As part of App Promotion Summit in London, Samuel Chorlton, CTO of Redbox Mobile, hosted a workshop that shed light on the data science that drives the success of Apple Search Ad campaigns. Strong attendance at the workshop highlighted the desire by app owners and marketers to understand more about the metrics that shift the dial, so expect more from us about this topic in the future. If you prefer to jump straight into the recoded session, the video can be found here.
So, What Does Sam Do?
Sam’s day-to-day job is to manage the smooth-running and evolution of the Redbox proprietary Search Ads Platform and algorithm; the Platform employs a data-led approach to identifying and verifying keyword selection, so Sam’s session set about explaining how this process can impact bid management and subsequent campaign ROI. His session also included an overview on the Apple Search Ads relevancy algorithm and the impact this has on effective bidding.
What is a "Second Price Auction"?
He kicks off with a simple explanation of the ‘second price auction’ bidding process – although this is a simple system to grasp (basically, it’s very similar to how eBay bidding works), the auction is influenced by more than just the price you’re willing to bid. A key influence is ‘technical relevancy’, or how Apple’s algorithm judges your app’s suitability as a search-match and therefore how likely a user will be to click on the Search Ad displayed.
Sam digs further into the technical relevancy topic, explaining the key areas of your app listing and metadata (your app’s image and text descriptions) that affect how relevant your app is deemed to be, and suggesting some crucial App Store Optimisation actions that app owners and marketers can take to positively impact the result.
Let's Talk Keywords!
Let’s get back to those keywords. How do you determine which keywords to use (which keywords are users using when searching for an app?); which of those words are the most relevant for your app and how much should you need to spend on them to meet your objectives.
Apple provides some insight to which keywords are being used, and how much they’re trading for. This will help you to understand whether your bid is strong enough and sits within the suggested bid range for the keyword, but the data doesn’t necessarily take into consideration the technical relevancy of your app - it’s a starting point and it’s data that you can access. However, more can be done to better determine the efficacy of keywords, and Sam goes on to explain how we use built-in Platform functions to further enrich your knowledge pool when it comes to that topic.
Time to Dive into "Natural Language Processing"
Redbox employs Natural Language Processing (an arm of machine learning) to make sense of metadata text and search terms; we utilise Semantic Analysis to start to understand the context within which App Store users are using a search term. This is quite subtle analysis but unlocks additional search terms being used on the App Store that you may not have thought of.
Sam’s gives us a simple example to run with here: Imagine there are two news apps on the App Store; what are the common terms that are being used to describe those apps (in the metadata) and are the words that you, as a news app marketer, may not have thought of? These might be ‘breaking news stories’ or ‘live events’, for example. Our ML and semantic analysis assess whether newly discovered phrases or words can be added as new keywords. To determine this, we go looking for those terms within the similar apps on the App Store; we look at what other words they’re using to describe their app, such as ‘politics’. So, what we’re trying to do here is look for semantic relationships between text that may reveal new, yet relevant, keywords.
What is The Right Amount to Pay for a Keyword?
Sam’s session then looks at determining what is the ‘right’ amount to pay for a keyword? Most app marketers will know what they’re willing to pay for a keyword, but what should they be paying? Every scenario will be different of course, but the Redbox Platform tests and learns from millions of scenarios, to determine the most efficient price at which you can win impressions – remember, some keywords might have very low availability of impressions and therefore low availability of installs, which might require price adjustments.
How do we do this? By looking back at historical data, we create this performance picture for as many keywords as possible and we allow the Redbox algorithm to set the bid prices correctly, to achieve the KPIs you’ve set. Within this, we must allow the algorithm to understand and benefit from occurrences such as seasonal events, which might push bid prices about and alter the number of available impacts.
Let The Redbox Platform do the Heavy Lifting for You!
Redbox has a proven way to achieve all this in a way that does the heavy lifting for you, but in closing his session and before he troubleshoots viewers’ questions in the Q&A, Sam encourages app marketers to really think about not only the keywords for their app, but the context within which users might be searching for an app.
Im Summary
The session highlighted the important role that data plays in ensuring the effective nature of campaigns within the App Store and how sometimes out of the box thinking can further enrich this performance. You can view the whole session online, here.
Redbox will be looking to share more in the future on how a data led-approach can improve App Store performance, but for those than can’t wait all of this and more is already available to all our users at www.redboxplatform.com
At Redbox Mobile, we combine award-winning data science with human ingenuity to deliver award winning app store marketing performance. Our advanced Machine Learning AdTech Platform has a proven record of Search Ads success, and in this workshop, Sam offers an insight to its algorithm and a sneak look at how the new UI of the Platform will look.