The Performance Marketing Auto-optimization Machine
Today we hear from Mike Silley on a recent project tackled by the Performance Marketing team.
Hello there! My name is Mike, and I am a Product Manager on the Performance Marketing team at GetYourGuide. Our mission is to ensure people can find the travel activities they love through online advertising. In its simplest form, it means connecting our amazing tours to the customer’s demand for an amazing experience, all the while minimizing showing ads to individuals who are not interested in buying our product.
Automation over Active Management
Ensuring we constantly target the right people is a question of active management or targeted automation. Active management of any company’s marketing accounts can be time-consuming and riddled with human error. Eventually the operational bloats of having an army of marketing managers working on optimizing different parts of the same portfolio will lead to performance inconsistencies and unwanted management overhead. Here at GetYourGuide, we do things a bit differently: we run autonomous teams where data analysts, engineers, and marketing specialists are embedded within the same team. We make data accessible to all team members, so opportunities can be identified by anyone, and we promote automation over repeated manual execution. One such project we developed was an auto-optimization framework for our performance marketing channel.
The Autonomous Optimization Engine
The goal of this project was to build a fully autonomous optimization engine that could identify poor performing marketing entities and pause them. This same engine should also consider unpausing entities that previously weren’t performing but might perform now . For example, it could pause if we run out of tickets to sell to our customers, and it could restart the campaign once we get more tickets.
Given we have active marketing for 40,000+ products, being marketed in a dozen and a half languages across 150+ countries, some of the engineering challenges of this project consisted of determining how we can regularly collect and consume performance and marketplace data to compute and push optimization changes to our marketing accounts. There is also the exciting analytics challenge of understanding how to maximize cost savings while minimizing the removal of positive revenue uplift potential. This is no small feat given the level of data scarcity we observe in some portfolio areas! Finally, measuring incremental impact to improvements in this framework lies between data science and analytics.
A project like this would be very difficult to execute without active participation from each of the team functions. Our cross functional team structure, emphasis on data accessibility, and grit, allowed us to drive this complex project forward in a way that was not only fun and a great learning experience for everyone involved, but also allowed each team member to make a large and lasting impact with their work. The outcome of such a project has led to a fully autonomous system that can improve our account performance, drive company growth, and ultimately improve the customer experience.