wizz Air’s strategy involves targeting high-value new users to its site to ensure that users can be captured and remarketed when they’re in-market.

To qualify the value of users, Wizz Air uses a KPI called ‘Cost Per Flight Search’ (CPS). A flight search is recorded when a user enters their origin and destination airports and dates into the Wizz Air website and generates a list of search results. This ensures that Wizz Air are measuring meaningful site engagement, beyond a simple landing page visit.

We identified an opportunity to reduce the CPS for the campaign by testing AI-powered custom bidding against Wizz Air’s existing automated bidding strategy.

The Strategy

Our test was to run within Wizz’s ‘Escape the Winter’ promotional period, encouraging users in the UK and Italy to ditch the bleak March weather in favour of sunnier weather by exploring flights to one of their hero destinations.

Existing consideration (middle-of-funnel) activity included Display & Video 360 (DV360) ads and utilised Google’s intrinsic automated bidding strategy, which optimised towards driving a minimum CPA.

Alongside the focus on CPS, Wizz Air was keen to have visibility over both Searches and Bookings within their reporting. The existing automated bidding solution restricted Wizz from differentiating between which conversion action the bid strategy optimised towards. This meant that the bid strategy used blended data from two conversion actions, which was theorised to likely drive up the CPS.

To give Wizz Air clearer insights into its reporting and to optimise Cost Per Flight Search (CPS) effectively, we needed a solution that limited the bid strategy to using only the Searches conversion action while still allowing us to track both Searches and other relevant metrics in its reports.

We tested Google’s custom bidding algorithm, which uses AI to optimise bids for specific conversion events. In DV360, we set parameters to minimise the CPS by assigning the Search conversion action the highest priority level (1) within the bidding algorithm. This ensured that while we continued tracking both key metrics, only Search conversions influenced the algorithm’s bidding decisions.

To ensure that the test between the existing bidding strategy and the AI-powered bidding algorithm was fair, the settings, targeting and creative were kept the same, and we set up an A/B test within the platform, which ran for 4 weeks. This ensured that the algorithm had enough time to learn and optimise.