Eric Sefert of Wooga company was a speaker at the latest GDC. He spoke about different traffic purchase strategies, how much does it cost to get tot the top of US App Store, whether there are games who do not need marketing, and how to evaluate the virality of the project and calculate the cost of traffic on the basis of it.
Wooga development approach
I'll start with how the development in our company works. The structure of the workflow resembles an inverted pyramid. The scheme is simple. It all starts with a prototype. We prototype a lot of ideas to choose the best. Then we proceed to production. The next phase is the preparation of a playable demo. This is followed by a “Soft-Launch" and the global launch. The last stage, which we call “Hit", represents the success of the project on the global market.
Promotion stages of the pyramid occur at special meetings where we decide whether the project is ready for a new stage of development or is it time to stop.
The overall strategy of development is as follows:
- Going to small development teams;
- These teams are doing a bunch of prototypes;
- Obtained prototypes are studied, and those who do not have the bright potential - simply get closed;
- By this method of natural selection we get 2 hits per year.
We do not have the goal to complete the project as soon as possible and launch it. We only have one goal - to understand that the project will not be hit and close it as soon as possible.
What is Jelly Splash
So, Jelly Splash is a project based on the familiar and widespread “match 3" mechanics . The "Soft-Launch” was set on June 14, 2013 in Canada. Global launch took place already on August 21, 2013. We have managed to immediately hit the "Top Free» (iPhone) in a number of countries: USA, France, Germany, Netherlands, Spain, Estonia, Brazil and many others.
There is a special marketing philosophy on our company. I also share this vision. These are the principles that we rely on when discussing the development of products.
There is a simple thesis: "I do not need to borrow a set of audience. If I did a great game - the audience will grow by itself.“ I do not agree with this approach and can not explain why.
I will give you a small example. Above you can see a screenshot from the App Store which depicts a promo card for game called Ninja Kid Run. It has 2500 reviews and a rating of 4.5 stars. People are raving about the game, they love it. If there are people who have left reviews like “AWESOME !!!” then your project is really good. But there is only one detail - Ninja Kid holds only 121st place in the "Top Free».
And here's another game - Farm Heroes Saga. It has the same rating as that of the Ninja Kid - 4,5 stars. But it is in 10th place in the "Top Free». So what's the difference between these two games? Players like them both. And even if you personally do not like these genres, we should agree that both projects are successful - they are being loved by their audience. But what went wrong? Why one project reached the "Top 10", and the other one did not even make the top hundred? The answer is simple - they are separated by marketing.
There is a good saying: "I need to attract an audience, because the App Store has hundreds of great games, and they are all fighting for the attention of users." There are no bad games that accidentally get fired up. Competition in the stores is so strong that a bad project simply does not have a place for competition there. The ones who are competing are the good and the better ones. And the winner of the competition is always determined by the marketing efforts.
Based on the above, I personally see a few goals for the marketing department of the company:
- Scaling. Marketing helps cover a larger and larger audience by acquiring it.
- Optimization. Spending to attract new users should be commercially viable i.e. not burn huge amounts of money for the sake of an abstract goal of "millions of users." The ultimate goal of any project is profit. Marketing helps to maximize it by investing part of the proceeds to the effective channels of user acquisition.
Metrics and marketing effectiveness
The basic unit of measuring the effectiveness of mobile marketing is CPI (cost per install) - the price of app installation.
CPM (Cost Per Millenium) - the cost of thousands of banner impressions - a figure which is usually dictated by the market. It is unlikely that you will be able to affect it, except that you find an advertising partner with more suitable conditions for you, but that's a difficult task.
CTR (click-through rate) - the ratio of clicks to impressions or "clickable". In fact, CTR reflects the interest in your game, manifested through the "clicks" on the banners. This is a metric that you more or less can control and influence. You have the ability to reduce the cost per click, thereby increasing the effectiveness of advertising campaigns.
IR - conversion to installation. It is not enough to get an "effective click”. The user should install the application. If the player clicks on the banner and encounters a bad description and ugly screenshots - you are starting to waste your money.
The picture above is showing the ratio of TAM (total addressable market, the target market) and CPI (cost-per-install). They are pointing in different directions. As soon as the number of users potentially ready interested in your game increases the cost of attracting a user is reduced. The more people are willing to click on the banner, the more clicks, the more installs.
How to improve TAM
Now let's get back to the pyramid of development. Usually, marketing begins with the fourth stage - “Soft-Launch", because it is possible to buy the users, show them the finished game and watch the reaction. We believe that marketing should begin with the prototype.
Marketing at an early stage of development (prototype) should assess the market demand for it and work out the criteria that will create a really good game, demanded an audience. If your game is not desired by the users - maybe it is not so good? That's why I fundamentally disagree with the statement, "We made a great game that does not need marketing, the audience will grow itself."
Who decides if the game is good?
It really is a very important question, which depends on the target audience. The picture above shows two humans. Good game in the understanding of one and the second one are two completely different games, with their potential audiences and opportunities for growth. It can be argued that if the game is like the guy on the left - a potential user group of the project is narrower than if it is liked by a sports girl on the right.
So, back to the question of the expansion of the target market. How can a marketing campaign affect the project that is still at the stage of prototyping? It is unlikely that a marketer will tell you the correct game mechanics, but if it is already selected, a good marketer will tell how to better serve the players and what they might want to see in the game, what do they or the whole market lack in terms of these mechanics. In other words, the marketer is able to take your central mechanics and suggest ways to increase its potential audience.
This analysis includes:
- Demographic evaluation of the project, as users have established a paying age group for which it is worth to "sharpen" the project;
- Identification of basic game mechanics that will work well on the selected demographic group.
The development that is based on data lies on three pillars:
- First - analytics. This includes all the means of collecting and analysing data using both custom and integrated third-party solutions.
- Second - experiments. Based on the last paragraph you see what does not work, where players encounter certain problems and do a number of experiments to improve the game.
- Third - Work on iterations. Gradually improve any component in order to quickly see the response from the audience. Since the changes made are usually small you might have to meticulously look out for the changes in user behaviour in the statistics, but that way you protect the players from possible bad experience with unsuccessful experiments.
The Wooga analytics structure
We have a separate independent BI-team that builds and maintains the infrastructure of the data - a "data layer" that runs through every Wooga game. Within the framework of BI the necessary tools are created, which later get used by all employees.
What follows is an independent product team. Each product team has its own analyst who get access to the database, that is, to absolutely any information about any project within the company. It can bring some unique experience to the team of a particular project and a unique experience throughout the company. Constantly studying the info, analyzing and making the necessary conclusions is more than important for a company like ours.
The "Soft-Launch" strategy
This strategy part lies between the two phases of our pyramid - the "soft-Launch" and release.
So, what is a "soft-Launch"
- Feedback from users. The most interesting thing at this point is to get feedback from the live users. And it's not just about the verbal description (in the comments or on the FB-page), but also about how to obtain data through your analytics. Each user, who is being tracked by the analytics system improves the understanding of the problems with the game.
- Feedback from the market. As far as users are willing to play your game, how good your click throughs of banners, what is the conversion of transitions to the installation. In addition, you can make the first rough estimate - how much will monthly marketing costs be for your project.
- Iterations and cyclical development. The earlier the release sets out the sooner you will begin to see how users behave with your game. This should help you stop pushing abstract hypothesis. This approach is justified from an economic point of view: as long as you're doing a big project without any real data, and is based only on assumptions, you are losing money. It is reasonably better to analyze a small amount of data from the “Soft-Launch" and spend money on a "smart" (calculated) to traffic driving model. In fact, you are improving project from iteration to iteration and cyclically repeating marketing efforts after each step, until you achieve the desired performance.
- Scalability. You have to understand how scalable your game is. And when I talk about scaling, I mean CPI formula to LTV - the price for installs to income over the lifetime of the player in the draft. You must be aware of the conditions under which your marketing and project as a whole profitable, in order not to get trapped with the user base who install the game, run it a few times and leave without paying a dime for it.
The duration of “Soft-Launch”
Usually, the duration of "Soft-Launch" depends on a few variables:
- First - how your current metrics are far from standard. Most studios do not have an exact plan of action, they simply examine the data and respond to the feedback of users. Here, it is still important to understand the difference between a feasible goal and an insurmountable barrier. For example, you have a clear understanding of what is required to generate the retention for the project you are developing in the genre that is preferred by your chosen audience. After a series of experiments with Soft Launches you realize that there is still a long way ahead towards success and there may not be any actions that will lead you quickly to your destination. You're left with a choice: to extend the stage of Soft-Launch or close the project until more useless effort is spent, draining your time and money.
- Second - how costly can a Soft Launch period be for your company. Many companies simply open their products to the world because of the need in generating at least some amount of cash. They can’t afford long and steady improvements to the game based on the results generated by their small audience.
- Third - your understanding of the competition on the market. Here’s and example of Clash of Clans clones: if you did a clone of this game in the first half of 2013, you should have been prepared for a limited amount of time for your Soft Launch, because by the end of the year the market was full of other Clash of Clans clones.
- Fourth - how polished your game is. If you start the game too early, the period of "soft-Launch" should take a little more time. This does not mean that it is necessary to improve the game as much as possible before the "soft-Launch." Conversely, if you have an almost finished, perfectly polished game, which in this case has not yet been tested in combat conditions, then any refinement and iterative processes can get dramatically complicated. You've spent too much time, effort and money on this version to be able to flexibly approach to its changes.
What is the purpose of the first release of the game and what you need to focus on?
There are two main metrics that need to be improved for the maximum period of "soft-Launch»: Retention (recurrence) and virality. Retention - is the most important metric that you should measure within your project. By “Retention" I understand a number of players that get back in the game every day vs. the number of all acquired users. In other words, how many people that are coming in today fall off tomorrow, the next day, a week or two, a month ...
Retention shows the attractiveness of your game - if the user likes it, then he will definitely be back. If not, then you have lost the user, and to it usually is very difficult to get him back. Besides, retention allows to predict life expectancy (LT, Life Time) average per user in the game.
The typical retention chart looks like this:
The indicator of returns on the first day was 50% of the initial volume of the audience. On the third, seventh and fifteenth days, we can observe the same dynamics - the seventh day showed a 50% return of players from the 3rd day, day 15 showed 50% of the returns from the players the 7th day etc. If you take this chart as a basis, we find that by the 37th day, there is 0 users of those who were acquired on day one. Thus, the Retention chart enabled us to identify that the user LT is 37 days.
The LTV formula:
You have already found out the retention rate. Based on retention you can find out the "lifetime” of users within your app. The next important step, showing the status of the project is LTV (Life Time Value). As you might guess, this metric shows how much cash the player bring a player during his life within the app.
Total there are two most popular ways to calculate LTV in your project: Top Dawn and Bottoms Up
The LTV calculation by the method of Top Down looks like this:
You already have a cumulative spending data for your users within the app: the total spend for the first day is X (amount of income divided by the number of users), the second day is Y, the third day is Z. Every day value on the graph is stored. Let us Suppose that the graph of monetization looks as shown above. Now you can take the obtained LT value and build an asymptote on the graph and find the value for your LTV. The picture above is about $ 3.7. You do not have to wait for the real end of life of the user to construct a graph. Part of the data could be predicted by the initial dynamics.
The second method of calculating the LTV is "Bottoms Up". In it - you should be using a curve based on your Retention and ARPDAU values.
This method is quite similar to the previous one - you take the Retention data for the first couple of days. On the graph we have Day 1, Day 2 and Day 7. This data enough to build the rest of the schedule. The resulting curve is power function:
Then you calculate the area under the graph by taking the integral of the resulting function - a lifetime. Multiplying it by ARPDAU, now you have the LTV.
Note that each part of the formula is segmented - that is, you can get your desired segment, for example, players in the United States, users of iPhone, select your performance ARPDAU and calculate the LTV for the selected segment.
Why is it so important for the app to be viral? It's simple - people who have come to your app or game thanks to virility do so free of charge. And people are buying, on the contrary, - very expensive. Of course, developers want as many free users to have more room to maneuver with marketing.
The chart above shows the steady growth of the purchased users database. In the beginning, the viral users’ growth is about repeating the graph of the purchased ones. But once the project reaches a large number of viral users - an explosion occurs and the graphics are beginning to look strikingly different. I usually compare virality with interest on my savings account. You get more interest thanks to the sum of interest that has accumulated. The mechanism of virality works the same - the more viral users you have - the more other viral users they bring in.
Determination of virality
Unfortunately, it is very difficult to give a clear definition to “Virality” term, especially in mobile games, where there is no single channel for for user acquisition.
Usually virality, as is customary in mathematics, is defined by the "K-factor". "K-factor”is the number of new users, which are brought thanks to another user of a game or a mobile app.
Thus, when the K-factor is equal to "1" - your project is indeed viral. One user causes a new user and the cycle is endless. If the "K-factor" is less than “1", this means that the virality of your project decreases with time.
Again, there are two ways to calculate virality:
- First - Top Dawn. "K-factor" is calculated as follows: number of invitations, which are sent to one user, multiplied by the conversion time of the invitations to the real new users. But for mobile gaming, this formula is not very suitable, because the players are sent an invitation to the great mass of potential users, and it is very difficult to calculate exactly where they come from. That is, the conversion by invitations is hard to measure. But this formula is perfect for social games on Facebook. It is very easy to calculate the "K-factor" because each invitation comes with a referral link, and it is easy to get the conversion data.
- The second method - Bottoms Up. You compare the two periods of time and identify users that have been brought to the game without tapping on banner ads. Imagine that a "Day X" - a day of release. You have a specific DNU number. That means you know exactly how many people got into the game thanks to your marketing efforts. And knowing this, you can calculate how many users came "organically." Now look at the "Day X + 1." If there haven't been any changes in the project or you haven’t been featured, the number of organic users does not change, there is no reason for this. Let's say you paid exactly the same for the users as in the "Day X". If you have a delta between these two days - "Day X" and "Day X + 1" - then this is virality itself.
If nothing has changed in the product, it is really easy to calculate virality. Simply divide the number of viral people "Day X + 1" by organic and purchased users on "Day X".
If, however, the project has undergone significant changes, this formula does not fit you. We use this formula in the period of "soft-Launch" - precisely because they do not change the key features of the project.
And yet, why virality is so important
Because it reduces the cost of engaging users (CPA). If I buy a single user, and the user gives me another one, I have two new members. The cost of CPA, respectively, divided by two. This is does not concur to acquiring users, but gives you a more return on the invested money.
So, what are the metric goals of “Soft-Launch", which you should really focus on?
It was only after the receipt of the data, you can decide whether to release the project or not.
Jaelly Splash "Soft Launch" example
The first launch took place in June 2013 in Canada. We were receiving approximately 2,000 new users per day.
The graph below shows installs during the “Soft-Launch":
At first, there were almost no installs. Then we went for a conditional plateau of 2000 users. A little later, some improvements to the project worked in way that the graph of installs began to grow steadily. At this point, we decided to "open up to the world"
What else have we seen during the “Soft-Launch":
- Strong virality. We got the "k-factor” indicator of about 92%. And I repeat, we have grown by about 2 000 new users a day without introducing new features and did not change the game globally. Here we had a chance to apply the “Top-Down” formula for the calculation of virality
- High retention. Prior to the international launch, our retention rate was 50%.
When we were still in the "soft-Launch" in Canada, our CPI on the iPhone was $2. But, unfortunately, this information was not useful for international release, because Canada was not our target market. We are focused on the United States, but we did not have these games on the market, and so we could not even estimate approximately how much it would cost to attract the user.
Besides price, CPI, and we know our "K-factor" - 92%, which is almost half the price drops installs. Therefore, an indicator of our CPI actually was - $ 1.04 dollars.
We did not know how much it will cost to attract the user in America, but if LTV is above $ 1.04, we will come out to profit from the CPI in $ 2.
We talked to other developers, and got to know that the Canadian market is very expensive. Many teams will "soft-Launch" here, and people know about it. Therefore, we have assumed that the market will fall in the United States, our profitability in terms of CPI.
Strategy for the global launch
So, about our common knowledge at that time:
- We understood that marketing costs will pay for the exponent 1.04 LTV, while maintaining the $2 CPI
- We had a large TAM - iOS, casual game with high CTR. The game is easy and not in a niche segment of games.
And we needed to make a decision about our exit strategy - “Stairway" or “Springboard".
This method is used under the condition of a stable marketing activity that drives a profit. You buy the users on a CPI that is less than LTV. You slowly build up the user base.
Pros: You have almost no money risk. Every day you systematically attract users. Once you notice that the attraction becomes unprofitable - you immediately stop and try to understand the reasons.
Many people forget that money spent at the beginning of the month will pay off only by the end of the month, this process takes time. But even after these 30 days you will not recoup all of your expenses, only the money spent on the cumulative LTV for the last 30 days.
Your marketing budget will not pay back earlier than 6 months after launch. Thus, the more money you spend on the launch of the project, the more money you make in advance.
Cons: you are leaving the money "on the table" and you can not predict market changes. If the “Sringboard” model gets you the money here and now, the “Stairway" can be profitable only after a couple of months, and none shall make you insured against having a strong competitor by that time.
Exactly the opposite strategy. You run a campaign where the LTV is less than CPI. In fact, these campaigns do not make a profit. But here you are aiming at a certain place in the TOP. You spend more money to take a higher position in the App Store. Once you have established the place in the "charts” - you are starting to get organic installs. And they will be giving you the money for your expenses.
Pros: You cover a larger number of users. If your project all is doing well with virality, the one-time "push" to get into the top can be a good approach. Here your working thesis "time is money" - you'd better get your dollar now than in six months.
Cons: Huge investment up front - you have to spend all your money on launch. And the ratio of "price / risk" is difficult to calculate, you do not know exactly what is coming. If you run a campaign and see that it is not profitable, it is difficult for you to predict when and how profitable it can be. With “Stairway” strategy you see the results every day. With "Springboard" you just throw a bunch of money to start the game.
The chart below shows the two types of marketing Launches:
The upper graph - one million users bought on the first day (we can say that it's stupid, but illustrative example), and then was not acquired a single user.
The lower graph - also bought a million users, but this purchase is distributed over a small period of time, ie the same number of users systematically acquired and not at once.
For both projects, "K-factor" is the same - 20%. On the one hand, to buy a large “push" is dangerous and reckless, but if you compare the dry result you can see that the effect of virality worked better on the first project it did get a larger amount of userbse. That's why virality is so important for mobile app developers - it affects the set of their audience.
Approaching the global launch
Before launching for the global market - you need to ask yourself a few questions - how much do we believe in our project?, are we ready to choose a marketing way with "springboard" or should behave in a more measured manner. To answer this question, it was necessary to calculate the expected costs.
To begin with, we had to decide which chart position was priority for us. We decided that we needed to get in the top 1 on the iPhone in the United States. We also had to reach the CPI index of $2.
Also, we decided that we needed to be getting 20,000 new units a day. Specifically, this decision dis not imply a scientific method, we were just talking with other developers and thought based on own experience.
Further, we have found out that we need to attract 140 thousand unit downloads to get to the Top 1 in the United States. Considering the $2 CPI - the cost of such campaign had to be $ 281 000.
Data analysis based Distimo service
We remembered that our "K-factor" equaled almost one - one user brought another user. Therefore, we've discounted the resulting indicators and received new data. Now you need to get 70 thousand units and spend 140 thousand dollars.
But something was still confusing us in our assessments. We’ve had the data that we needed 30,000 new users to get into the top 10 applications, 56 thousand to get in the top-3 and 70 thousand to get the first place. None of the models based on freemium model did not behave that way. There is no linear relationship, it just does not exist. We thought that this linear progression should look like this, because that's what usually happens in all freemium models:
- This is an increasing exponential function;
- The rate of prices growth increases;
- Price of the transition from position №10 in the top chart position to №3 and then №1 differ by an order.
We have defined a standard set of target countries (France, USA, Germany, Brazil, the Netherlands) and decided that we should become №1 in all of those markets.
Below is a table with the estimated outlay for getting into TOP1 for each of the selected countries:
- And we managed to become №1 all selected markets with the K-factor index of 92%; $2 CPI for USA market.
- We had $150,000 for the first week after launch. The total budget for the launch was $250 thousand.
The logistics of international release
1. Stepped launch. We had our own scheme of work, we did not carry out a simultaneous release on all five markets. We have only five people in the marketing team, and the release to the global market is always a lot of stress. We wanted to keep the team from this release and separate it into stages. Therefore, each day we launched in one country, with the exception of France - it was launched in conjunction with the United States.
2. World release was set out on 23 networks. With this launch, we have chosen the “Springboard” strategy. We knew how much we users needed to break even, and bought more. Purchases were unprofitable business, but it was assumed that all will pay off by taking the Top 1 on the U.S. AppStore.
- We have exceeded the budget for each grid by 25%;
- We were actively using social media (Wooga has more than 22 million subscribers on Facebook);
- We have resorted to cross-promotional mechanics to other Wooga games.
The results of campaigns:
- We were in the Top 1 in all the targeted countries;
- 5 million installations over 5 days;
- Became №1 in other key countries, such as in the Nordic region.
Once we stopped investing in promotion of the product, the reduction in revenues did not follow. We’ve had a very high retention rate in Jelly SPlash.
Here you can see a DNU graph and "Profit" - we stopped investing a lot of money, and DNU stopped growing, but the income has continued to rise.
What have we learned
- "Explosive campaign" can make a profit. Yes, we went to the model of “Springboard", but before that had a deep market research and suggested that such a model is really suitable. We still made a profit even considering the large up-front expenses for marketing;
- We were in the top 10 applications in the United States, and it gave us organic growth of 35k DNU. Staying in the top 3 gave us 50,000 additional units to DNU;
- Most of the networks have promised excessive amounts of traffic. Error was more than 20%;
- There is a "damping virality": The users who install the app first are the most active ones and remain the most active throughout the app lifecycle.
- The "K-factor” has decreased to 84% in the first month and CPI rose by 28%;
- But we kept the CPI at the same level by working with a large number of networks. In the end, we started working with 30 networks and CPI was kept at the same level.
4 main conclusions:
1. The long period of the first run gave us the opportunity to develop the game and iteratively make changes that have been important to large audience. We had a chance to measure the "K-factor", and in the future we knew the game and the users so well that were not afraid to run an international launch with a budget of a quarter of a million dollars;
2. Assessment of LTV did not just to calm us down but made it possible to prepare for a global release;
3. Analysis of the market - we were able to accurately assess our goals and objectives and to identify where we can achieve the goal of becoming №1, based on planned expenditures;
4. Virality is incredibly important. Without the 92% "K-factor” figure we could have never become №1. At least, not with a budget of a quarter million.
Source: siliconrus.com/2014/07/woogaOur G+