AppsFlyer: how to analyse, measure and optimise user acquisition campaigns

Peter Keung

In App Marketing. July 21, 2014

Oren Kaniel AppsFlyer
App publishers spend money on advertising but how do they know if the money is well spent?  AppsFlyer provides independent app tracking service that allows evaluating each user acquisition channel, measuring user their lifetime value and the channel profitability.  In this episode of AppInTop mobile app marketing podcast we talk to the CEO of AppsFlyer Oren Kaniel, CEO and marketing director Ran Avrahamy.  Listen to this podcast episode or subscribe on iTunes.
Which element of AppsFlyer technology are you the most proud of?
When we started the company back in 2011, as a small DC firm on the East Coast, I realised that app developers have no idea how users are discovering their applications.
The reason for this is the lack of cookies and the different technologies of mobile devices as opposed to the web. A mobile device does not have cookies and the code is not HTML, it’s native code. The rise of App Store and Google Play as opposed to the web posed another problem. These three things posed a great challenge to the user acquisition and there was no technology to solve this problem. This drove me back to my hometown where I developed this unique technology called NativeTrack. NativeTrack simulates the cookie environment in the cloud thus overcoming the three main challenges.
When we launched it at the end of 2011, our investors asked us about our competition and we didn’t have any. Now, there are companies that are trying to replicate what we did. However, engineering power can not create a search engine. Let’s take Google, for example. Google business is driven by the users. If, one day everyone suddenly stops using Google and starts using Yahoo, Yahoo will have that information and data to provide search service.
The same is true for attribution because we’re planting machine learning into these algorithms so that we can study and improve tracking algorithms over time. If you don’t have the data, you can’t improve. To get into this market right now, you have to work very, very hard to get this data so you can do your job correctly.We are integrated with almost five hundred different ad networks. Before platforms like us existed, the developers and advertisers were buying ads based on CPC or CPM, whereas now, they can buy CPI or CPA, which are actions within the applications. They need a reliable source of data for optimisation and to charge the clients.
Let’s talk about metrics and methodologies you are using.
We didn’t invent anything though. Everything is taken from the investment world. If you’re investing in Facebook stock, for example, you would like to measure the return and the ROI just like you would if you were buying real estate. The whole industry of advertising transformed from spending into investing. Even just five years ago, all the advertising campaigns were very hard to measure outright if you’re talking about TV adverts etc. Now, with the internet, you can measure each and every dollar spent.
How do you define loyal users and how do you calculate the ROI?
That’s very easy actually. You need to measure the return and you need to know your acquisition cost. That needs to be a positive number. Let’s say the cost is one dollar. You want to make more than one dollar in lifetime value. You can compare your mobile app to home building for rent with an endless apartment. That’s equivalent, but you have to ensure that the cost and user acquisition cost is lower than the lifetime value.
Can your technology monitor or measure the virality of the app and  how does that work?
Firstly, we track organic installs. Let’s say you have an advert in the Superbowl. You can see the jump in the user acquisition in organic terms which means that we don’t know where the users came from. You have to consider why the jump was on the organic installs. You can correlate this to paid acquisition, for example, if you ran a burst campaign through the weekend and you got 20K installs as opposed to the normal organic 10K installs you receive every day during the weekend. You can correlate that to the paid campaign and the paid campaign delivered more users. You can calculate the K-factor and the virality factor for your application based on this data.The technology we provide is not only for paid acquisition. It’s also for user invites, emails, texts etc. You can measure the virality of your application if you’re using social share or user invite and that sort of thing.
Can your service track an install and user action in the scenario where they install immediately and then delete the app offline?
We only track it if the user actually uses the application. If they don’t use it, it’s worthless to track it. We can track losses of data when a user moves from 3G to Wi-Fi and that sort of thing though.We can track data about pre-installed apps too, that’s just like an organic install. We have a special solution for the baseline of apps that will be pre-installed. The advertiser or developer will be able to see exactly which pre-install it came from, because you can have multiple pre-installs. We like to know exactly where the users have come from so we work with the device vendor and the carrier(s) to provide a better granularity of the data of the organic installs.
Have you ever run into the problem of dealing with data loss because of a server load and how did you deal with it?
We’re fairly used to that actually. We’ve seen a steady increase in the volume or traffic volume with a 30-50% increase every month in the last year and a half. We work on top of Amazon Web Services and we are using ultra scanning groups so the number of instances can grow automatically if the traffic grows. The uptime of our platform is five-ninths which is very transparent. If you went onto you can see the status of our servers in real time.
Why should a smaller app developer use Appsflyer? Are they not discouraged by the cost of using your service?
Firstly, of all of Appsflyer’s cost, the measurement cost is only around 1-2% of the media cost according to our internal statistics. That’s very low compared to the value we generate.  It doesn’t really matter whether you’re a smaller player or a bigger one. We aren’t charging for anything else, there’s no minimum fee. We found out that our pricing model is in alignment with developer campaign success. If a campaign wasn’t successful and they didn’t get a lot of installs then they aren’t paying enough. If the campaign was very successful they’re paying for the installs they got from the campaign. We don’t charge for organic installs or for every cost per event or click like some providers do. We only charge when they’re acquiring some new users.
In your reports, you state social media as the most effective source of users. Does it make sense to place ads on alternative platforms?
We aggregated all the media sources and campaigns we took into one report. It doesn’t mean it will work for all app developers though.Social media such as Facebook is performing very well in terms of user acquisitions. I would recommend that developers try all media sources that they can, including social media.
What are your thoughts on CPI networks and the potential for fraud through incent installs?
The market is all about CPI (cost per install) which is great. If you aren’t using CPI directly, you’re probably calculating the effective CPI cost. You can do that and then calculate the cost per registration or cost per first to turn booking etc. Fraud is not necessarily coming from incent installs. These can degrade for certain applications, yet with higher levels of engagement and for other applications, they won’t. Incent installs and fraud aren’t related.Sometimes for incent installs, users are installing the app, deleting it and reinstalling it just to get some sort of benefit. We do not allow that, you can only install the application once. You will not get any benefit by deleting and reinstalling the app, no matter how many times you do it.
What is an advantage to use Appsflyer compared to Facebook tracker, for example?
Firstly, we’re independent and unbiased. Unlike the other players in the industry, we aren’t selling media. When you measure and sell media, there’s a conflict of interest, just like an estate agent selling a flat, but also working on getting you to buy the flat and thus getting the bonus at both ends. Therefore, we picked a side. We only get paid by the advertiser; we aren’t selling to them as well. We’re not selling for the ad networks, if we get offered revenue share from them for example, we refuse it but ask them to pass this on to our advertiser. We’re actually welcoming more innovation from the ad networks such as eTargeting for example. A few ad networks are getting into mobile eTargeting and we’re supporting this, although it takes time.
What needs to change within the current analytics services?
In the industry, there are a billion things we want. For the first time, the advertisers are thinking about where the data is; considering how to separate measurement and the ad networks. We’re looking into eTargeting which is a newcomer to the mobile advertising industry. It’s not as easy to calculate the ROI for engagement ads. Do you open a one day attribution window? A seven day one? Is it the lifetime value again? How do you measure that? That’s what we’re currently working on this.
You recently announced Appsflyer for agencies. One of the features is optimisation of the budget in real time. Can a system distribute the budget automatically or should marketers do it manually?
Definitely manually. However, media agencies that are playing in the exchanges and doing algo-buying within different exchanges can plug into our API’s and get conversion data, the cost, the CPI and the cost per specific action in real time to back feed their algorithms. Again, we’re not handling the budgets and the money is not going through us. That is what the market has to offer in different DSP’s and exchanges.
What do you see happening in the next years for Appsflyer?
In five years we would like to be the industry standard for mobile attribution so it will be a platform where both advertisers and ad networks can trust the data both for optimisation and performance marketing billing settlement.Both parties need to be able to trust and use a platform that can actually tell them what the performance is. As for the next big thing for us, that’s a surprise. I can tell you one thing though; it will be in the toolbox of every mobile developer!
This is an abridged and edited interview from the AppInTop mobile app marketing podcast produced by AppInTop, an automated mobile app marketing platform. Listen to this podcast episode or subscribe via iTunes.