EARLY RETENTION, or what players want

What is retention and what is it for?

In mobile free-to-play games, retention, that is, retention of players, is the most important metric that directly affects the project economy. If the main goal in pay-to-play games is to sell the game, in free-to-play games it is important that the player stays as long as possible and occasionally makes purchases. As a consequence, there is a fundamentally different approach to game design and game promotion.

For instance, one study shows that about 37% of games bought on Steam have never been launched. But for the developer and publisher, this is not a significant problem. After all, the game is already bought, the income is received. When it comes to mobile free-to-play games, it is just the opposite — a game that has never been launched is a dead game. And vice versa, the game is more alive, the more often it is played and the longer players stay in it.

Economic viability is determined by classical mathematics — the sum of the costs associated with attracting and retaining a player must be less than the income from the same player.

Retention is perfect for calculating this, besides, the metric works great on raw projects, so the first test that all free-to-play games pass is the retention test.

Why is early retention important?

Retention is calculated as a percentage per day and shows the number of players who stayed in the game. For example, a 7-day retention at 15% tells us that 7 days after the game was launched, only 15% of the players continue to play it.

The actual division of the metric into days allows you to evaluate the effectiveness of the game from different angles. The 1-day retention shows the effectiveness of onboarding and the overall ability of the game to captivate a player, and the 30-day retention allows you to evaluate how sophisticated the game mechanics are, how exciting the player’s progression is, etc. In modern analytics, rates on days 1, 7 and 30 are considered the most indicative.

Naturally, a low early retention automatically suppresses a late one as well — no one comes into the game, which means that there is no one to stay in it. Therefore, it is so important to make the best possible impression at the first stage and keep as many people as possible in the game in the first days.

What affects early retention?

So, hereinafter we will be talking about the most interesting thing — a 1-day retention, or for the sake of simplicity, R1.

The question of what affects R1 is debatable. Some believe that the quality of onboarding is important, while others think that generous meta reward cycles are necessary, and, finally, there are those who claim that the overall quality of the game is above all. In fact, everyone is right, and the specific answer depends on the genre of the game and its features.

In casual genres (match-3, hog), it is extremely important not to change the rules of the game, as the audience is looking for recognizable elements. Therefore, it is extremely important at the first stage to show the player that everything is clear, familiar and comfortable. In the mid-core segment, everything is exactly the opposite. It is important to come up with new moves, and at the same time convey them to the players as intelligibly as possible, so that they understand right away that they have downloaded a special game that can be challenging.

At Donut Lab, we are creating a mid-core shooter with an unusual combination of PvE and PvP. At the same time, our PvP is in most cases indirect, which is generally unique for online games.

Daring to work with mechanics that were unexpected for players, we tracked the retention from the very first tech demo, which was released back in December 2019. And now, having accumulated experience in our project Donut Punks, we would like to share the story of our mistakes and misconceptions, unfulfilled hopes and discoveries that led to breakthroughs.

The first fail

Let us briefly explain the mechanics of the game. In Donut Punks, the player collects a collection of characters, each of which has a unique weapon for PvE and an absolutely unique skill for PvP. The trick is that PvE and PvP work at the same time — in a battle, the player fights off zombie mobs with the help of weapons, and plays tricks on other players with the help of skills.

At the same time, tricks work in completely different ways. They include stealing resources, invulnerability, freezing an opponent, provoking mobs, etc. And in every battle, another important thing is to collect resources — you need to get donuts, take them to storage, defend yourself from thieves or steal yourself.

If, after reading this description, you were able to get a clear idea about the game — congratulations! The rules of the game are multifaceted, which is what makes Donut Punks so exciting. However, we were afraid that this versatility would turn out badly, and if you do not explain all the nuances to the players, they will be lost.

Therefore, in one of the first iterations, the player was greeted by a scrollable animated comic. Before the first fight, it showed how to play tricks and fight zombies. For convenience, the comic was equipped with a progress bar so that the player understands how many frames with explanations are left and when the actual game begins. Also, the comic could be scrolled not only forward, but also backward — in case the player wanted to re-read something. For the sake of high aesthetics, the comic was fully voiced, well-animated and weighed 20 megabytes. We were sure that we took care of the player as much as possible and did everything in good faith!

And what was the result? The analytics showed that the appearance of the comic did not affect the retention at all. And the starting funnel even worsened. We saw that scrolling through the comic was an extra obstacle on the way to the game, and some of the players left already at that stage and never saw what we wanted to offer.

The size of the comic at 20 megabytes was a pain in the neck. That content was used only once in the game. But at the same time, it required constant attention from the testing department.

And we also suffered losses at the production level. Creating animations, voice acting and illustrations took a lot of effort and resources. But it didn’t pay off at all. We really liked the quality of the comic and it was truly amazing. But, unfortunately, the players did not appreciate that. The audience was waiting for a spectacular action game, and not a visual novel.

Corrected mistakes

When we saw that the comic was not working, we unashamedly removed it. The next step was an interactive tutorial — a linear level where the players got acquainted with the basic mechanics themselves, step by step.

Initially the steps were:

1. character control and movement;
2. attack;
3. super attack charging and its application;
4. PvP for collecting resources;
5. delivery of resources to the bank.

All of this took about 2 minutes. Continuing to experiment, we removed the delivery of resources to the bank. The tutorial became even more lively, and now ended with a bright PvP episode.

But there was also the opposite effect. It turned out that the players no longer understood the importance of delivering resources to the bank. As a result, they began to lose more often, and that, in turn, did not have a very good effect on late retention. At that point we started thinking about the steps that the player takes right after the tutorial.

Is there a benefit to variety of mechanics?

We grasped the idea that the best way to teach the players is to let them play. Therefore, we suggested that it would be good to give the player two game modes with different rules right after the tutorial. By design, that was supposed to impress the players with an abundance of opportunities, and as a result, to involve them more in the gameplay and give the opportunity to practice all the nuances of the mechanics that we could hardly push into the tutorial.

We had a basic game mode where 4 players competed in the amount of collected resources. In addition to that, we quickly developed another mode — a cooperative with the capture of points on the map. It brought a significantly different experience, felt fresh, and the development price was in line with our expectations.

We launched an AB test. One half of the players continued to play the old way with one mode, the other one received two modes at once after the tutorial. Looking at the retention, we were surprised — in the early stages, both groups showed comparable results. There was no difference!

Formally, the second mode did not give advantages. However, we did not rush and, taking into account all the pros and cons, distributed the version with two modes to all players. It seemed reasonable. We did not lose anything, and at the same time gave life to the mode, which we had spent development resources on.

At the same time, we continued to watch the analytics. It turned out that the introduction of the second mode completely deceived our expectations. We thought that diversity would work in the early stages (R1), but in fact we saw that the growth started in the later ones (R14, R30).

Why is that? It turned out that one or two game sessions in the new mode are not enough to stick with the player. At first, players are eager to fight with approximately the same tactics, regardless of the rules of the mode. But over time, when the basic things become familiar to the players, they begin to see and, more importantly, appreciate the difference between the modes, realizing that they get different experiences, can apply different tactics and, as a result, get a lot of impressions.

Fragile logic

With care for the players and relying on common sense, we decided to put the game terminology in order. We had two words describing similar gameplay elements — perks and traps. Perks were the characters’ skills, and traps were a game event when someone became a victim of perks (in fact, a classic kill assist).

The semantic connection of those concepts seemed very important. We saw that the players rarely used perks and assumed that their benefits were not obvious to everyone. That benefit was expressed by the concept of a “trap”, which brought generous bonuses to the players. But due to the fact that “perk” and “trap” are two different words, players could not catch the connection.

We assumed that giving it one name instead of two would make things easier for everyone. Therefore, instead of perks and traps, we introduced pranks into the game. And that is how we reached order. The character’s skills are now called pranks, and when someone becomes a victim of a prank, it means that “they were pranked.” You can directly tell the player — do pranks, get bonuses and don’t let someone else prank you.

We changed the texts, released an update and… were horrified. R1 collapsed by a few percent. Instead of using pranks more actively and joining PvP, the players, on the contrary, stopped noticing that unique part of the game, and therefore left. We quickly rolled back the names — and hallelujah! R1 recovered.

Why did it happen? Probably overly creative terms are not very conducive to understanding the mechanics. The words “perk” and “trap”, although they made the terminology heavier, remained intuitive. But in the case of “pranks”, the meaning was lost. The word is well-known, but it is not obvious what is behind it within the framework of game mechanics.

Fine tuning

You can also improve R1 in subtle ways, tweaking the core gameplay and making small improvements here and there.

In one of the versions, we dramatically increased the number of zombie enemies in a battle. Although that decision formally belonged to the PvE part of the game, it consequently improved PvP. Now the players’ characters who would fall into traps were getting into the clutches of zombies faster, which brought a bonus to those who would throw a perk. R1 immediately rose by 2%.

Then we took it one step further. All the characters who would fell into traps received a forced zombie aggro — it means enemies in the first place attacked those who were trapped. As a result, those characters were more likely to be devoured. The intensity of PvP grew, just as R1, in that version by another 3%.

Next, we improved the interface. Now, every time one player lost due to the fault of another, a “teasing screen” (a small plate with custom animation, where it was reported who defeated whom and what they received for that) would appear. The retention of the first day increased by another 2%.

These are just three small improvements, but in total they increased our R1 by 7%. Given how moderate the expenditures for those innovations were, that was an excellent result.

It gets even better. Initially, the players could knock out the main resource (donuts) from almost any zombie. At some point, we decided to create a separate type of zombies that carry loot, rarely appear, hit hard and are overloaded with HP (hitpoints).

That increased the intensity of competition between the players, because now you had to naturally hunt for zombies which have the resource you need to win. Moreover, the answer to the question of where to get donuts became more obvious — a new kind of zombie showed with all its appearance that he had the donuts. Thanks to that, it became easier for the players to navigate the turmoil of the battle. Now they could easily tell apart useful zombies from unworthy ones.

Mistakes and lessons

As you can see, all the described failures and success show how amazing and incomprehensible the nature of retention is. Even the most obvious decisions and the best intentions can turn into failure or, conversely, lead to success.

How to survive in such chaos? Follow the numbers! Retention is a metric that needs to be continuously monitored and analyzed. The indicators are affected not only by product changes, but also by seasonal traffic, inquisition parameters, server response time. It is extremely important to separate one from the other and learn to understand which retention fluctuations are related to the product, and which ones to the technical side.

Free-to-play games are a marathon for both the player and developers. This means that the creators of the game will need flexibility, a willingness to learn from mistakes, doubt and think a lot.

There are many ways to increase early retention, and, most likely, it is necessary to look for a unique approach to each project and invent new solutions. And yet, if we try to give a universal advice that suits everyone, it will be something like: “Always think about the gameplay, follow different tactics, experiment and give your game a chance to reveal itself through mechanics.”