Why Algorithmic Marketing is a Good Thing for Mobile Marketing

This post is a summary of a presentation by Gabe Kwakyi at the 2018 App Growth Summit event in New York, NY. You can find the slides here. This post is part 1 and focuses on why the age of algorithmic marketing is a good thing for mobile marketers. Part 2 will focus on how to analyze and optimize the success of your marketing campaigns using data.

 

The age of algorithmic marketing has become mainstream in mobile marketing with the launch of Google UAC; we think this is a good thing.

If you don't yet understand the change or why it's a good thing, let's explore a few concepts:

What Algorithmic Marketing Changes

The following is a list of the major (although not fully inclusive) levers that mobile marketers were able to pull in an AdWords web campaign prior to algorithmic marketing campaigns:

• Top-level
  • Geographic (country, state, city)
  • Platform, device (iOS/Android, phone/tablet)
• Mid-level
  • Network (display vs search)
  • Interest (word games)
  • Keywords (word games)
  • Demographic (age/gender)
• Lower-level
  • Lookalike (purchasers)
  • Remarketing (installed not purchased)

While the top-level levers will remain (except for device), algorithmic marketing deprecates the ability to select different mid-level targets in favor of a single lever: event objective. 

This is the crux of the new age of mobile marketing: marketers can now rely on algorithms to do the research in identifying the right user traits to target, enabling marketers to focus on deciding which usage outcomes or user behaviors to target (e.g. users who install, sign up, or purchase).

Why Machine Targeting > Surface-Level Human Targeting

You may be thinking (as we initially did) that the ad networks are crazy and mistaken to do away with the levers mobile marketers have come to know and feel confidence in pulling.

Consider this example that illustrates why targeting a user by outcome is more useful than surface-level traits:

In this example we are targeting two users who share a great deal of surface-level traits, but are worth considerably different values as users of our app. Being focused on targeting users by surface-level traits ends in the same creative, the same bid, and the same budget being used to target both Joe Schmo here and Jeff Bezos.

Consider another example of two users who are both high LTV users of Wordscapes:

• Wordscapes user A
  • Likes word games
  • Searches for [word games]
  • Ad relevance: 10/10
• Wordscapes user B
  • Likes traveling
  • Searches for [flights to new york]
  • Ad relevance: 2/10

In the age of surface-level targeting, Wordscapes would only be able to show ads to user A, because Wordscapes is ostensibly not relevant for traveling or [flights to new york]. The fact that user B does not present surface-level interests to Facebook does not mean user B is less valuable.

Let's return to the concentric circles example for a machine-learning driven/algorthmic marketing campaign.

By targeting users by event-targeted campaigns, we can now effectively target Joe Schmo and Jeff Bezos with different creatives, different bids, and different budgets.

The algorithms (pulling thousands or more back-end levers) are much more capable than humans (pulling hundreds or fewer front-end levers) in identifying that Joe Schmo values his money more than his time and that Jeff Bezos values his time more than his money, and thus the potential value of each user to our app is significantly different, despite the fact that they share many surface-level traits.

How to Target Algorithmic Marketing Systems

Targeting by events does not necessarily mean creating just one or two campaigns. Events can themselves be segmented into more events, such as degrees of event completion; rather than closing off the ability for humans to target, algorithmic marketing opens up a whole new slew of segmentation abilities, all based on studying actual behavior instead of inferring behavior from surface-level traits.

Algorithmic marketing and the ability to acquire higher value users (and the requirement to pay more per user acquired) also means that remarketing campaigns become more important.

Where Algorithmic Marketing is Not Yet Good Enough

While algorithmic marketing is great in many cases, there are two main cases in which machine learning-based marketing does not work out so well:

1. Apps occupying small niches
2. Apps with insufficient marketing investment

Machine learning is great at making statistically significant inferences and driving better results overall, but also requires a lot of data to work properly. While driving installs is something that the algorithms can handle at mostly any scale, optimizing algorithmic marketing campaigns for downstream events requires enough users completing enough events to raise the chances of success. For example, Google recommends a minimum of 10 events/day (e.g. 10 purchases) for UAC campaigns, yet this is often insufficient to maintain or scale campaigns.

Getting around this challenge involves a careful strategy of shifting the focus to higher-in-the-funnel events (e.g. sign up) that occur more often (but are less valuable) and gradually improving the quality of users acquired, so that the number of events that occur at the desirable levels of the funnel (e.g. purchases) increases to the right level to optimize campaigns directly at the desired level.

 

Stay tuned for part 2 and more articles on how algorithmic marketing affects the world of mobile marketing!

 

Be sure to bookmark our blog, sign up to our email newsletter for new post updates and reach out if you're interested in working with us to optimize your app's ASO or mobile marketing strategy.

Incipia is a mobile marketing consultancy that markets apps for companies, with a specialty in mobile advertising, business intelligence, and ASO. For post topics, feedback or business inquiries please contact us, or send an inquiry to projects@incipia.co