How Little Alchemy generated +39% of incremental revenue with web interstitial ads, without harming UX?

As a high-impact format, Interstitial needs to be introduced with skill and analytical background. For many digital publishers and website owners, money matters, and revenue generated by the product is the only measure of success – yet testing new advertising products requires looking at different aspects, such as performance metrics, and also User Experience. Regarding earnings, it is worth verifying whether we are thinking about incremental revenue – that is, additional earnings generated from an additional quantity of sales. The performance of other advertising products should therefore be monitored during the test, which involves assessing whether the new implementation has affected their results. Should other products register losses during the test period, then such losses should be included and recalculated in the test summary. User Experience is another challenge in adding new products to the advertising stack. Users visit a website to perform a certain task, whether it is to find some information, read an article, or, in this case, play a game. When we put obstacles in their way, there is a possibility that some of these users will find other platforms to interact with. Therefore, in addition to monitoring the performance of all running advertising products, the daily analysis of key metrics in Google Analytics should also be added to the test evaluation.

Interstitial ad format has been implemented on Little Alchemy under the supervision of a team consisting of an Analyst, AdOps expert, and Customer Success Manager, who is a dedicated publisher’s assistant. It was carried out separately on the mobile and desktop versions. The individual measures of success were strictly supervised by the team for several days, and any deviations from the norm were analyzed taking into account the current market conditions (seasonality). The test was completed when the behavior of the Interstitial units, both in the mobile and display versions, became predictable and when we could conclude its lack of influence on the users’ behavior.