Customlytics is a Berlin-based full-stack mobile technology and marketing consultancy, solving challenges around everything from mobile tracking, analytics, CRM and deep linking to ASO, UI / UX design and performance marketing.
The company was running programmatic UA campaigns for a large non-gaming app developer using the largest legacy DSP in early 2020. During that time, Customlytics struggled to meet its client marketing KPIs on this DSP with little success. Something needed to change. So the team turned to Kayzen, a mobile specialized player, to help them achieve their business and marketing goals through UA campaigns with a robust self-serve platform. Customlytics’ main objectives were to:
- Get new installs for their customer in the GSA markets
- Convert app installs to valuable subscribers at a specific cost per action
- Improve the existing results on programmatic UA vs. legacy DSP
And doing all this while being in full control by complete self-serve campaign management.
Step 1: Onboarding
Work began with onboarding the Customlytics team with Kayzen’s training materials and support from a dedicated Programmatic Success Specialist, whose role is to help clients by providing consultation and support. Following completion, campaigns started within a few working days.
Step 2: Initial Setup
The next step in any Kayzen campaign set up is to integrate all relevant events with the platform. Kayzen has modular integrations with major MMPs. This enabled Customlytics to integrate all relevant install and post-install events in a matter of minutes, in this case, the MMP was Kochava.
Thanks to Customlytics’ strong marketing knowledge and Kayzen’s deep expertise on mobile programmatic we were able to design targeting criteria to navigate the UA exploration process successfully. Two strategies were created:
- Contextual targeting: We created a whitelist based on the analysis of the client-app audience, and were able to identify a set of publisher apps with high overlap of users. The underlying assumption was that the most used apps could be a good place to reach new subscribers for the client.
- Lookalike model: Together, we created a user profiling model to get similar users to the app-client top performers.
With this strategy, Customlytics launched its mobile programmatic campaigns and started bidding across all major RTB exchanges with display, native and video ad formats.
Step 3: Optimization
Once we accumulated enough data to analyze, we developed a few key drivers for a successful optimization process.
- Algorithmic transparency and control:
- Kayzen’s advanced bidding platform shows the aggressiveness and state of its ML algorithm — at a campaign level — in the UI at all times. This allows the clients to have full transparency on the algorithm stage and behaviour.
- With appropriate training on attributes and data interpretation. The Customlytics’ team was able to gain control over the status of the algorithm and decide on whether to put more emphasis on exhausting budgets or reaching target KPIs. As well as to understand the impact of the algorithm state across all metrics.
- With comprehensive analytics features available on the UI, the Customlytics team was able to quickly detect specific trends and optimize at different levels:
- Creatives: We found that creatives with a human face or top-selling products were outperforming other design concepts, thanks to that insight, the Customlytics’ team was able to quickly iterate and introduce more assets that used this knowledge.
- Publisher Categories: We found that puzzle games, certain entertainment, coloring, shopping and news apps were among the categories bringing in the most new subscribers. This was the key for further iteration on publisher targeting.
- Customized Bidding: Certain exchanges were driving most of the traction, so we customized bidding using bid multipliers to boost up exploration on those inventory sources.
- Differentiated Max Bids by Auction Type: Many exchanges in the mobile space have transitioned to first-price auctions, but there is still an important part of the RTB traffic accessible on second-price. Thanks to this feature, we were able to customize the bidding according to those nuances.
After five weeks of running, collecting data and gathering learnings, the Customlytics team switched the algorithm state from exploration to optimization. These were the results:
- The average number of installs increased by 32% WoW while reducing CPIs by almost 50%
- The CPA was reduced by 58% achieving new subscribers within the expected KPIs
Overall, the CPI and CPA achieved through Kayzen are now about 50% lower than what Customlytics used to achieve on the legacy DSP, while spending a USD four-digit daily ad budget.
“We are happy to say that our expectation was met and look forward to growing our partnership with Kayzen” – Christian Eckhardt – CEO, Customlytics