Mining and processing of gaming traffic
Introduction
In 2007, in order to optimize the advertising of our games, an advertising rotation engine was created. So our advertising network was born. It quickly grew into a large banner network focused exclusively on advertising online games and near-game resources. Today we want to share our observations and experience. We hope that the material will be useful to young teams, but I will be able to learn something interesting from experienced colleagues.
The material is devoted to the following topics:
')
First, you should go through the analysis and processing of traffic, because before you get traffic, you need to understand what kind of traffic we need. And how much it should cost.
Obviously, there is no point, for example, to drive tons of expensive, super-high-quality traffic into the game, if it cannot keep users, or if its site cannot convert traffic into registrations.
Before determining the cost and quality of the traffic that we need to purchase, it is necessary to determine and eliminate the main thresholds. Otherwise, you simply do not distinguish good traffic from bad.
Let's take a visual look at the entire path of the traffic, starting with the transitions to the game site.
For example, suppose we drove 100,000 users to the site from a high-quality, normal source to our landing page.
LANDING PAGE is a page where users come from traffic sources. For a browser or client game, this is usually a promotional page, from where the user goes to the registration page. For social projects, this is the page that the user sees before the game installation. Consider the numbers on the example of a browser game.
The normal percentage of conversion of users who come to the landing page in the registration is from 15 to 30% for browser projects, depending on the traffic sources. For games with a downloadable client, the rates are naturally lower. For social programs above, 50-80%. The number of losses at this stage, if you do not consider the quality of the traffic itself, depends on two factors:
To check the quality of the landing page, make a test purchase of traffic from various sources, as good as possible. This check is not suitable for traffic from offers in which users receive buns in other projects for registration in your game - the conversion will be 40-45% for the browser and close to 100% for social projects and it will not show the quality of the landing page.
Examples of the perfect Landing Page, see Blizzard. They have all projects are about the same made, and this is not by accident.
In fact, this is the most important page of your project. At this stage, with incorrect landing page conversion can drop down to 4-5% with the same quality of traffic, which would give 30% on a good landing page for the browser. The only purpose of this page is that the user wants to register in the game. You do not need to load it with anything else, it still doesn’t care about the ratings in the game, it still does not understand the gaming news. All he needs to know by and large is a brief info about the game, screenshots, a movie, and a big “PLAY” button. But he needs to give it beautifully, so as to achieve only the wow effect. In other words, you should not save on a landing page, because it is fraught with losses of tens of thousands of dollars for poor conversion.
If the quality of traffic is low, for example, a "click call" (about it a bit lower), then a good landing page will give 1-1.5% conversion. Strangely enough, sometimes it is not always a tragedy, if the traffic is cheap enough so that the cost of registration ultimately suits us. At the same time, a bad landing page will not allow to work with click-type traffic in general, since conversion will tend to zero.
Plus, an inconvenient or poor-quality registration form can easily lose up to 95% of users, tested by bitter experience.
This problem is not worth it for social networks, where registration takes place in 1 click by the means of the network itself, and where the landing page is often essentially a registration confirmation page with screenshots of the game. We are now talking about the registration on the website of a browser or client game, where the user gets from the landing page.
Consider an example of registration from one game, we will not point the finger which one. To register in “that game” you need to enter your nickname in the game, mail and password. It seems to be all simple and standard. The user fills out the form, presses okay, and then he is told that the password must be at least 6 characters, with at least one of them must be a number and one capital letter. The user came up with a password, entered it twice, clicked OK.
And here he is told that the nickname he entered is busy.
The programmer thought why to pull the server and first check the nickname, if the password can be checked right on the page for validity. And, of course, the password field is cleared. That is, the user must pick a nickname, each iteration by entering a password of at least 6 characters in length. Double enter.
The easiest way to measure the effectiveness of a landing page is to put a Google Analytics counter. Set goals for the registration page and the page after registration. Some projects have a registration form right on the landing page. This is normal, but in this case it is a little harder to keep track of those who have fallen off due to the inconvenience of the form.
The next threshold, at which another 90% of users of an average browser project fall off, is the conversion from registered users into active registrations, and from them into payers. Active in this case are those who did not fall out of the game after logging in until they reached a certain control point, for example, level 2, or did not complete the training course.
The quality of traffic greatly affects the loss at this threshold. For example, when buying traffic from offers in which payment is made for registration, and not for user actions in the game, the conversion of registered users into active registrations may not exceed 1%. Only 5-10% of the remaining users become payers (again, on average in the hospital).
See what happened to our 100 thousand users. I intentionally omit other thresholds that are not affected by traffic quality.
In order to adequately assess the quality of traffic, it is necessary to eliminate these basic thresholds. At the same time, one should be aware that they should be monitored constantly and best of all - automatically.
Ideally, after the elimination of the main thresholds, their daily monitoring should be done by the machine, i.e. script on your server. If the indicators suddenly go beyond the specified values, then a notification should be sent to you. Otherwise, there is a chance to simply not notice how the indicators have changed at some point. And only after a week to see that several tens of thousands of dollars from the advertising budget were merged, because all users simply were not kept in the game due to some minor problem.
Correct traffic
After the main thresholds in the game are eliminated on the test traffic, you can measure the real ARPU.
The correct traffic is traffic, the cost of active registration of which is lower than the ARPU for the period that you consider normal for the payback of advertising in your project. Most often this is 1 month.
Not being able to see the separation of ARPU by users from different sources, you will not be able to adequately estimate the traffic from these sources. Especially it concerns the purchase of traffic with payment for registration. Be sure to ask the programmers for such a table in the admin of the game, where you could compare the conversions of users from different sources at the main control points and ARPU.
Now let's talk about the methods of extracting traffic, what works, what doesn't, and what quality traffic gives.
"Popander"
Let's start with the "pop-under" of advertising, as with the very "trash" method.
“Popander” - these are banners that crawl over the content of the site, often in the middle of the screen, have annoying content and a “close” cross. Nobody calculates that the user clicks on the banner itself, it is expected that the reaction to closing the banner is expected, and by clicking on the cross, a transition is made to the advertiser's site, which has nothing to do with the banner. A click is needed to bypass the browser protection to open new windows with scripts, this is called “breaking through”.
"Click-under"
Separately, we say a few words about the "click-under." This is a format similar to the “pop-under” in that the advertiser's site opens in a separate window under the current one, and is visible to the user when he closes the browser. But unlike “under-ander”, for “punching” a user click on any place of the site is used, and not a pop-up banner in the pop-up. This is not so annoying to users and generally gives a good result if the platform is thematic and the audience on it is gaming. At the same time, the role of the banner itself is essentially the Landing Page, which opens in a separate window. It should be noted that the quality of one format does not play such a role so much as in the case of a “click-under”
The undeniable advantage of the traffic coming from this format is its cost. Near the cent for the transition.
We put several experiments on this topic in different games, including with the downloadable client, and we managed to achieve conversion from this format in 2-3%, i.e. at the price of a switch to a cent, registration cost 30-50 cents. Normal, valid registration.
The disadvantage of this format is that large thematic areas still do not seek to place it in their own. At the same time, the volumes and coverage should be really large in order to provide the sane number of registrations. It makes no sense to turn the clicker on one site for a long time, because he quickly burn out. If it is even simpler: when most of the regular audience of the site has already seen your site in the window of a clicker, there is no longer any sense to turn it on this site. And this in turn means that such traffic is usually not very much.
We recommend using "click-under" as an addition to the main advertising campaign. From experience, it is also noted that clickkander works well together with a strong banner on the same site.
Text advertising
It also shows itself very well as an additional format - plain text ads placed on thematic sites. We consider this format unfairly undervalued. For example, on one of the top sites of our network, a block of text ads generates about 10,000 clicks per day. At the same time, the quality of the traffic received in this way is very high and is not inferior to the quality of traffic with well-targeted banner advertising.
Branding
Separately, let us say a few words about a promising format that is gaining popularity lately - branding. In the simplest version - this is the wallpaper on the right and left of the site content. The price of a click is higher than that of a regular banner, but at the same time:
There is no sense, of course, to brand very small or non-thematic resources.
Offers
The user is offered (in a game or application) to go to another game, do this and that, then come back and get a bun, for example, some object or game currency.
Offers work relatively well in social networks, within the same social network. As for the traffic from the offers to browser or client games, it most often comes across of poor quality.
Sometimes a really mad conversion in registration is made from the offers, if a bonus in the offer is given for registration. If the bun is given for reaching the level - then users swung to this level with a specialist to get a bun, and then they will score for the game. The game remains a tiny percentage of users. The specificity of this traffic is such that users who come to you because of a freebie in another game, so as not to spend 10 rubles on it, are not particularly inclined to pay.
With offers you can collect traffic, but again the question is in the cost of such registration. In any case, before working with an offer, you need to test its traffic before converting to payers and then monitor it tightly. An offer with payment after some actions in the game (“reach the level”) is much more effective than payment for registration. More users are drawn into the game, but still quite a few remain.
But in fact, it is much more sensible to make your affiliate program, with a meager payment for registration, such as 1 rub / registration, but with a percentage of user payments, and offer it to various traffic sources. Then, in general, you will not care how much trash you drive traffic and how specific partners mine it. If the affiliate is promoted, it is a very good and stable addition to the main advertising campaign. But it may never become the main source of traffic; it should be regarded as a good additional source.
Banner advertising
Well, the most effective and most diverse way to extract traffic is banner advertising. There is an opinion based on personal experience that it is on banner advertising that the main advertising campaign should be built. The main thing in the banner is optimization. This topic is quite extensive and it is better to consider it in more detail.
The main enemy of quality and cost of traffic is the quality of the banner itself.
The banner has two main indicators that interest us: the CTR and the quality of the generated transitions.
CTR is the ratio of the number of clicks to the number of banner impressions in percent. The concept of a good CTR varies depending on the site and banner format. For example, when placing a small banner in a sidebar on Facebook or “vkontakte”, the average CTR is one-hundredth of a percent. At the same time, there are sites where the normal CTR is more than 1%.
There are two fundamentally different approaches to the creation and placement of banners. The first approach is focused on raising the CTR by any means, i.e. "Let him click as many people as possible, no matter what." Let the conversion be very bad, but the clicks will be many and cheap. Such an approach is often justified when traffic from the site is already trashy, poorly targeted, and of poor quality. But there are a lot of hits. In this case, the landing page already plays the role of a filter, converting around 1%. But nevertheless, due to the cheapness of the traffic, the cost of registration can be kept within reasonable limits, and these are much more honest and high-quality registrations than from the pay-per-offer offers.
The second approach, as it is not difficult to guess, is focused on ensuring maximum quality of traffic. Those. People who clicked on the banner are already interested in the product. The banner corresponds to the essence of the product, the placement is targeted at the maximum target audience. At the same time, naturally, a lot of effort is applied to also increase the CTR of the banner.
But a banner that gives super quality traffic, but with a low CTR, is the same file as a very comfortable car, but without an engine.
Few observations: animated banners give an average of 30% on average higher CTR than similar in quality static. Video Banners provide 20% more CTR than just animated. This means that you can get one and a half times more traffic for the same money.
If half of your advertising budget is more than the amount you need to spend on good video banners, make them.
Video Banner is a flash banner that is first loaded as a regular banner, but after loading it loads the video insert and plays it inside. Thus, the banner itself loads fairly quickly, it is no more than a regular flash banner, i.e. about 50-70 kb. Instead of a video, while it is being loaded, a freeze-frame, static picture is shown.
Perhaps there are only two reasons not to use video banners: either you do not have the budget for creating a sane video for a banner, or the specific site where you want to place it does not allow you to do this due to some technical reasons.
If you have a beautiful in-game video or, better yet, a video trailer, just cut a few seconds of a connected video out of it that can be played in a circle.
Total: Do not save on banners, as well as on the landing page. Quality advertising materials are our everything.
Despite all of the above, you need to understand that there is always a chance, much more than zero, that your super banner, on which you had hopes, took and did not shoot. Those. showed so low CTR or poor quality of traffic that the registrations received from it simply do not pay off.
The conclusions are simple:
In addition, there is the concept of "burnout" banner. This means that the same banner that hangs on the same platform loses the CTR over time. At the same time, it does not matter how many different banners we will launch on the same platform in rotation at the same time. They will simultaneously lose about the same CTR over time.
On average, by experience, one banner on one site lasts about 1 to 3 weeks. Further, it becomes ineffective, and it is better to replace it with a new one.
Therefore, when planning an advertising campaign, it is necessary to stock up with a certain number of banners for future use, in advance, and taking into account that 2/3 of them may have to be thrown out.
Naturally, there are exceptions, for example sites, where either the majority of the audience is search traffic, or where the audience is incredibly huge, but visits the site rarely and irregularly. For example, torrent trackers and file sharing. Unfortunately, the traffic on such sites is not thematic, i.e. It can be targeted only by region, as a rule, but not by hobby for games. eg.
You can extend the life of banners by using the services of a large banner network where your banner is:
Also, the above allows to significantly increase the CTR, and thus reduce the cost of advertising. For example, in the course of experiments, we derived the following rule: the optimal number of banner contacts with 1 unique is 3 shows per day, no more than 7 shows per week, no more than 14 per month. A similar tuning of the shows gives a stable + 50-60% to the CTR for a good, correct banner, if you apply this restriction on sites with frequent attendance and a stable permanent audience. And at least + 20% on the remaining sites.
Monitoring and analysis
However, when placed on a large number of sites, the question of monitoring and analysis is very acute.
Firstly , even different banner places on one page of one site can give completely different CTR for the same banner. But the placement on them may cost differently, and it’s not a fact that placing expensive on the CTR site itself is more profitable than, for example, on the second pages, but cheaper.
Secondly , different sites give different CTR and quality of traffic from the same banner, due to slightly different audiences for example. Somewhere among the core of the audience of the site there are more fans of browser strategies, somewhere client-side games.
All this translates into the fact that manually optimizing impressions on a large number of sites becomes more expensive than the gain from this optimization, or even impossible.
Here are a few exaggerated, but life options, showing what this can lead to:
Option 1 . The marketing department completely abandons self-advertising, looking for partners who are willing to sell conversions (i.e. ready traffic) or registration. At best, they give their promotional materials. They ask partners for test traffic, if they are satisfied with the quality of traffic, then they continue to buy registrations. All forces are concentrated on finding sane partners with normal traffic.
As a result, partners are driving first normal traffic, using the materials of the advertiser. Then they burn out, CTR falls. Starts twist-twirl. For example, banners are placed in the rotation, not carrying a semantic load, but giving a large CTR. Or the traffic with offers begins to be added. The quality of traffic inevitably falls slowly. For a while, there is a surge of poor-quality traffic at an inflated price that is diluted to reasonable quality limits, as much as the advertiser’s marketing department allows. Often - for years. The exception is only very large platforms with constant rotation of the audience, or banner networks that merge non-purchased displays on their own platforms into various affiliate programs.
Option 2 . Buyers work only with large partners. That is, large sites and banner networks. Often, just the whole places on large sites are redeemed, banners are not put into rotation. At best, banners change more or less regularly to fresh ones. Each partner, including banner networks, is treated as one banner space. The fact that the network consists of a set of heterogeneous sites is simply not taken into account, the banner is placed immediately on all more or less suitable sites of the network. It is bought immediately or a large package of shows or places are redeemed for some time. It turns out that from a part of the sites there is a frank trash, somewhere the CTR sags, from somewhere it goes normally. As a result, such a middling at the exit, the average traffic at a price slightly above average. And if you mix it with good, but expensive traffic from a couple of large sites, then in the report the authorities will get a bit cheaper traffic than the allowable registration price for this project, and the quality is slightly above average, you can live. But you don’t need to sit and monitor day and night how there are shows from 100,500 small and medium sites, change sites, disable, change targeting settings and so on. As a result, no one knows that a third of the advertising budget simply merges. Optimization is carried out once a month, if lucky. Those. based on the results of the monthly report, we see that from such a site the last month is fig traffic, or the CTR has dropped sharply, then we don’t buy further shows on it for the next month. All mishn eccomplish.
Option 3 . Most insidious, because he has the most unobvious disadvantages. Social projects often suffer from this, because an advertising manager is one, it’s difficult to optimize impressions, and social networks offer to place ads both with pay-for-impressions and pay-per-click. The bottom line is that once it’s difficult to optimize impressions, let's pay for the clicks! Still simple! And often do not care that the banners are not very CTR-style, we do not pay for the type of shows, the cost of a click for us is the same, that click on the banner that do not click.
For those who are not aware of the subtleties of banner rotation algorithms, we explain.
First , any banner network, well, or at least imputed, that wants to make money on advertising, always buys shows, rather than clicks, from sites. It is elementary more profitable and sites and networks. Sites - because the clicks depend on the quality of the banners, i.e. CTR, and the quality of banners depends on the advertiser. The point is that it is cheaper to sell impressions than clicks. Therefore, all sane sites with high-quality and large audience sell only shows. Well, or rent advertising space for a fixed number of days.
Secondly , the banner network, buying impressions, and selling AND shows and clicks, always seeks to show on the advertising space that will bring her more money. Therefore, every banner has a constantly updated economic coefficient inside the rotator database, expressed in the profit from the display of this particular banner on this particular advertising space. To translate the cost of a click into the cost of the show is very simple: multiply the CTR of the banner by the cost of the click. We get the amount that the site and the network will bring his show. But CTR banner thing is not constant, and it still needs to be measured, for this we need to give a certain number of hits. At that time, as if the banner was sold for impressions to the advertiser, the cost of display is stable and known. Therefore, the advertising network is a little pessimizing the priority of showing banners sold per click, for example, by 10-20 percent, depending on many more indicators. Thus, in order to get the same amount of traffic as a banner sold for impressions with the same CTR, the cost per click must be raised by the advertiser by the same 10-20%. That is, the advertising network assumes the risk that the banner is not clicked. For this, it raises the price. Naturally, not only CTR is taken into account, there are many factors, but it will be enough for a common understanding. With an advertising budget of $ 100,000, ten to twenty thousand is simply given to the banner network for optimizing impressions and taking risks. Think about it.
Thirdly . The client knows what the cost per click for his project is acceptable. And the CTR at banners falls with time, plus it may initially not be so wonderful. Moreover, the lower the CTR, the less traffic comes from this banner. If it is sold for clicks, the number of clicks on it falls, not only because it is less often clicked on, which is the actual drop in CTR, but also the network shows it less often, i.e.The drop is much more noticeable than just the result of a decrease in CTR when buying impressions. To compensate for it, the advertiser’s buyer starts to increase the price per click to the threshold, which is valid, and then leaves this campaign in this form - it brings some amount of traffic, the price is within tolerance, why kill it?
In order to optimize the advertising campaign as efficiently as possible, the most correct thing, in our opinion, is to buy banners for impressions from many sites using banner networks, but to shift the main part of the optimization of impressions to automation. Or do at least monitoring. There are simply no universal machines that would help fully optimize impressions when working immediately with large, stand-alone sites, and with a bunch of sites when purchasing displays in banner networks. Some large banner networks have their own automatic optimization tools. If the networks with which you work do not have them, make the programmers output monitoring of the traffic from different sites of the network to your project statistics as thoroughly as possible and make convenient understandable notifications about changes in the quality of traffic coming from different sites.
We can separately mention how we solved this problem in our banner network, which we prefer to leave unnamed (in order not to be suspected for the mercenary purposes of this material). In fact - very simple. We’ve made a tool for our advertisers with which you can almost automatically optimize your impressions. Called his autofocus. The essence is as follows.
Advertiser, placing a banner for impressions in our network, can indicate a limit on the cost per click. At the same time he pays for the shows. Further, our rotator makes a certain number of test impressions on all the sites of the required category, for example, on the portals about games, and then distributes the shows so that on the sites with the maximum audience response the banner will be shown with higher priority. For each advertising space where the banner is shown, the CTR is calculated during the test impressions. Based on this, it is calculated how much an advertiser’s cost per click from each specific site to this banner is at the current price of impressions. If the cost per click is higher than the advertiser’s settings in the settings, then the impressions on this site stop. As a result, banner displays are concentrated on those sitesfrom which there is a maximum of traffic with minimal cost per impressions. And as soon as the shows on any of the sites cease to give the desired price of a click, they stop there.
As a summary, we list the main points:
In 2007, in order to optimize the advertising of our games, an advertising rotation engine was created. So our advertising network was born. It quickly grew into a large banner network focused exclusively on advertising online games and near-game resources. Today we want to share our observations and experience. We hope that the material will be useful to young teams, but I will be able to learn something interesting from experienced colleagues.
The material is devoted to the following topics:
')
- processing and analysis of traffic;
- traffic extraction methods;
- traffic mining optimization tools.
First, you should go through the analysis and processing of traffic, because before you get traffic, you need to understand what kind of traffic we need. And how much it should cost.
Part 1. Processing and traffic analysis
Obviously, there is no point, for example, to drive tons of expensive, super-high-quality traffic into the game, if it cannot keep users, or if its site cannot convert traffic into registrations.
Before determining the cost and quality of the traffic that we need to purchase, it is necessary to determine and eliminate the main thresholds. Otherwise, you simply do not distinguish good traffic from bad.
Let's take a visual look at the entire path of the traffic, starting with the transitions to the game site.
For example, suppose we drove 100,000 users to the site from a high-quality, normal source to our landing page.
LANDING PAGE is a page where users come from traffic sources. For a browser or client game, this is usually a promotional page, from where the user goes to the registration page. For social projects, this is the page that the user sees before the game installation. Consider the numbers on the example of a browser game.
The normal percentage of conversion of users who come to the landing page in the registration is from 15 to 30% for browser projects, depending on the traffic sources. For games with a downloadable client, the rates are naturally lower. For social programs above, 50-80%. The number of losses at this stage, if you do not consider the quality of the traffic itself, depends on two factors:
- quality landing page;
- quality of the registration form.
To check the quality of the landing page, make a test purchase of traffic from various sources, as good as possible. This check is not suitable for traffic from offers in which users receive buns in other projects for registration in your game - the conversion will be 40-45% for the browser and close to 100% for social projects and it will not show the quality of the landing page.
Examples of the perfect Landing Page, see Blizzard. They have all projects are about the same made, and this is not by accident.
In fact, this is the most important page of your project. At this stage, with incorrect landing page conversion can drop down to 4-5% with the same quality of traffic, which would give 30% on a good landing page for the browser. The only purpose of this page is that the user wants to register in the game. You do not need to load it with anything else, it still doesn’t care about the ratings in the game, it still does not understand the gaming news. All he needs to know by and large is a brief info about the game, screenshots, a movie, and a big “PLAY” button. But he needs to give it beautifully, so as to achieve only the wow effect. In other words, you should not save on a landing page, because it is fraught with losses of tens of thousands of dollars for poor conversion.
If the quality of traffic is low, for example, a "click call" (about it a bit lower), then a good landing page will give 1-1.5% conversion. Strangely enough, sometimes it is not always a tragedy, if the traffic is cheap enough so that the cost of registration ultimately suits us. At the same time, a bad landing page will not allow to work with click-type traffic in general, since conversion will tend to zero.
Plus, an inconvenient or poor-quality registration form can easily lose up to 95% of users, tested by bitter experience.
This problem is not worth it for social networks, where registration takes place in 1 click by the means of the network itself, and where the landing page is often essentially a registration confirmation page with screenshots of the game. We are now talking about the registration on the website of a browser or client game, where the user gets from the landing page.
Consider an example of registration from one game, we will not point the finger which one. To register in “that game” you need to enter your nickname in the game, mail and password. It seems to be all simple and standard. The user fills out the form, presses okay, and then he is told that the password must be at least 6 characters, with at least one of them must be a number and one capital letter. The user came up with a password, entered it twice, clicked OK.
And here he is told that the nickname he entered is busy.
The programmer thought why to pull the server and first check the nickname, if the password can be checked right on the page for validity. And, of course, the password field is cleared. That is, the user must pick a nickname, each iteration by entering a password of at least 6 characters in length. Double enter.
The easiest way to measure the effectiveness of a landing page is to put a Google Analytics counter. Set goals for the registration page and the page after registration. Some projects have a registration form right on the landing page. This is normal, but in this case it is a little harder to keep track of those who have fallen off due to the inconvenience of the form.
The next threshold, at which another 90% of users of an average browser project fall off, is the conversion from registered users into active registrations, and from them into payers. Active in this case are those who did not fall out of the game after logging in until they reached a certain control point, for example, level 2, or did not complete the training course.
The quality of traffic greatly affects the loss at this threshold. For example, when buying traffic from offers in which payment is made for registration, and not for user actions in the game, the conversion of registered users into active registrations may not exceed 1%. Only 5-10% of the remaining users become payers (again, on average in the hospital).
See what happened to our 100 thousand users. I intentionally omit other thresholds that are not affected by traffic quality.
In order to adequately assess the quality of traffic, it is necessary to eliminate these basic thresholds. At the same time, one should be aware that they should be monitored constantly and best of all - automatically.
Ideally, after the elimination of the main thresholds, their daily monitoring should be done by the machine, i.e. script on your server. If the indicators suddenly go beyond the specified values, then a notification should be sent to you. Otherwise, there is a chance to simply not notice how the indicators have changed at some point. And only after a week to see that several tens of thousands of dollars from the advertising budget were merged, because all users simply were not kept in the game due to some minor problem.
Correct traffic
After the main thresholds in the game are eliminated on the test traffic, you can measure the real ARPU.
The correct traffic is traffic, the cost of active registration of which is lower than the ARPU for the period that you consider normal for the payback of advertising in your project. Most often this is 1 month.
Not being able to see the separation of ARPU by users from different sources, you will not be able to adequately estimate the traffic from these sources. Especially it concerns the purchase of traffic with payment for registration. Be sure to ask the programmers for such a table in the admin of the game, where you could compare the conversions of users from different sources at the main control points and ARPU.
Part 2. Traffic extraction
Now let's talk about the methods of extracting traffic, what works, what doesn't, and what quality traffic gives.
"Popander"
Let's start with the "pop-under" of advertising, as with the very "trash" method.
“Popander” - these are banners that crawl over the content of the site, often in the middle of the screen, have annoying content and a “close” cross. Nobody calculates that the user clicks on the banner itself, it is expected that the reaction to closing the banner is expected, and by clicking on the cross, a transition is made to the advertiser's site, which has nothing to do with the banner. A click is needed to bypass the browser protection to open new windows with scripts, this is called “breaking through”.
"Click-under"
Separately, we say a few words about the "click-under." This is a format similar to the “pop-under” in that the advertiser's site opens in a separate window under the current one, and is visible to the user when he closes the browser. But unlike “under-ander”, for “punching” a user click on any place of the site is used, and not a pop-up banner in the pop-up. This is not so annoying to users and generally gives a good result if the platform is thematic and the audience on it is gaming. At the same time, the role of the banner itself is essentially the Landing Page, which opens in a separate window. It should be noted that the quality of one format does not play such a role so much as in the case of a “click-under”
The undeniable advantage of the traffic coming from this format is its cost. Near the cent for the transition.
We put several experiments on this topic in different games, including with the downloadable client, and we managed to achieve conversion from this format in 2-3%, i.e. at the price of a switch to a cent, registration cost 30-50 cents. Normal, valid registration.
The disadvantage of this format is that large thematic areas still do not seek to place it in their own. At the same time, the volumes and coverage should be really large in order to provide the sane number of registrations. It makes no sense to turn the clicker on one site for a long time, because he quickly burn out. If it is even simpler: when most of the regular audience of the site has already seen your site in the window of a clicker, there is no longer any sense to turn it on this site. And this in turn means that such traffic is usually not very much.
We recommend using "click-under" as an addition to the main advertising campaign. From experience, it is also noted that clickkander works well together with a strong banner on the same site.
Text advertising
It also shows itself very well as an additional format - plain text ads placed on thematic sites. We consider this format unfairly undervalued. For example, on one of the top sites of our network, a block of text ads generates about 10,000 clicks per day. At the same time, the quality of the traffic received in this way is very high and is not inferior to the quality of traffic with well-targeted banner advertising.
Branding
Separately, let us say a few words about a promising format that is gaining popularity lately - branding. In the simplest version - this is the wallpaper on the right and left of the site content. The price of a click is higher than that of a regular banner, but at the same time:
- very good quality traffic
- greatly increases the effectiveness of the main advertising campaign.
There is no sense, of course, to brand very small or non-thematic resources.
Offers
The user is offered (in a game or application) to go to another game, do this and that, then come back and get a bun, for example, some object or game currency.
Offers work relatively well in social networks, within the same social network. As for the traffic from the offers to browser or client games, it most often comes across of poor quality.
Sometimes a really mad conversion in registration is made from the offers, if a bonus in the offer is given for registration. If the bun is given for reaching the level - then users swung to this level with a specialist to get a bun, and then they will score for the game. The game remains a tiny percentage of users. The specificity of this traffic is such that users who come to you because of a freebie in another game, so as not to spend 10 rubles on it, are not particularly inclined to pay.
With offers you can collect traffic, but again the question is in the cost of such registration. In any case, before working with an offer, you need to test its traffic before converting to payers and then monitor it tightly. An offer with payment after some actions in the game (“reach the level”) is much more effective than payment for registration. More users are drawn into the game, but still quite a few remain.
But in fact, it is much more sensible to make your affiliate program, with a meager payment for registration, such as 1 rub / registration, but with a percentage of user payments, and offer it to various traffic sources. Then, in general, you will not care how much trash you drive traffic and how specific partners mine it. If the affiliate is promoted, it is a very good and stable addition to the main advertising campaign. But it may never become the main source of traffic; it should be regarded as a good additional source.
Banner advertising
Well, the most effective and most diverse way to extract traffic is banner advertising. There is an opinion based on personal experience that it is on banner advertising that the main advertising campaign should be built. The main thing in the banner is optimization. This topic is quite extensive and it is better to consider it in more detail.
Part 3. Optimization tools
The main enemy of quality and cost of traffic is the quality of the banner itself.
The banner has two main indicators that interest us: the CTR and the quality of the generated transitions.
CTR is the ratio of the number of clicks to the number of banner impressions in percent. The concept of a good CTR varies depending on the site and banner format. For example, when placing a small banner in a sidebar on Facebook or “vkontakte”, the average CTR is one-hundredth of a percent. At the same time, there are sites where the normal CTR is more than 1%.
There are two fundamentally different approaches to the creation and placement of banners. The first approach is focused on raising the CTR by any means, i.e. "Let him click as many people as possible, no matter what." Let the conversion be very bad, but the clicks will be many and cheap. Such an approach is often justified when traffic from the site is already trashy, poorly targeted, and of poor quality. But there are a lot of hits. In this case, the landing page already plays the role of a filter, converting around 1%. But nevertheless, due to the cheapness of the traffic, the cost of registration can be kept within reasonable limits, and these are much more honest and high-quality registrations than from the pay-per-offer offers.
The second approach, as it is not difficult to guess, is focused on ensuring maximum quality of traffic. Those. People who clicked on the banner are already interested in the product. The banner corresponds to the essence of the product, the placement is targeted at the maximum target audience. At the same time, naturally, a lot of effort is applied to also increase the CTR of the banner.
But a banner that gives super quality traffic, but with a low CTR, is the same file as a very comfortable car, but without an engine.
Few observations: animated banners give an average of 30% on average higher CTR than similar in quality static. Video Banners provide 20% more CTR than just animated. This means that you can get one and a half times more traffic for the same money.
If half of your advertising budget is more than the amount you need to spend on good video banners, make them.
Video Banner is a flash banner that is first loaded as a regular banner, but after loading it loads the video insert and plays it inside. Thus, the banner itself loads fairly quickly, it is no more than a regular flash banner, i.e. about 50-70 kb. Instead of a video, while it is being loaded, a freeze-frame, static picture is shown.
Perhaps there are only two reasons not to use video banners: either you do not have the budget for creating a sane video for a banner, or the specific site where you want to place it does not allow you to do this due to some technical reasons.
If you have a beautiful in-game video or, better yet, a video trailer, just cut a few seconds of a connected video out of it that can be played in a circle.
Total: Do not save on banners, as well as on the landing page. Quality advertising materials are our everything.
Despite all of the above, you need to understand that there is always a chance, much more than zero, that your super banner, on which you had hopes, took and did not shoot. Those. showed so low CTR or poor quality of traffic that the registrations received from it simply do not pay off.
The conclusions are simple:
In addition, there is the concept of "burnout" banner. This means that the same banner that hangs on the same platform loses the CTR over time. At the same time, it does not matter how many different banners we will launch on the same platform in rotation at the same time. They will simultaneously lose about the same CTR over time.
On average, by experience, one banner on one site lasts about 1 to 3 weeks. Further, it becomes ineffective, and it is better to replace it with a new one.
Therefore, when planning an advertising campaign, it is necessary to stock up with a certain number of banners for future use, in advance, and taking into account that 2/3 of them may have to be thrown out.
Naturally, there are exceptions, for example sites, where either the majority of the audience is search traffic, or where the audience is incredibly huge, but visits the site rarely and irregularly. For example, torrent trackers and file sharing. Unfortunately, the traffic on such sites is not thematic, i.e. It can be targeted only by region, as a rule, but not by hobby for games. eg.
You can extend the life of banners by using the services of a large banner network where your banner is:
- rotated with other banners;
- rotated on many small and medium thematic areas;
- It is possible to limit the number of contacts with a unique user.
Also, the above allows to significantly increase the CTR, and thus reduce the cost of advertising. For example, in the course of experiments, we derived the following rule: the optimal number of banner contacts with 1 unique is 3 shows per day, no more than 7 shows per week, no more than 14 per month. A similar tuning of the shows gives a stable + 50-60% to the CTR for a good, correct banner, if you apply this restriction on sites with frequent attendance and a stable permanent audience. And at least + 20% on the remaining sites.
Monitoring and analysis
However, when placed on a large number of sites, the question of monitoring and analysis is very acute.
Firstly , even different banner places on one page of one site can give completely different CTR for the same banner. But the placement on them may cost differently, and it’s not a fact that placing expensive on the CTR site itself is more profitable than, for example, on the second pages, but cheaper.
Secondly , different sites give different CTR and quality of traffic from the same banner, due to slightly different audiences for example. Somewhere among the core of the audience of the site there are more fans of browser strategies, somewhere client-side games.
All this translates into the fact that manually optimizing impressions on a large number of sites becomes more expensive than the gain from this optimization, or even impossible.
Here are a few exaggerated, but life options, showing what this can lead to:
Option 1 . The marketing department completely abandons self-advertising, looking for partners who are willing to sell conversions (i.e. ready traffic) or registration. At best, they give their promotional materials. They ask partners for test traffic, if they are satisfied with the quality of traffic, then they continue to buy registrations. All forces are concentrated on finding sane partners with normal traffic.
As a result, partners are driving first normal traffic, using the materials of the advertiser. Then they burn out, CTR falls. Starts twist-twirl. For example, banners are placed in the rotation, not carrying a semantic load, but giving a large CTR. Or the traffic with offers begins to be added. The quality of traffic inevitably falls slowly. For a while, there is a surge of poor-quality traffic at an inflated price that is diluted to reasonable quality limits, as much as the advertiser’s marketing department allows. Often - for years. The exception is only very large platforms with constant rotation of the audience, or banner networks that merge non-purchased displays on their own platforms into various affiliate programs.
Option 2 . Buyers work only with large partners. That is, large sites and banner networks. Often, just the whole places on large sites are redeemed, banners are not put into rotation. At best, banners change more or less regularly to fresh ones. Each partner, including banner networks, is treated as one banner space. The fact that the network consists of a set of heterogeneous sites is simply not taken into account, the banner is placed immediately on all more or less suitable sites of the network. It is bought immediately or a large package of shows or places are redeemed for some time. It turns out that from a part of the sites there is a frank trash, somewhere the CTR sags, from somewhere it goes normally. As a result, such a middling at the exit, the average traffic at a price slightly above average. And if you mix it with good, but expensive traffic from a couple of large sites, then in the report the authorities will get a bit cheaper traffic than the allowable registration price for this project, and the quality is slightly above average, you can live. But you don’t need to sit and monitor day and night how there are shows from 100,500 small and medium sites, change sites, disable, change targeting settings and so on. As a result, no one knows that a third of the advertising budget simply merges. Optimization is carried out once a month, if lucky. Those. based on the results of the monthly report, we see that from such a site the last month is fig traffic, or the CTR has dropped sharply, then we don’t buy further shows on it for the next month. All mishn eccomplish.
Option 3 . Most insidious, because he has the most unobvious disadvantages. Social projects often suffer from this, because an advertising manager is one, it’s difficult to optimize impressions, and social networks offer to place ads both with pay-for-impressions and pay-per-click. The bottom line is that once it’s difficult to optimize impressions, let's pay for the clicks! Still simple! And often do not care that the banners are not very CTR-style, we do not pay for the type of shows, the cost of a click for us is the same, that click on the banner that do not click.
For those who are not aware of the subtleties of banner rotation algorithms, we explain.
First , any banner network, well, or at least imputed, that wants to make money on advertising, always buys shows, rather than clicks, from sites. It is elementary more profitable and sites and networks. Sites - because the clicks depend on the quality of the banners, i.e. CTR, and the quality of banners depends on the advertiser. The point is that it is cheaper to sell impressions than clicks. Therefore, all sane sites with high-quality and large audience sell only shows. Well, or rent advertising space for a fixed number of days.
Secondly , the banner network, buying impressions, and selling AND shows and clicks, always seeks to show on the advertising space that will bring her more money. Therefore, every banner has a constantly updated economic coefficient inside the rotator database, expressed in the profit from the display of this particular banner on this particular advertising space. To translate the cost of a click into the cost of the show is very simple: multiply the CTR of the banner by the cost of the click. We get the amount that the site and the network will bring his show. But CTR banner thing is not constant, and it still needs to be measured, for this we need to give a certain number of hits. At that time, as if the banner was sold for impressions to the advertiser, the cost of display is stable and known. Therefore, the advertising network is a little pessimizing the priority of showing banners sold per click, for example, by 10-20 percent, depending on many more indicators. Thus, in order to get the same amount of traffic as a banner sold for impressions with the same CTR, the cost per click must be raised by the advertiser by the same 10-20%. That is, the advertising network assumes the risk that the banner is not clicked. For this, it raises the price. Naturally, not only CTR is taken into account, there are many factors, but it will be enough for a common understanding. With an advertising budget of $ 100,000, ten to twenty thousand is simply given to the banner network for optimizing impressions and taking risks. Think about it.
Thirdly . The client knows what the cost per click for his project is acceptable. And the CTR at banners falls with time, plus it may initially not be so wonderful. Moreover, the lower the CTR, the less traffic comes from this banner. If it is sold for clicks, the number of clicks on it falls, not only because it is less often clicked on, which is the actual drop in CTR, but also the network shows it less often, i.e.The drop is much more noticeable than just the result of a decrease in CTR when buying impressions. To compensate for it, the advertiser’s buyer starts to increase the price per click to the threshold, which is valid, and then leaves this campaign in this form - it brings some amount of traffic, the price is within tolerance, why kill it?
In order to optimize the advertising campaign as efficiently as possible, the most correct thing, in our opinion, is to buy banners for impressions from many sites using banner networks, but to shift the main part of the optimization of impressions to automation. Or do at least monitoring. There are simply no universal machines that would help fully optimize impressions when working immediately with large, stand-alone sites, and with a bunch of sites when purchasing displays in banner networks. Some large banner networks have their own automatic optimization tools. If the networks with which you work do not have them, make the programmers output monitoring of the traffic from different sites of the network to your project statistics as thoroughly as possible and make convenient understandable notifications about changes in the quality of traffic coming from different sites.
We can separately mention how we solved this problem in our banner network, which we prefer to leave unnamed (in order not to be suspected for the mercenary purposes of this material). In fact - very simple. We’ve made a tool for our advertisers with which you can almost automatically optimize your impressions. Called his autofocus. The essence is as follows.
Advertiser, placing a banner for impressions in our network, can indicate a limit on the cost per click. At the same time he pays for the shows. Further, our rotator makes a certain number of test impressions on all the sites of the required category, for example, on the portals about games, and then distributes the shows so that on the sites with the maximum audience response the banner will be shown with higher priority. For each advertising space where the banner is shown, the CTR is calculated during the test impressions. Based on this, it is calculated how much an advertiser’s cost per click from each specific site to this banner is at the current price of impressions. If the cost per click is higher than the advertiser’s settings in the settings, then the impressions on this site stop. As a result, banner displays are concentrated on those sitesfrom which there is a maximum of traffic with minimal cost per impressions. And as soon as the shows on any of the sites cease to give the desired price of a click, they stop there.
Total
As a summary, we list the main points:
- eliminate thresholds, “lick” landing page;
- The optimal basis for an online game advertising campaign is banner advertising, the rest of the methods are good as ancillary, combine them with a banner;
- make a lot of banners;
- Pay for impressions, if there is a choice between impressions and clicks;
- make test shows, select the best banners, twist them;
- require test placement from networks and sites when you start working with them. All normal networks and most large sites give free test placement, if you introduce yourself and are going to advertise a particular project on an ongoing basis. Test impressions will allow you to determine a reasonable price for impressions on a specific site, based on the price of active registration and ARPU, after you measure the quality of traffic;
- monitor test shows of rltime, 10 thousand impressions are enough to evaluate how a particular banner behaves on a specific site;
- work with the largest possible number of thematic sites;
- monitor the shows on each site separately;
- If possible, monitor the quality of traffic from each site and each banner separately. Different banners can give different quality of traffic from the same site.
Source: https://habr.com/ru/post/147266/
All Articles