Machine Learning Is A Key Component In Managing Mobile Advertising

Machine Learning Is A Key Component In Managing Mobile Advertising
Machine Learning Is A Key Component In Managing Mobile Advertising

A large number of mobile devices, the volume of apps on each phone, and the basic mobility of the devices all mean there is a lot of information being creating in the mobile world. Managing that large volume of information is impossible in a reasonable timeframe using older technologies. Machine learning (ML) is critical to mobile advertising in a number of ways.

Advertising is complex even in the older channels of print and broadcast. Cable increased the need for better data to more finely segment the audiences. The Web has meant even more finely tuned narrowcasting, focused on smaller groups and even individuals. Mobile devices, primarily the growing presence of smartphones, add another, massive, layer of complexity to the advertising challenge. Some of the key concerns in the mobile world include:

  • APP BASED ADVERTISING
  • LOCATION-BASED ADVERTISING
  • ADVERTISING FRAUD PROTECTION

The first two issues directly address narrowcasting and the third has to do with appropriately allocating advertising costs and preventing loss. This column focuses on describing those three areas, showing why ML can be an appropriate solution. Later columns can delve a bit more into the specific solutions for each risk.

APP BASED ADVERTISING: What many people think of as the most powerful and most annoying tool on their smartphones is the app store, whether it is Apple AAPL +0%’s or Google GOOGL +0%’s. The large number programs (apps), give a lot of choices but also can make that choice difficult. That is true for the advertisers as well as the consumer.

Smartphone users regularly see advertisements in most free apps. There is a long supply chain involved in ad placement, linking advertisers with app creators through both ad agencies and ad networks. With so many networks and apps available, how is an agency or business choose the right placement?

That’s where another part of the chain comes in. The mobile measurement provider (MMP) is a company that provides an SDK for app developers. The purpose is to capture information about the mobile phone and usage in order to provide feedback to all the companies in the chain so that adverting can be better tracked and planned. “The mobile advertising ecosystem is complex and deals with billions of transactions,” said Charles Manning, CEO, Kochava. “The MMP adds value by tracking and analyzing the unique usage in mobile devices in order to help marketers better allocate their advertising spend.”

That tracking includes information about clicks on a phone, users clicking on an ad for a new app and whether or not that new app was installed, to clicking on an ad for a product and going to the eCommerce site for more information and purchase. The unique information about a phone allows the MMP to combine that information with other sources to build a demographic picture of the device owner, allowing better placement of advertising. The companies must, of course, follow privacy rules for the customers, especially in relation to the EU’s GDPR. While that means the information would be anonymized, that still allowed better spending decisions.

LOCATION-BASED ADVERTISING: The app-based decisions don’t require real-time, but the move to location-based advertising will show as a clear benefit to using an MMP’s services. The ability to quickly identify locations and provide information will help advertisers provide more immediate options to an owner of a mobile device. The ability to locate a device is even more fine-tuned than the obvious question of advertising a restaurant that’s a block ahead of somebody walking. In-store advertising can be improved by providing coupons and advertisements relevant to the grocery aisle a customer is currently walking through.

That same benefit is clearly relevant to the advertiser. If someone is driving through a town, an ad for a restaurant or a store that’s a mile ahead is more valuable than one for a business that’s a mile behind the driver. A business that can use location information even more precisely than “the person is in San Diego” can even more finely narrowcast to an audience.

ADVERTISING FRAUD DETECTION: While the previous two issues help to link business advertising to better revenue per spend, fraud detection is a way to prevent loss. Just as with so much of the world, mobile advertising is vulnerable to fraud. In this discussion, we’re not talking about fraudulent ads, as that issue is dealt with in a non-technical issue through the FTC in the US and other agencies and laws around the world.

As mentioned, the mobile ad world has a complex ecosystem. However, the core issue in ad success is the “click”, did someone click through the ad?

The path for a business that wanted to advertise on a mobile device is mentioned above. In the most complex chain, a company hires an advertising agency that works with multiple ad networks to place the ad in a number of apps. An SDK such as that of Kochava’s sits between the app and the app store or ad network. When it sees a click in an application, it notifies the appropriate network, and then back upstream, so that the advertiser then pays the app provider (the entire chain taking their commissions on the way) for the click.

The problem comes that security on mobile devices is still very new. Malicious apps can watch what the user does and then send an attribution message downstream so that the wrong software gets paid for the click. That’s especially an issue when it is an ad for another app. That can be complex, and a more detailed discussion is past the scope of this article, but the end result is a lot of money can be stolen by the providers of fraudulent applications.

To show the extent of the problem, just last December Google removed 22 apps from the Google Play store due to click fraud. Twenty-two apps might not seem like a lot, but they were downloaded 2 million times.

“Analytics on clicks can help quickly identify fraudulent behavior,” said Manning. “That not only prevents money from being stolen from advertisers, but it provides increased trust needed in a new and rapidly growing advertising channel. Confidence in a solution that quickly addresses fraud allows advertisers to feel more comfortable about increasing ad spend in the mobile world.”

AI IN MOBILE ADVERTISING: In all three of these areas, there is a large volume of clicks and transactions. While each one has different demands for response time, they all have a volume level that makes human interaction in a reasonable response time a near impossibility.

Mobile advertising, both the rewards and risks, comes at the business from multiple angles. To quickly choose ads for the right apps, provide attribution for app downloads, provide location-based ads, and identify fraud are all areas in which machine learning’s ability to analyze large volumes of data can have an impact.

originally posted on Forbes.com by David A. Teich