Eric Sefert of Wooga company was a speaker at the latest GDC. He told what’s better: to buy several thousands of installs per day or purchase traffic for a hundred thousand dollars once while trying to get to a TOP. He also spoke about how much TOP-1 costs in the USA, are there games which need marketing and how to evaluate virality of a project and calculate traffic price based on it.
Let’s start from how the development process is built in the company. The structure of the working process reminds a reversed pyramid.
The scheme is quite simple. Everything starts from a prototype. We prototype many ideas in order to choose the best one. Then we get down to production. Further on, a gameable demo version is prepared. After this, “soft-launch” and global launch follow. The last stage, which we call a “Hit”, embodies project success on the world market.
Advance on pyramid stages happens on special meetings where we decide whether the project is ready to a new development stage or it’s time to stop.
Common development strategy looks like this:
We don’t have a goal to finish a project and launch it as fast as possible. There’s only one goal — to understand quickly that a project won’t become a hit and close it.
Jelly Splash is a project based on widely known match3 mechanics. “Soft-launch” took place on June, 14th 2013 in Canada. A world launch was on August, 21 2013. We got to the “TOP Free” (iPhone) at once in several countries: the USA, France, Germany, Netherlands, Spain, Brazil and others.
There’s an own marketing philosophy in the company. These are the principles we rely on while discussing product development.
There is a simple thesis: “I don’t have to grow audience. If I’ve created a great game, the audience will grow on it’s own.” I don’t agree with it and can explain why.
Let me provide an example. There’s a screenshot from an app store above. It displays a promo card for Ninja Kid Run game. The project has 2500 reviews and 4,5 stars rating. People are crazy about the game, they like it a lot. If there are people who leave reviews like “SUPER!!!”, it means that the project is really good. But Ninja Kid is only on the 121 line of TOP Free rating.
And here’s another game – Farm Heroes Saga. It has the same rating as Ninja Kid – 4,5 stars. But it’s on the 10th line of TOP Free rating. What’s the difference between the games? Gamers like both of them. And if you personally don’t like the genres, you have to recognize the games succeeded. But what’s wrong with one of them? Why one project has got to TOP-10 and the other one couldn’t get even to the 100 best apps? The answer is simple – marketing.
There’s a good idea: “I need to attract audience as there are hundreds of great games in stores and all of them are fighting for users.” There are no bad games who succeeded by chance. Competition is so strong in the stores that a bad project doesn’t have a single chance. The good and the best are competing. And the winner is the one which has the best marketing.
There are several ideas for marketing department:
CPI (cost per install) is the basic unit of evaluation of marketing efficiency.
Here’s the formula of CPI calculation:
СРМ (Cost Per Millenium) is a number set by market as usual. You can’t affect it. Though, you may try to search for an advertising partner with the most beneficial conditions for you. But it’s a rather difficult task.
CTR (click-through rate) is a correlation between clicks and displays or “clickability”. In fact, CTR reflects an interest to your game shown through clicks on banners. You can control this metric yourself, affect it, decrease a click price, raising advertising campaign efficiency.
IR – conversion into installs. It’s not enough to get “effective click”, it’s also necessary that a user installs an app. If a gamer clicks on a banner and encounters low quality screenshots or unclear description, it will affect the quantity of installs at once.
Correlation of TAM (total addressable market) and CPI are displayed on the picture above. They are directed to different sides. As quantity of users who are potentially ready to be interested in your game grows, the price for attraction of 1 user decreases. The more people are ready to click on the banner, the more clicks there are, the more installs. TAM and CPI correlation is displayed on the picture above.
Let’s get back to development pyramid. Usually marketing starts from the 4th stage — “soft-launch” because you can buy users there, show them a ready game and watch their reaction. We think that marketing should start along with a prototype.
On an early stage marketing should evaluate market, its demand and develop criteria on which a really good game will be developed. If audience don’t need your game, may be it’s not good enough? That’s why I totally disagree with saying that : “We can make a great game which don’t need marketing as audience will grow on its own.”
It’s a really important question which depends on target audience. There are 2 people on a picture above. They have totally different understanding of what is a good game. We may say that if a guy likes it, then the project audience is smaller than if a girl will like it.
Let’s get back to the question of target market extension. How can a marketing company affect a project on the stage of prototyping? A marketing specialist won’t hint you at the right game mechanics but a good marketing specialist will point at how it would be better to present it to gamers, what do they want to see from realization, what do they or market lacks in frames of this mechanics. In other words, it’s a marketing specialist who can take your central mechanics and offer ways of increasing its potential audience.
The analysis includes:
It has 3 main aspects:
Analytics structure in Wooga looks like this:
We have a separate independent BI-team which builds and supports data infrastructure — a “layer of data” which all the Wooga games pass through. A needed toolset is created within BI which all the employees use later on.
Further on, there’s an independent product team. There’s an own analyst in each product team who gets an access to a data bank, i.e. to all the information on any project. He/she can add to a team of a definite project unique experience of the whole company, constantly studying it, analysing and making necessary conclusions. Such a person is needed, nobody should combine his/her functionality. Otherwise, the potential on analysed data capacity will narrow and the process will become less useful.
The strategy lies between the 2 stages of our pyramid — a “soft-launch” and a release.
“Soft-launch” duration depends on several basic moment:
There are 2 main metrics which should be enhanced to the maximum during a “soft-launch” period: Retention and virality.
Retention is the most important metric which you’ll change in your project. Retention is the number of gamers from an initial group who come back to the game daily from. In other words, it’s how many users who have come today will fall out tomorrow, in a week, in a month…
Retention shows the appeal of your game. If a user likes it, he/she will come back to it. If not, you’ve lost a user and it’s very difficult to win him/her back. Besides, retention allows to forecast life time of an average user in a game.
A typical graph of Retention looks like this:
Retention rate from an initial number of users was 50% on the first day. The same dynamics is for the 3rd, 7th, 15th day — the 7th day has shown 50% of comebacks from the 3rd day, etc. If we take the graph as a basis, it means that we have nothing from the traffic on the 37th day. This way, Retention graph allowed us to learn user LT in 37 days.
You have built a Retention graph and on its basis learnt life time of a user in a game. Next important step displaying project condition is LTV (Life Time Value). This metric shows how much income do you get from a user during a life time.
There are 2 the most popular ways to calculate LTV in our project: Top Down and Bottoms Up.
LTV calculation graph on Top Down method looks like this:
You have cumulative data on user expenses in a game: X (sum of income divided by user quantity) is spent on the first day, Y on the second day, Z on the third day. The value is accumulated on the graph each day. Let’s assume that a monetization graph looks like on the picture above. Now you can take LT number and build an asimptota on the graph and learn your LTV number. On the picture it’s $3,7. You don’t need to wait real end of a user life time cycle in order to build such a graph. You can forecast some of the data by initial dynamics.
Another way of LTV calculation is Bottoms Up. You use a curve on the basis of your Retention and ARPDAU data.
The calculation method is similar to the previous one. You take retention data of the first days. These are Day 1, Day 2 and Day 7 on the graph. The first 3 points will be enough to build the rest of the graph. The curve you get is a power function:
Then you calculate the place under the graph, taking an integral from the function you get. It’s a life time. In order to get LTV, multiply it by ARPDAU.
Pay attention that each part of the formula is segmented, i.e. you can take a segment you need, for example, gamers in the USA, iPhone users, select your ARPDAU indices and calculate LTV for a chosen segment.
Why virality is so important? It’s simple. Users who have come via it are free. And, vice versa, purchasing users is expensive. Surely, developers want as many free users, as possible in order to have more space for marketing maneuvers.
A planned growth of purchased user base is displayed in the graph above. At first the graph of viral users is similar to the one of purchased users. But as the project gets to the peak of viral users, there’s a burst. The graphs start to differ totally. I usually compare virality with per cent on my savings account. You get the more per cent on the per cent you have already saved. The same with virality. Viral users bring more viral users and your gamer base grows faster and faster.
Unfortunately it’s difficult to give virality a precise definition. Especially in mobile games, where there’s no a single channel for getting users.
Usually virality is presented (as in maths) by defining it via “K-factor”. “K-factor” is a number of users a user brings to a game.
Thus, when a K-factor of your game equals 1, it means that your game is viral. One user brings a new user. And the cycle is endless. If a K-factor is less than 1, it means that virality of your project decreases with some time.
There are 2 ways to calculate virality:
Because it helps to decrease expenses on attracting users (СРА). If I buy a user and this user brings another one, so I get 2 new users. CPA cost is divided by 2. It doesn’t make user attraction less significant but it gives a possibility to buy them cheaper or more (it depends on what way is more pleasant for you).
So what are the goals of “soft-launch” you should orient to?
Only after getting these data you may decide whether to release a project ot no.
The first release took place in June 2013 in Canada. We got approximately 2 000 new users per day.
The graph of installs in the period of “soft-launch” is presented below:
There were almost no installs at first, then we got to a conditional plateau of 2 000 users. A little bit later some project changes made the install graph grow stably. We decided to “open up to the world” at that moment.
What else we noticed during the “soft-launch” period:
When we were on a “soft-launch” in Canada, our CPI on iPhone was $2. But unfortunately, the information was useless for the international release as Canada wasn’t our target market. We targeted on the USA but we had no similar games on that market and we couldn’t evaluate how much will user attraction cost.
Apart from CPI price, we know our “K-factor” — 92%. It decreases install price twice. That’s why our CPI is $1.04.
We didn’t know how much user attraction will cost in the USA but if LTV was higher than $1.04, we would have a $2 income from CPI.
We talked with other developers and knew that Canada market was rather expensive. Many teams carry out “soft-launches” here and users know about that. That’s why we assumed that the USA market would fit in our indices of income from the point of view of CPI.
As for common knowledge:
And we had to make a decision about launch strategy — “stairs” or “springboard”.
The method assumes that any marketing activity brings income. You purchase users by a formula CPI less than LTV. You slowly grow user database.
Advantages: you almost don’t risk your money. Each day you attract users. In case user attraction isn’t profitable anymore, you stop and search for the reason.
Many people forget that the money spent in the beginning of the month will pay off only in the end of the month, the process needs some time. But all the expenses won’t pay off in these 30 days. You won’t pay off the marketing budget in 6-month time since the launch. Thus, the more money you spend on the project launch, the more money you pay in advance.
Disadvantages: you leave money “on the table” and you can’t predict market changes. Using a “springboard” model, you get money right here and now, but a “stairs” model may bring you income in several months only. And nobody can guarantee that there won’t be a strong competitor by that time.
It’s a totally different strategy. You launch campaigns where LTV is less than CPI. The campaigns won’t bring you much income. But the goal is to get to a definite place in a TOP. You spend more money in order to get a higher position. Having taken a definite place in a TOP, you get organic installs. And they pay off your expenses.
Advantages: you cover more users. If virality of your project is good, some “pushes” for getting to a TOP is a good approach. A thesis “time is money” works for you. It’s better to get a dollar now than in half a year.
Disadvantages: huge investitions in advance. You have to spend all the money on release. It’s difficult to calculate “Price/Risk” correlation as you don’t know exactly what’s ahead. If you carry out a campaign and see that it’s not profitable yet, it’s difficult to forecast when and how profitable it can be in future. In “stairs” you see the results every day but in a “springboard” model you throw a pile of money into the game release.
The graph below displays 2 types of marketing launches:
On the upper graph — they bought a million of users on the first day (it may be stupid but it’s a vivid example) and they haven’t got a single user.
On the lower graph — they bought a million of users as well but the purchase was divided in time, i.e. they were purchasing the same amount of users systematically.
Both the projects have the same “K-factor” — 20%. On the one hand, it’s risky and thoughtless to buy a big “push”, but if we compare a dry result, we can see that virality effect has worked for the first project more. And it has a larger userbase. That’s why virality is so important. It affects attracting your audience.
Ask yourself one question before a global release — do you completely believe in your project, are you ready to choose a marketing way with a “springboard” or you think you should take a more quiet way? In order to answer it, it’s necessary to calculate expected expenses.
For the beginning, we need to decide which chart position is a priority for us. We decided that we want to get to TOP-1 for iPhone in the USA. We also had to reach $2 CPI index.
We also decided that we have to get 20 000 installs per day. This definite decision has no scientific basis. We just talked to other developers and based on our own experience.
Further on we found out, that in order to be in the TOP-1 in the USA we needed to attract 140 000 installs. It would cost $281 000 for $2 CPI.
Data on the basis of Distimo analysis
We remembered that our “K-factor” equaled almost 1. One user brought another one. That’s why we discounted the old data and get new ones. The we had to get 70 ooo installs and spend 140 000 dollars.
But there was something strange in such an evaluation. We got the data that we needed 30 000 users in order to get to a TOP-10 apps, 56 ooo users to get to TOP-3 and 70 000 to get to the 1st position. No model based on freemium looks like this. There’s no linear dependence, it just doesn’t exist. We thought that linear progression should look like this as that’s the way it happens in freemium models:
We defined a standard set of target countries (France, the USA,Germany, Brazil, Netherlands) and decided that we should become №1 on all of the markets.
There’s a table below with expected expenses for getting to TOP-1 in every of the chosen countries (with “K-factor”):
We succeeded in becoming №1 on all the markets with 92% “K-factor”; $2 CPI on the USA market.
We got $150 ooo on the first week after the release. Common budget for the release was $250 ooo.
What we’ve got:
Results of the campaign:
When we stopped investing money in product promotion, there wasn’t income decrease. Jelly Splash had a really high Retention.
Here you can see a DNU and “Income” graph. We stopped investing lots of money, DNU stopped to grow but income was still increasing.