The developers of Taxi Run from Azur Games tell us about their methods to increase profit and retention rates.

Mobile game development is a process that rarely leads to instant success. When a game is first released, many players churn and delete games because they don’t offer the right experience for gamers.

Taxi Run was released two years ago and is still profitable. However, this article is not so much about revenue and performance as it is about a game that failed two tests but came to profit after redesigning the levels. A new approach to using assets in the hypercasual game also played a big part in this process.

The beginning of development

The idea behind the game is for the player to drive a car around the city and use simple controls. The company did not repeat the popular street racing setting but used the concept of a cab where passengers are picked up and dropped off.

The team worked on the project together with Ural Games. It took 2.5 weeks to create the prototype, which was immediately sent for testing. It turned out not very good:

  • R1 is about 20-25%.
  • CPI in the U.S. is 64 cents.

The peculiarity of the hyper-casual market is that after such figures, the project is closed.

The original gameplay turned out to be boring. There was no environment, just the desert and a lot of long, straight, boring road sections. At the same time, we spent a decent amount of time on writing the level generator, so there were many levels, about three hundred. I had already decided to close the project, but decided to make a couple more iterations.

Kirill Turov, Team Leader at Ural Games

Azur Games publisher did not expect to get a hit game at once and helped the developers to improve the project because the metrics had objective reasons for being low. The first thing they paid attention to was the visual component.

Proprietary assets

The thing we didn’t like the most was the cars. These are the standard models from the Asset Store that are used in hyper-casual games. Some of the assets fit, but you needed a lot of cars for the meta, and there are no such large packs in the right style.

In 2019, Azur Games recruited 2D and 3D teams to make a pack of unique, beautiful cars and put them in recognizable car shapes. It took 2-3 days to create each car.

For the first pack, they made 20 cars; since then, Taxi Run has used unique cars for their production. Today the number of models exceeds 40 pieces and is regularly updated.

The analytics show that premium cars with a special badge are more popular with players. Typically, these are cars with visual style references and unique effects, such as a rainbow trail. Players are much more likely to watch ads with rewards to unlock or improve cars.

Kirill Turov, Team Leader at Ural Games

After the new cars were introduced, R1 increased by a couple of per cent, but the gameplay still had a problem. It was heavy, monotonous and not hyper-casual, which caused players to leave en masse. To fix this, the team took on level design in addition to working on the cars.

Level design and environment

First, we gathered assets for the city. Collaborating with the Ural Games development team, released another game — City Gangs, which uses paid construction assets. They perfectly fit the style of cars, as if from one package. All materials were combined into one project and collected in the first 20 levels instead of the previous three hundred.

The old levels had two problems:

  • Monotony. Too long stretches of road where nothing happened;
  • Complexity. Sections of levels with very dense traffic;
  • The winning percentage sometimes fell below 50%, which is too difficult for simple gameplay.

The team solved the first problem by reducing the time to pass the level several times and dramatically changing the environment. For 10-15 seconds, the player had time to make turns, jump off a diving board, go out of town, cross the railroad tracks, and so on. It was dynamic: drive — get a reward — drive again.

The winning percentage was harder to fix. The team raised the score to 90%, and the players left because there was no challenge. After experimenting and balancing, the score stopped at 70-80%.

The team broke the levels into blocks and made a level builder to support the dynamics and analyze the repetition of elements. The end result was one hundred unique blocks, and each level consisted of an average of three. Thus, a section of road from level two could be repeated on level 30, but in a new context. The order of the blocks was never repeated. The color scheme, time of day, and weather conditions changed. This created a new sense of environment.

Kirill Turov, head of the Ural Games team

They also detailed the city to improve realism: they added lampposts, trash bins, and hung clotheslines in the street. Created interactive moments and mini-stories: players knocked down boxes, drove through the lock, and instead of meaningless dead ends, there was a road accident or a destroyed bridge. Thus, the mechanics didn’t change, but there was a sense of exploration in the game, as the scenery and surroundings constantly changed.

Initially, each level began with a passenger boarding a train and ended with a passenger being dropped off. But after testing, the team noticed that it was tedious to see this on every level. That detail was updated. After the changes, the passenger only got on the car once per game, on the very first level, and it increased the metrics. The second test showed that it was better not to sit the passenger at all, even on the first run, and this, too, had a positive effect on the game.

Kirill Turov, Team Leader at Ural Games

We also removed the graphics settings: antialiasing, anisotropic filtering, etc. After this, the performance increased. This is due to the frame rate, which has become more stable.

New test and current state

After iterations, the game is much deeper than typical hyper-casual projects. You could say that the level design saved the game. Creating complex levels for hyper-casual games is expensive, but it helped to stand out firmly from other genres’ games.

A separate role was played by the unique creatives that went to the marketing and ASO teams. They replaced the images for the advertising campaigns and store pages, which impacted CPI.

The resulting numbers were as follows:

  • R1 rose to 40%;
  • CPI in the U.S. dropped to 21 cents;
  • Playtime was 9-10 minutes.

Over time, the metrics have sagged as the game is two years old, but the project is still going strong. Regular updates help keep the numbers stable.

The game needs more content. In Taxi Run, players are particularly sensitive to level design. The team did a lot of testing and found that players notice the same twist that repeats itself in two adjacent levels, and this is immediately repulsive. The second point is that players like the challenge, but don’t like the difficulty. When creating a level, you have to constantly maintain a balance between elements that create tension and elements that impede getting through the level.

Milana Gelmanova, producer of Azur Games

The team found an exciting solution for players who need an extra challenge: added a shield that activates if you avoid a collision with another car three times. The player passes next to another vehicle but does not crash into it. This motivates gamers and adds risk to gameplay. For LTV, this means more reward ads and interstitials if the player fails the first time.

Side note: Advertising and in-game monetization: busting myths.