Even though the word incent is often misused, its underlying principle is straightforward: you want someone to perform an action and you reward them once it’s done. Applied to mobile user acquisition: you need a user to download your app and you’ll reward them for doing so. The complexity comes after, as the quality of the users you’ll gain from incentivized campaigns will depend greatly on the nature of the incentivized action versus the prize you are offering.
Regardless of the different methods, incent campaigns all show similar characteristics that make them attractive for UA (User Acquisition).
- They are cheap compared to more traditional UA campaigns
- They are efficient. Performed well, they can deliver several thousands of users to your app per day
However, nowadays, these campaigns are often perceived as fraudulent and we endeavor to understand why.
Incent as a means to elevate app store rankings
Up until a few years ago, incentivized campaign were extremely popular. They used to play an essential role in any UA campaign. The reason? When used well, they could push your app to the highest rankings on app stores.
The Google Play and iTunes store ranking algorithms used to be elementary compared to what they are today. Using incentivized inventory to acquire several thousands (or hundred thousands depending of the country) of users the first week of your app launch was a guaranteed top 5 spot in your app category and thanks to this spot on the stores which brings a lot of exposure,a great amount of users were driven to download your app organically. Those organic users have especially valuable LTV and they make the seemingly low quality incentivized users ROI positive in the long run.
Today, the efficiency of this type of campaign has been neutralized majorly by the new ranking algorithms deployed in the main app stores.
Incentivized campaigns done right
Once the app stores changed their algorithm, incentivized campaigns grew out of fashion and left very few advertisers investing in them. As a result, all the industry’s incent specialists had to pivot to implement performance based user acquisition campaigns. This development is the key to nowadays good and bad incent campaigns.
Prior to this transition, incentivized campaigns were performed to stimulate the maximum number of people to download your app in exchange for a reward. This was the most effective way to drive volumes, and even though the LTV (LifeTime Value) of the acquired users was extremely low, the campaigns where valuable.
Some of the actors of the old incentivized industry therefore decided to change the rules to make sure their campaigns would remain valuable for advertisers. The new trend became to reward a user for discovering a new app rather than installing it. The key here is that the reward is handed over for learning about your app, nothing more. This can be done through several techniques but the most popular are rewarded video and playable ads, the reward often being some sort of in-app currency.
A potential new customer is compensated to watch a 15s or 30s descriptive video about a new app. If the user is interested, he will download it. If he is not, he will simply continue onto the next video to gain other prizes.This ensures that you will only pay for users genuinely interested in your app. The acquired users through this process show high quality and display a profitable LTV.
This new incent technique, if done right, can be very successful. Nowadays, some apps are created solely for the purpose of offering rewarded playable ads in exchange for virtual currency. Lately, there have even been discussions about creating a cryptocurrency dedicated to this purpose.
Unwanted incentivized campaigns
Up until now, we saw all the good faces of incent. One can think that incentivized campaigns are actually not deserving of their poor reputation. However, we’re going to dive into the dark world of unwanted incent and explore the keys to understanding it.
First, the reward. We saw that an incent reward is often a virtual currency. However, it’s nature is only limited by someone’s imagination. Some apps reward users with real money to download and even interact with your app. Let us take an example. We are doing user acquisition for a casual gaming app which pays it’s partners 10$ CPA for a level 10 reached. A publisher will offer, through an incentivized app, up to 5$ to each users that downloads your app and plays with it up to the level 10. In this scenario, users might get addicted to the game after playing or might simply switch on to another game to get more rewards. The key here is transparency. As long as the advertiser is aware of the method used to get the users to download it’s app, there is no reason to worry. But when this method of acquisition is hidden in the middle of a classic user acquisition campaign this method can be judged as fraudulent.
Most worrying, incentivized campaigns can be turned into extortion-like campaigns by extremely unscrupulous publishers. The user will not be incentivized to download your app, but forced to do it. This is done through malwares that will infect a phone and block a specific content (snapchat account, pictures etc.). The user is then forced to download one, two or three apps to unlock its phone and regain ownership of its content unharmed.
Finally, the transparency. For some actors,the most lucrative use of incentivized traffic is to disguise it as regular performance based traffic and sell it at a high CPI. This is the most common form of fraud associated with incentivized traffic. In order to cloak it as if it’s legitimate traffic, the fraudsters are using various techniques.
Firstly, mixing valid traffic with incentivized traffic to grow volumes safely. This method is usually used by networks that have multiple sources of traffic. They will use a large amount of regular traffic and a small portion of incentivized traffic to remain undetected. The issue is that the advertiser will be paying for conversions that are not driven by users interested in using their app. These kind of techniques are widely spread but can be easily detected using in depth analysis of in-app activity and transparency at the sub-publishers level (as detailed below).
Secondly, fraudulent publishers will try to incentivize users not only to download the app, but also to perform some sort of in-app activity to mimic legitimate traffic. For example, the promised reward can be given once the user plays the game for a few minutes, reaches a certain level or simply registers. That way, the fraudsters insures that their incentivized traffic will remain undetected if the advertiser is looking at simple in app event metrics.
How to fight it
The first step to fighting it, is simply to detect it. Once it is clear that a publisher is running unwanted incentivized traffic on your campaign, it is relatively easy to get a refund and blacklist it.
In order to detect it, we are going to focus on the user journey from the incentivized banner to the install or in app activity. This journey will translate into specific key metrics that shows very recognizable patterns. Once these patterns are known, we can detect incentivized traffic inside any campaign.
- The journey from the click to the download. Users that see the incentivized banner and are interested in the reward will click on the add and download it. Because the users will want to download the app solely for the purpose of the reward, there will be very few people quitting between the click and the download. This will result in an abnormally high conversion rate (CR) also known as click to install ratio (CTI).
- Journey from the click to first app open. Here again, users are seeking a reward, and it’s unlikely that they will quit before getting it. The reward is only given once the app is open for the first time (equivalent to the conversion time of the tracking platform). Hence, users will want to open the app as soon as possible after the download to make sure they get the reward straight away. This will transfer into a very short time difference between the click time and the conversion time (a.k.a first app open time).
- Finally, after the app is open for the first time, there are two scenarios which will translate into two different patterns :
- The user is solely rewarded for the install. They will likely delete the app after the first app open. No real in app activity of these users will be recorded and their quality will be extremely low. Hence in-app event KPIs will be low or nonexistent.
- The users is rewarded for the realisation of a particular in app activity. Let’s take “Registration” as an example. The install to registration rate will be very high for this type of traffic. However, once the required action is performed and the reward collected, the user will tend to delete the app. This will transpose into several metrics: The install to action ratio will be high, the install to action time delay will be low and finally, there will be very low in app activity after the first action is performed.
To summarize, in order to differentiate incentivized campaigns from non incentivized ones, all those metrics should be audited. One KPI alone is not sufficient to determine for certainty that a publisher is performing unwanted incent. However, taken together, those metrics are the fingerprint of incentivized campaigns.
Example of incentivized campaign analysis
Here, we will show in detail the technique to effectively detect incent coming from a sub-source hidden in one of your performance based UA campaigns.
First let’s take a look at the CR/CTI of all the sub-sources of a campaign. It is the most obvious metric and a good way to filter out the suspicious and non suspicious sources.
Table 1 : Campaign X: Conversion rate per source and sub-source
Thanks to this quick analysis we can detect three sources with abnormal CR. This shows that we need a deeper look to understand the nature of this traffic. We are now going to use the mean time to install metric. This will allow us to observe the average delay between the click and the install for those sources.
Table 2 : Campaign X: CR and Mean time to install (a.k.a. MTTI) per source and sub-source
The table shows that out of the three suspicious sub-publishers we identified, two seem to show a very low MTTI compared to the others. This allows us to understand that sub_source A_2 might not be as fraudulent as we originally thought. It doesn’t mean however that it is clean of all suspicions and a follow up conversation with the source is therefore necessary.
Now, for the two highly suspicious sub sources, A_3 and B_2, we are going to perform further analysis to be certain of their nature.
Table 3: Campaign X: CR, Registration rate, mean time from install to registration and level 5 reached rate per source and sub-source.
The in-app activity of the users allows us to gain a deep insight into the sources. For example, source A_3 is flagged on all our KPIs for incent. If the analysis is done halfway, for example, looking only at CR and Registration rate, source A_3 appears as very healthy. It is only through a deep dive into several performance metrics that we can detect fraudulent activity. Fraudulent sub-publishers will always try to hide their traffic. For example, looking at the metrics of B_2, we can highly suspect that they are mixing incentivized traffic with healthy traffic.
As always with fraud, one metric and one benchmark is very inefficient. If we take the example of source A_3, it’s CR might look suspicious at first, but all the other metrics look normal. The high CR is not always indicative of incent. This applies to all the other KPIs in this analysis. It is only through multiple data points that we can understand the true nature of the traffic.
Once you are confident in the quality of the activity of each sources, it is much easier to go back to them and get refunded for all the unwanted/fraudulent traffic that you detected.
The future of incent
As mentioned at the beginning of this post, the underlying principle of incent is basic. The different uses of incent are only limited by the different actions and rewards that exists, and they are infinite. The key to performing clean incentivized campaigns is transparency. In the future incentivized campaigns might attract more and more advertisers and become once again the trend in UA. For example, in the APAC region, these campaigns can perform extremely well as users are very familiar with the concept of getting rewards for actions performed online. Then again, the increasing amount of fraud in classic user acquisition campaigns may push advertisers to look into other types of user acquisition campaigns as well.