Will Cost Per Engagement Become The Go-To Metric For In-App Advertising?

omerkaplanData-Driven Thinking" is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is by Omer Kaplan, co-founder and CMO at ironSource.

Performance advertisers looking to acquire users for their apps traditionally considered cost per install (CPI) their key metric. But as both brand and performance marketers become more focused on incorporating clearer KPIs into their campaigns, we’re starting to see other metrics emerge and bridge the gap between brands and direct-response advertisers.

One example is cost per engagement (CPE), which offers a more nuanced pricing model than CPI. Based on specific engagement actions taken by a user once inside an app, it’s easy to see how CPE can appeal to performance advertisers wanting more than just an install. Mcommerce, dating, gaming or content-rich apps can pay only for specific actions taken within the app, such as completing a level or opening the app a certain amount of times, while brands can pay based on specific engagement with ads, such as hovering over an ad to expand it.

CPE can significantly increase the quality of the users being acquired – and therefore ROI. While costs rise proportionately to the quality and rarity of the user, many advertisers are comfortable paying higher CPEs because they believe that getting a user who has reached a certain level of retention or completed a certain amount of actions in their apps are likely to stick around for the long term and become ROI-positive.

CPE also lends itself to retargeting or re-engagement campaigns. Instead of paying a subscription for a re-engagement service, mobile marketers can run a CPE campaign based on inspiring dormant users to become active again by driving specific in-app engagements.

Upside For Brands

While CPE seems to be the mainstay of performance advertisers, it can also offer compelling advantages to brands looking to add measurability and transparency to their campaigns.

While clicks, conversions or other specific actions might not suit a brand looking for a high-impact campaign, measuring whether users are engaging with an ad at all can help brands evaluate the impact of an awareness campaign. More preventative than a cure, running CPE campaigns can also work as an effective barrier to fraud. Faking an engagement, while doable, is significantly harder than faking an install.

Moving forward, we may see CPE find success as a “performance lite” metric, with just enough performance built in to satisfy an increased focus on ROI and measurability.

The Future Of CPE

While promising, there are several obstacles that may inhibit widespread adoption of CPE. Engagement, for example, is a broad term. CPE can fall between the cracks of brand needs, which are impressions based, and performance KPIs, such as completed actions, purchases or subscriptions. CPE could potentially be undone by the broadness of its own definition.

For pure performance advertisers, an engagement is only valuable in the sense that it can be connected to a concrete KPI, such as a completed sale. In this context, CPE would get passed over for more clearly defined metrics, such as cost per action.

CPE could also become irrelevant for big brand marketers looking for high-impact campaigns that drive awareness and create a specific public perception, as opposed to particular engagements.

Since it’s reliant on post-install activity and data, CPE can also prove problematic for technology providers, as they work to build models that can convert between the CPE advertisers are willing to pay for the end-goal engagement – for example, $5 for each subscription post-install – and the CPM or CPI they are paying publishers to get in front of those users in the first place.

For example, an ad network would have to account for the thousand impressions they bought, calculate the conversion rate to installs and then calculate what percentage of those installs met the specific KPI set by the advertiser. It’s a game of inches, and since advertisers lack the ability to know in advance which traffic sources will bring high CPEs, success is dependent on effective optimization on the part of the tech provider.

Either way, in order for CPE to have any chance at becoming a go-to metric there will need to be greater transparency and more sharing of data between advertisers and technology partners. By sharing the right data, marketers will empower technology providers to optimize more effectively and drive more high-quality users.

Follow ironSource (@ironSource) and AdExchanger (@adexchanger) on Twitter.


  1. I fully agree with Omer. We at Performance Revenues identified this judged shift towards real performance based need of our clients and have structured our company and technology only to achieve this goal.
    I hope that the industry will move in this direction, despite the barriers mentioned above, as this is the logical way to work.

  2. I agree with this article to a certain point. Sure, CPA has many pros. But how practical in implementation is it? I believe it is almost impossible. At the same time we at Appbooster offer cheap and effective CPI. I know it's not perfect, but it's better than theoretical CPA.

    • It's not a theory. We can actually deliver on CPA to our advertisers while buying on CPI and CPC and CPM. We developed specific technology for that.
      I dont think you can fight it as this is what advertisers want, soon they will be asking for that, and they have the power... so I believe networks should align themselves accordingly or they will be out of the game.

  3. I really appreciate your prediction of the future of CPE. As an advertiser, nothing could be better for us to see a more active engagement with less cost. However, it would be quite struggling for publishers. The advertiser just simply shifts the risks to the publishers and have more reasons not to pay them. At the same time, some new tricks in this game may be created.


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>