For marketers and advertisers in the user acquisition space, mobile app cost per install (CPI) has long been the default bidding model. Cost per install campaigns, where advertisers pay ad networks and other partners only when they drive installs of their mobile applications from the app stores, are ubiquitous.
But is this really the best use of ad spend for getting users for mobile apps? In our latest Whiteboard Wednesday video, Srinivas KC, VP and GM of InMobi DSP, argues why marketers should worry less about average CPI costs and instead using CPX bidding to buy against the most important metrics for the business.
Hello, welcome to another edition of InMobi Whiteboard Wednesdays. My name’s KC, and today we’re going to talk about CPX bidding. Everybody’s heard about CPC, CPM, CPI and all of the kinds of bidding in the programmatic world, but what in the hell is CPX bidding? Well, let’s get into it.
Traditionally what mobile marketers have done was to depend on CPM bidding, so when they’re buying media from an exchange or an ad network, they typically ended up paying on a CPM basis.
That evolved over time, and over the last many years, models have evolved from paying CPC - that is cost per click - to potentially CPCV and CPP, which is cost per video view or cost per playable, and those kinds of things. And more and more recently, the model everybody’s depended on is CPI or CPA, which is a cost per install or a cost per new user acquisition.
What’s happening, though, is that as marketing has become more and more a growth function, and a function that drives revenue and not just new users, the onus of delivering revenue and delivering business outcomes is increasingly on marketers as well.
What we see here at InMobi is that most of the marketers we work with are gravitating towards what we refer to as the CPX bidding model. What is X? X is a business outcome. It is anything that is considered a business performance metric.
Some examples of that are CPFT. That is cost per first transaction, for an e-commerce company that knows that a first transaction means the user that is going to be a long-term user. It’s a real user. They’ve made a transaction and they’re probably going to be a high value user. Similarly is CPFR, which is cost per first ride, if you’re a rideshare app.
CPTB: That is cost per tenth burrito. Let’s say you’re a food app or a restaurant, a QSR restaurant, with an application. You know a user is hooked and regular when they’ve had their tenth burrito. Can you put a price on that? Can you put a price on that event? That is the challenge a marketer has to face today and optimize towards.
And another example is CPFC, which is cost per first cappuccino. Let’s say you’re another QSR example where you have an app, you want users to order off the app and pick up a coffee. What constitutes or what defines a user that is hooked, addicted? Potentially it’s first coffee.
So that’s what CPX is all about. Cost per X, where X is the event that is a business outcome or a performance metric that directly impacts the business outcome.
Why is this happening? Part of it I talked about before, which is marketers increasingly have the onus of delivering business outcomes. But there’s other reasons too.
Purely from a trading perspective, from a programmatic buying perspective, buying CPM, the amount of effort required to deliver a campaign to the scale you can potentially achieve, the tradeoff just doesn’t work. You’re taking as an advertiser a large amount of the risk of buying media. You’re paying publishers but you don’t know if your business outcomes are going to be met. You don’t know if you’ll acquire new users. All you know is that you’re getting impressions.
Obviously the effort required to optimize towards business outcomes starting from there is huge, and therefore it limits scale because it’s all manual. People have to do a lot of work.
The other end of the spectrum is CPX, where what you’re able to do is depend on a lot of automation, depend on bidders that do most of the work in an automated manner. Take all the data points and analytics from the actual business, the actual application, and build it back into the bidding. And what it means is you can deliver, as marketers, significantly higher scale at a much smaller effort, which is why you see the transition from CPM to CPX has been accelerated in the recent years, as AI and ML models have kicked in, in the performance world.
There’s more tradeoffs associated with this as well. If you think about the amount of risk one would take running a CPM performance campaign versus a CPX performance campaign, there’s no comparison. CPX is no risk. A business only pays when a business outcome is achieved, so there’s absolutely no risk attached with that. On the other hand, CPM, you pay for media and not the business outcome. Same thing we talked about before.
The second thing is effort. Now, in a world where media buying teams have tens of people, many people hands on keyboard, analytics, a lot of people working on creative, a setup of that kind could potentially do a pretty good job optimizing the CPM campaign as well. But it’s a lot of effort. It’s a lot of bandwidth. It’s a lot of investment, a lot of people.
CPX typically tends to be fully automated. What you need really is somebody to run a strategy. So you need people who are careful, people who are able to craft the right metrics - the kind of people who will make the decision that the cost per first cappuccino is the right metric for the business or not. And once the marketer makes that decision, the hands-on keyboard effort and the operational effort is really minimized and CPX bidding takes care of most of it.
And the last point is of course control. The biggest risk with automation typically tends to be that when I was running my manual campaigns, I had a lot of control over not just the outcomes but also the inputs, and on a day-to-day basis we’re able to optimize campaigns. But in a fully automated system like CPX, we lose that control to a certain extent.
The idea though is that CPX releases the bandwidth of marketers and marketing teams to move away from optimization logistics and move towards strategies and outcomes. And as long as marketing teams spend the right amount of time defining the right outcomes for the business, defining the right KPIs and right performance metrics, the machine does the rest.
And what it means is the tradeoff between having a significant amount of control on the campaign and delivering results is actually not there anymore. And what it means is as long as the right results have been defined by marketing, automation achieves all of that.
So to conclude, the CPX bidding is all about the business outcomes. It’s all about the performance metrics and it’s all about automation, therefore delivering significantly higher scale at a much, much lower effort.
Our recommendations to marketers for 2019 and beyond are basically three things. One is work with partners that do support CPX. Working with partners that support only CPM, CPC and old school bidding models that really go way back when in timeline terms, it’s not good enough anymore. We’re in 2019. Marketers are part of the business outcome. Marketers are part of revenue generation and growth, and partners that support marketers also have to be.
The second recommendation we have is use analytics in apps to heavily drive through any outcomes. The reason to say that is to define the right outcome metric, define the right X, whether it’s the first coffee of the fifth coffee or the third coffee. As a marketer making that decision, it has to be very, very data driven. And that can only be achieved through a significant amount of analytics put into the app, to know the user’s journey, to know where the users drop off and how do you take that data and understand who the sticky users are versus not and define an event that will then help you identify the sticky users. So to that extent we do believe that analytics plays a huge role in this entire equation.
And last but not the least - and this is oft lost in terms of prioritization in the performance world - is you’ve got to shoot for scale when it comes to CPX. CPX is not for experimentation. CPX is not for just figuring out what is the right business outcome, and CPX is definitely not just to test out automation to then scale manually. CPX bidding is all about shooting for scale. You have the entire universe of programmatic supply as a marketer in front of you.
How do you take all 5 billion users globally that potentially are an addressable market for your business and target them effectively using the highest amount of automation in the smartest way possible? That is the right way to think about this problem and therefore shooting for scale is probably the precursor to even using CPX or 100% automation.
And that’s it for today.
Interested in learning more about app installs, acquisition costs and other related topics? Be sure to check out these blogs for more information: