Analysts from AppInTop told in their column on App2Top.ru why, when buying traffic, you do not need to look towards the CPI.
The CPI model itself is not perfect, but in the current realities it is not viable. Why? Let’s figure it out.
As a rule, traffic can be bought either with a pay-per-click or with a pay-per-install. It would seem that it is better to pay for the installation, because you pay not for clicks, which may not lead to installations, but for the result.
But there is a nuance.
By purchasing CPC traffic (per click), you buy only advertising traffic (the one where installations occur due to conventional advertising: banners, videos, and so on). Of course, no one guarantees that its source will lead to an increase in downloads, because there are a number of other variables here: a lot depends on how “tasty” the banner turned out to be, how many “screens” between the banner and the application. But you do know where it is shown, how often, and so on.
With CPI (per installation), not everything is so smooth. The fact is that you can get a guaranteed CPI only if you buy motivated traffic.
This, we recall, is the one where installations are made by users performing an action for a reward (a reward in another game, a small amount of money). The application itself may not interest them much, the main goal is to get a reward. Such traffic is used only to raise the application to the TOP.
In other cases, the CPI will have a floating (read, very high) price, since the conversion of people who clicked on the application banner into its users constantly depends on a huge number of variables.
As Wooga ‘s lead analyst Eric Sefert wrote:
“Advertising networks cannot guarantee the exact number of installations at a given price, because they have no idea how many views sold at a fixed price are converted into installations. An advertiser can buy a certain number of views at a specific price, but the number of installations with a given number of views is unpredictable, it can only be predictable.”
For this reason, if they look at CPI, then post factum, not as an applied value, but rather as a kind of thermometer.
However, many still tend to buy installations, so the following situation often develops in the market:
A developer comes to the market. He orders 1000 advertising installations per day at a price of $ 2 in the CPI network.
CPI network takes the order and distributes it to its partners (minus 30% commission already at $ 1.4). Partners purchase motivated traffic for $0.4 from their partners, who pay $0.2 of remuneration to their performers.
As a result: you received 1000 installations for $2. CPI network earned $600, the partner of the first level earned $1000, the partner of the second level earned $200, all performers earned $200. Everyone is happy and happy!
The only downside: instead of users, you got empty installations, which are unlikely to lead to organic traffic.
Conclusions: if you are unable to distinguish a motivated installation from an advertising one, you should not buy a CPI.
But what if you really want to use the CPI model anyway?
Behavioral indicators of attracted users should be monitored. This requires the use of analytics systems, for example, Mixpanel, Adjust or Flurry. They allow you to calculate:
- ARPU — average earnings per user;
- LT — the time the user spent in your application;
- Churn Rate — user churn rate;
- playing or not playing;
- passes registration or not.
And other indicators that allow you to identify “live” users.
If the majority of users log into the application and spend no more than 30-60 seconds there, then this indicates that they were brought there by motivated traffic.
In addition, advertisers should be alarmed by high conversion rates from clicks to installations. For example, on Facebook they rarely exceed 15-20%, on Google AdMob — 1-2%. Therefore, the conversion rate in an unknown advertising network at the level of 10-20% should make you think.
By the way, on July 2 (Wednesday) from 12:00 to 13:00 Moscow time there will be a webinar from AppInTop dedicated to buying traffic. .