How does advertising monetization work? — Part I

This article is devoted to how advertising networks function from the point of view of game developers who place ads in their application. In it, we do not touch on the aspects of promoting and buying traffic.

1. How do advertising networks function for a developer who places ads in his game?

Working with an advertising network for a developer is usually divided into the following stages:

• creating a game;
• thinking through the advertising inventory (this is the name of the places where advertising will be running);
• integration of the advertising network SDK into the game (it is thanks to the SDK that ads will be automatically inserted into the inventory);
• getting revenue from ad impressions.

2. How does advertising monetization work in general?

Four sides are involved in advertising monetization:

• developer;
• advertising grid;
• advertisers;
• users.

Their interaction in the most general form looks like this:

• advertisers give ads to the advertising network (various creatives — banners, videos, interactive formats) and pay it on various terms for their placement;
• the developer, in turn, sends a request to the network to display ads (in other words, asks to send ads);
• in response, the network gives the developer one of the available ads, as well as part of the amount it received from the advertiser for advertising (usually a small percentage of this amount);
• as a result, the developer shows the resulting creative to one of his users.

Important: the one who requests advertising for display in mobile marketing is called a publisher, that is, the party that published the creative.

A primitive scheme of advertising monetization. The network receives an ad request from the developer and gives him a suitable impression from its advertiser (AppQuantum)

3. Why not work with only one advertising network?

There are two main problems when working with a single network:

• a) When only one network is built into the game, all advertising impressions go to her, because she is an exclusive supplier, a monopolist. This means that without the absence of any competition, it begins to take a larger and larger percentage from each impression, up to 75-90%. At the same time, the developer does not even suspect this, because he does not know what margin the network takes. He cannot compare the payment with the market value, because other networks are not connected.
• b) Each advertising network is limited to a relatively small number of advertisers. The developer is interested in the widest possible coverage, because in this case he is more likely to come across the advertiser who will be willing to pay the highest price for the impression.

4. So, maybe you should just connect a couple more advertising networks?

Starting to embed the SDK of advertising grids on your own alone is not the best solution for several reasons.

• First. Technically, they can conflict with each other.
• Second. Without comparing the revenue per impression from each network, the developer will not know which network is currently offering a better offer. It turns out that any display can go into the inventory, not necessarily the most expensive.

It is for this reason that it is customary to work with mediators.

5. What is a mobile advertising mediator?

A mediator is a service that:

• a) includes a group of different advertising networks;
• b) ensures that there are no technical conflicts between them within its own SDK;
• c) and, most importantly, distributes impressions between networks, which ensures competition.

You can write the logic of distributing impressions between networks yourself (you will get your own mediator), but it takes a lot of time, plus you need a deep understanding of mediation and a willingness to independently keep the SDK of different networks up to date.

This also makes no sense, since there are a lot of third-party proven solutions on the market today, for example, Google AdMob, ironSource, AppLovin Max, MoPub, and so on.

6. How do networks treat competition with each other?

The networks do not directly know that they have entered into competition (except for the network whose mediator you use), but they can guess from indirect data.

This is a common practice. In addition, it allows not only to get the highest price for impressions on the market, but also to “knock out” an increase in the share of advertising from network managers.

The fact is that the networks are primarily important volumes. They want to unscrew ads and give advertisers installations. Therefore, the networks are ready to sacrifice their commission for the sake of increasing the volume of impressions.

The “knocking out” of the share usually works according to the following scenario:

• the number of impressions that the grid buys from the publisher is called;
• then the number of impressions is called, which the publisher is additionally ready to sell to her;
• in the event that the additional volume is significant, the advertising grid usually meets halfway and increases the share of the publisher.

Important: although the network pays the developer for the impression, the advertiser can pay both for the impression and for the installation, click or conversion, that is, this system is complex, it has a place for both very large and negative margins.

7. How exactly is the competition between networks in the mediator?

A little earlier we mentioned that the most important function of the mediator is the distribution of impressions between advertising networks. Distribution can occur according to several principles.

The principle of “waterfall” (waterfall)

(I) Description of the principle

Waterfall is a list of advertising networks that are interviewed in turn for the availability of suitable advertising.

Surveys of advertising networks are called “calls” (call). The number of calls to the waterfall depends on the mediator server. The standard is about 20. Some create their infrastructure in such a way as to support 100 calls each.

By default, the mediator optimizes the order of the networks in the waterfall itself based on the impression data. Networks also give impressions for the price they consider necessary. The developer does not know if this is the best price.

To optimize the process, you need to set the price of each impression for the network yourself, and in the mediator — the order of network polling.

II) Setting up a “waterfall”

When setting up the publisher specifies:

• desired CPM for each of the networks;
• arranges the grids in the order in which the mediator will poll them.

Note: in advertising, the term CPM refers to both the price paid by the advertiser for placing an ad, and the income that the publisher / developer receives from the network.

In this article, we are talking about CPM only in the context of advertising revenue, so when setting up in waterfall, we mean that CPM (cost per mile) is the income that a publisher receives from the network for 1000 ad impressions in his application.

Why 1000?

It is much easier to estimate income in dollars than in hundredths of a cent.

Important: by setting a price for 1000 impressions, the publisher allows the network to redeem individual users for different costs, on average receiving a set value, except in cases with hard floor, which are discussed below.

III) Deep tuning of the “waterfall”

When setting up, the publisher puts up several price tags at which he is ready to sell traffic. Each must have a price specified.

The cost of each call to the network is of three types:

• Hard floor (we will continue to call them “flora”) — advertising is given clearly according to the CPM specified by the developer or higher. The most disliked flor in networks is often coordinated for weeks at the level of an HQ company.
• Soft floor — CPM in a spread from the one specified by the developer: it can be either higher or lower. But, on average, the result will be lower. Most often, the networks work precisely on soft flora.
• Target floor — an advertising network can advertise both above and below the specified CPM, but on large volumes, the average price will have to fall within the specified values.

After installing flora, this call is added to the mediator, where the search for a suitable offer begins.

In fact, the mediator simply goes through the advertising networks for each call in turn, asks them if they are ready to give advertising for a set price.

For example:

• mediator offers the advertising network “A” in waterfall call for CPM $150;
• if the advertising network is not ready to buy advertising inventory for such a cost, the mediator shows her the next call, whose price is less, for example, $ 140;
• this happens until the cost for which the network will be ready to buy traffic is named.

The developer decides which step to set between calls — $10 or $1. But here it is important to keep in mind two things:

• if you make a big difference between calls, for example, at $10, then you can miss some offer for $148;
• if the difference between calls is too small, then polling all networks will take too long, as a result of which the user will simply close the application after a long wait for advertising, and you will get nothing.

Example of a waterfall

The trick is to set up a waterfall with the maximum number of calls to different networks in the range where the user is most likely to be given the impression, but not lose in its effectiveness. To do this, waterfalls are configured taking into account many factors: OS, GEO and others.

At the same time, if you set the price too high, there is a risk that no network will buy the impression.

You can solve this problem with presets — sets of certain calls to certain networks with requests that are most suitable for these networks. This allows you to call the network only at those moments when it can buy an impression. Thanks to this, you can get 15-20% more income, due to the fact that you do not waste space in the waterfall on “empty” calls.

The principle of “Bidding”

I) Description of “Bidding”

Above we described how network polling (“waterfall”) works in the mediator. In fact, this is a bruteforce of networks with a search for calls until a suitable advertisement is found (that is, one that will suit the price tag for your traffic)

An alternative to this approach is bidding (from bid — bid). When working with it, you do not need to prepare a lot of calls. The call goes one to the grid.

II) Principles of bidding

There are two types of bidding:

• standard bidding is when an advertising network requires you to run an ad display and only after its results announces the amount that it will pay;
• real-time bidding (also known as RTB or real-time bidding) is when the network is ready to announce the amount before the show, in response to which the publisher sends her a suitable call.

The principle of “Hybrid waterfall”

Mediation may consist of:

• from one waterfall;
• from bidding;
• from a waterfall and a bidding (then such a waterfall is called a hybrid one).

A hybrid waterfall is a set of calls to bidders (from bidders — bidders, advertising networks here) and a regular waterfall, in which networks are polled.

The mediator polls bidders (here RTB), what price they are willing to pay for the display, and selects the highest result. The waterfall is being polled at the same time. If there is an advertisement more expensive than bidder’s, the display goes to the network from the waterfall that named the highest price. If the price of the waterfall goes below the price that the bidder is willing to give, then the display goes to the bidder.

An approximate scheme of operation of a hybrid waterfall. For clarity, a sequential survey process (AppQuantum) is described here

8. Optimization

Working with mediation requires constant experimentation, testing new waterfall setups to understand which one works better.

Waterfalls are optimized by split tests (A/B testing). Usually mediators have tools for this (at least they allow you to compare two waterfalls with each other).

But even on large volumes, the results are not always obvious and logical. Therefore, it is important to evaluate them at a lower level than comparing the metrics of the two groups. Manual evaluation of the results gets an advantage due to the quality and time of the tests without having to wait weeks for a ready answer.

Split tests should be carried out not only on new games, but also on long-optimized waterfalls. For example, on Idle Evil Clicker, we ran a test in the USA and compared the current waterfall and a longer one, to which we added 9 new positions. These positions were selected based on the performance of the control waterfall. We needed to make sure that the new positions would work with the efficiency that we expect from them, and would not cause a strong increase in the delay time of impressions.

According to the results of the split test, a longer waterfall showed greater efficiency and increased rhubarb by 15% with a small loss of users. Further optimizations, taking into account the new balance in the waterfall, made it possible to reduce this loss. Conclusion: it is not enough to optimize the waterfall once: you need to constantly look for new, more effective setups.

9. The future

Now the daily setup and optimization of mediation can be very tedious and take a lot of time. But the more networks will switch to bidding, the less time you will have to spend on optimization. And, perhaps, in a couple of years — this problem will completely disappear.

10. The result

Finally, let’s run through the main theses:

• to successfully monetize your application, do not choose only one advertising network — it is better to immediately install a mediator and connect several networks;
• it’s better not to write your own mediator — use ready-made solutions;
• build your waterfalls, do not neglect experiments, look for opportunities for optimization.

11. What’s next?

In the second part of the article, we will talk in detail about the process of setting up mediation from a technical point of view.

12. Terms

• Mediator is a service for monetization of your application, which you install into it using the SDK, which allows you to distribute impressions between advertising networks in order to increase profits.
• Call (or call) — a request from the publisher to the advertising network for the availability of advertising at a set price until someone gives it away.
• Bidding is a single call to the network with data on the user, relative to which it is assumed that the network will give back the most expensive impression.
• Waterfall (or waterfall) is a list of advertising networks that are gradually interviewed by a mediator for the presence of advertising at the requested price.
• Hard floor — advertising is given at the CPM specified by the developer or higher.
• Soft floor — CPM in the range specified by the developer (can be either higher or lower).
• Target floor — an advertising network can advertise both above and below the specified CPM, but on large volumes, the average price will have to fall within the specified values.
• Setups (or presets) are sets of calls to bidders and a waterfall, in which networks are polled.
• CPM (cost per mile) is the revenue that you will receive from the network for 1000 ad impressions in your application.

---