October 28, 2020
The best-kept secret in mobile advertising is that a low cost per install does not equate to profitability.
Need a primer on mobile game and app marketing? Check out Mobile User Acquisition 101: A Beginner's Guide.
In the early days of the App Store, CPI made sense as a performance metric. But in the freemium age, post-install behavior is all that matters, making CPI far less relevant. One driver of this trend is that mobile attribution has grown more sophisticated, allowing advertisers to get detailed data on post-install behavior. The other is that installs alone rarely result in net-gains: A 2016 study of 37,000 apps found that one in four users abandons an app after the first use.
As monetization strategies have evolved, so too has user acquisition. Specifically, the media buying landscape has shifted towards programmatic. CPI is not built for the programmatic world — CPM is more universal and thus more scalable in a large, complex ecosystem. Unlike SDK networks, programmatic platforms ingest a high volume of impression-level data and use it to train machine learning models. Those algorithms then target users based on your goals, such as whether they are likely to pass a key in-app threshold or make a purchase. On any platform where such capabilities exist, aiming for installs is aiming low.
The interplay of these dynamics means that CPI is misleading in the pursuit of actual revenue. So why is no one talking about it? And why is CPI featured in every industry report, and factored into every UA manager’s media planning? The truth is that those who benefit from the status quo have a vested interest in maintaining CPI as a benchmark metric and insist on its value. I say it’s time to take the blindfold off and focus on what matters: Do your mobile UA campaigns drive revenue?
Sometimes it’s easy to forget that mobile advertising is still in its infancy. Most people alive today remember a time when the word “app” wasn’t listed in the dictionary. More broadly, digital advertising is still evolving. We’re actively negotiating the balance between user privacy and the free flow of data on which advertisers have come to rely.
In the last decade, we’ve seen performance metrics enter and exit vogue. First, we co-opted the impression from broadcast television. Then we developed the technology to track clicks, installs, and video views. Now UA managers can target post-install actions, such as an in-app purchase or completing an action like level completion. The question is, why do we continue to talk about CPI when cost per action tells a fuller story?
The simple answer is that it’s a crime of convenience — it’s easier to stick with what we know. CPI is one of the few metrics that MMPs, SDK networks, and DSPs all measure consistently, so it appears to be a valuable baseline for making apples-to-apples comparisons. Meanwhile, regions with less-developed mobile ecosystems continue to rely on legacy revenue models. For example in APAC, CPI plays a major role in UA strategy.
But there’s another side to the equation: In an industry obsessed with data, some companies are more focused on churning out flashy numbers than delivering value. And that’s a real problem.
According to recent data from AppsFlyer, mobile ad fraud cost the industry $2.3 billion of ad spend in the first half of 2019. Its CTO correctly cites a rise in bot attacks, click farms, and device emulators as key factors behind the increase. But equally insidious are those traffic providers that knowingly allow organic poaching to occur on their platforms.
Organic poaching is when a traffic provider assumes credit for an install that would’ve happened anyway. It’s especially challenging to identify because these installs are linked to real users — often high-quality users with a genuine interest in the app. Organic poaching can involve any of the following tactics:
Traffic providers that have poor attribution hygiene can be the hardest to identify. Reports will show a low CPI and a steady stream of quality users. This is because organic users are the highest quality both in terms of retention and ROAS.
Such networks often offer fraud rebates as a business practice, but quantifying the impact of fraud can quickly become a gray area, and advertisers aren’t always reimbursed what they feel they’re owed. At MOLOCO, we prefer to take a more proactive approach to preventing ad fraud, and implement a multi-layered system including automated and human-layer checks, to keep our traffic as fraud free as humanly possible.
Despite the enduring prevalence of fraud, there are steps you can take to protect your campaigns. First, choose traffic providers with strong machine learning capabilities. These algorithms optimize targeting based on CPA because that’s a stronger predictor of ROAS than CPI ever could be. Next, run incrementality tests with all partners. When you start with a new channel, do your organics take a nosedive? If so, that’s a red flag.
It’s also important to opt for short attribution windows. For click-through attribution, we recommend a 24-hour window. For view-through attribution, we recommend using a 24-hour window for video, and a shorter window for banner placements (typically between 1-6 hours). This is especially crucial given the launch of iOS 14. The new OS will make Limit Ad Tracking the default, effectively eliminating IDFA on iOS. Apple has delayed the feature but plans to enforce the policy early next year. Also, the new SKAdnetwork requires you to use the same attribution window across all ad networks. As a best practice, use a short attribution window to minimize your risk.
Finally, start the conversation with your peers. Next time you hear a CPI success story, ask how that campaign impacted overall growth. Stop settling for results that look good on paper and start actually making paper.
Operational ML through real-time data offers precision in targeting and bid optimization in advertising. Learn how to master machine learning and navigate the nuances of real ML solutions to grasp the future of advertising for measurable growth.
The transformative impact of machine learning on modern advertising can provide real-time decision-making, targeting, and measurable outcomes. Learn how advanced machine learning-powered solutions can elevate advertising strategies and drive tangible results.
Explore how the transition from generative AI to operational machine learning is reshaping mobile app marketing and learn how leveraging this cutting-edge technology can unlock exponential growth and unparalleled success in the digital realm.
Uncover user engagement insights with Engaged-View Conversions. Gain a deeper understanding of how users interact with your video ads beyond clicks and views. With EVC, discover that users who watch skippable videos for more than 10 seconds or complete shorter ads are 4 times more likely to convert. Now available for all Moloco advertisers, EVC provides comprehensive data on views, engaged-views, and clicks to shape your marketing strategies. Enhance decision-making and drive business growth. Visit our Help Center or contact your Moloco representative for details.