Creating an Optimization

Optimization is the main tool of Hyperstone. You specify the parameters to be optimized, indicate the target metric, and the ML algorithm begins automatically selecting the best values for different segments of your users.

To create an optimization, open the required project and click the New Experiment button.

How the page is structured

The optimization creation interface is divided into three columns, which must be filled out strictly in order — from left to right:

Column	Name	What you do
1	Metrics	Select the target metric, secondary metrics, and configuration “stickiness” duration.
2	Parameters	Select project parameters that will be tested in the optimization.
3	Configurations	Specify concrete values (variants) for each selected parameter and indicate the baseline production value.

Fill in all three columns, and then either save the optimization as a draft or launch it immediately.

The bottom line of the screen displays summary information about the current state of the optimization:

Field	Value
Sec. Metrics	Number of connected secondary metrics
Configs	Number of configuration variants
Est. Duration	Estimated minimum and recommended duration of the optimization
Testing Scope	Number of selected parameters
Target Metric	Selected target metric
Stickiness	Selected type of configuration “stickiness”

---

Step 1: Metrics

In this step, you define the indicators the system will use to evaluate the effectiveness of the optimization.

Target Metric (Target Metric)

This is the main indicator that the ML algorithm will optimize. The system relies on it when deciding which configuration variant is better.

Choose one of two modes:

• Recommended — ready-made metrics pre-configured by the system based on typical monetization tasks.
• Custom (Events) — a custom metric built based on your own events.

Select the required metric from the Metric Name dropdown list. Over 30 ready-made metrics are available in the system, divided into categories: revenue, advertising, retention, sessions, engagement, and in-app purchases.

See the full list of all metrics with descriptions, types, and calculation formulas in the Optimization Metrics section.

Which metric to choose?

If your goal is maximum revenue, choose ARPU. The algorithm will select parameter values so that each user brings in as much as possible. For retention, use Retention D1 or D7. If you need a custom metric based on your events, choose Custom (Events) mode.

Secondary Metrics (Secondary Metrics)

Secondary metrics do not affect the optimization algorithm, but they are displayed in reports — they help see the full picture and avoid negative side effects.

By default, the system suggests four recommended secondary metrics:

Metric	Identifier	Type
IAP Revenue per User	iap_revenue_per_user	per user
Ad Revenue per User	ad_revenue_per_user	per user
Revenue per Day	revenue_per_day	per day
Revenue	revenue_total	total

You can remove unnecessary metrics or add your own by clicking + Add Secondary Metric.

Why are secondary metrics needed?

Imagine: you are optimizing ARPU, and it is growing. But at the same time, the secondary metric “Ad Revenue per User” drops sharply. This is a signal that ARPU growth is only achieved through IAP, while advertising revenue suffers. Secondary metrics help notice such situations in time.

Configuration: “Stickiness” (Configuration Stickiness)

Stickiness determines how long a user will receive the same configuration after the first request to the system. If the user opens the game again during this period, the timer resets. If the user has not visited for longer than the set time, they may receive a different configuration upon their next visit.

Available options:

Variant	Stickiness Time	When to use
Session	10 minutes	For optimizations within a single game session.
Same day	16 hours	For optimizations within a single day.
User lifetime (default)	30 days	For most optimizations — the user consistently receives one configuration for a month.

Important

The shorter the “stickiness,” the less stable the user experience. Using Session is justified only when you are sure the configuration should change between sessions. In most cases, use User lifetime.

---

Step 2: Parameters

In this step, you choose which specific parameters of the project the optimization will test.

On the left, a list of all parameters you added to the project during the parameter creation stage is displayed. For each parameter, its type (string, number, boolean) and default value are specified.

Click on a parameter to include it in the optimization. Selected parameters are highlighted, and the Testing Scope counter in the bottom line is updated.

How many parameters to test?

Hyperstone handles multi-dimensional optimizations well — dozens of parameters can be tested simultaneously. However, for the first optimization, we recommend starting with 3–10 parameters to make the results easier to interpret.

---

Step 3: Configurations

In this step, you specify concrete values that the system will test for each selected parameter.

For each parameter, two blocks are configured:

Test Values / Variations

Enter all the variants you want to check. Each variant is a separate “card.” Click + Add Value to add a new variant.

Example: for the parameter ad_interstitial_interval (type number), you can specify variants 30, 60, 90, 120.

Baseline Value

Select from the dropdown list the value that is currently used in your game. This is the reference point — the system will compare the results of all variants exactly with it.

Important

Baseline Value is not just a “default value.” It is the current working value in the production version of the game. If you do not specify it correctly, the comparison results will be incorrect.

---

Launching the Optimization

After configuring all three steps, you have two options:

Button	What happens
Create as Draft	Optimization is saved as a draft. The system does not start the optimization — you can return and refine the settings later.
Start Now	Optimization starts immediately. The system begins distributing users among configuration variants and collecting data.

Tip

If you are unsure about the settings, save the optimization as a draft and double-check the selected parameters and metrics. It can be launched at any time.

What’s next?

After starting the optimization, you can track its results in the Optimization Analytics section.