Optimization Analytics

After starting an optimization, the system begins collecting data and automatically evaluating configurations. To go to analytics, open your project page and select the desired optimization from the list — you will land on its dashboard.

---

Experiment Overview

The main optimization screen provides a quick answer to the question: what is happening right now and what result has already been achieved.

Dashboard Header

Key information is displayed at the top of the page:

Element	Description
Status	Current state of optimization (Running, Paused, Completed).
Target Metric	Name of the metric the system is optimizing (e.g., Ad Revenue per User).
Best result	The best value of the target metric achieved so far among all configurations. The baseline value — the result of your current production version — is shown next to it for comparison.
Participants	Number of unique users participating in the optimization. The more participants, the higher the statistical significance of the results.

Optimization Progress

The progress bar shows what stage the optimization is at. Scale milestones:

• Min — minimum amount of data for initial preliminary conclusions. Results may change significantly before this point.
• Good — recommended amount of data for reliable conclusions. Results stabilize after reaching this point.
• Done — optimization is complete and has collected the full volume of data. Results are maximally reliable.

Don’t rush to conclusions

Make data-driven decisions only after progress has reached Good or Done. In the early stages, the system has not yet collected enough data, and the “leading” configuration may change.

Top 3 Optimal Configurations

Hyperstone automatically determines three best combinations of parameters based on accumulated data and displays them as cards.

Each card contains:

• Target metric value — result of this configuration (e.g., $0.014).
• List of parameter values — specific values of each parameter in this configuration.
• Metrics Impact — a table with results for all tracked metrics:

Column	Description
Metric	Metric name.
Progress	Visual indicator of change relative to the baseline (green — growth, red — decline).
Result	Current metric value for this configuration.
Baseline	Metric value in your current production version (control group).

• Apply Parameters button — apply the values of this configuration to the project as new default parameter values.
• Recommended badge — the system marks the configuration it considers best based on the specified target metric.

Tip

Don’t rush to click Apply Parameters immediately. Wait until progress reaches Good, and ensure that secondary metrics do not show negative dynamics. Only then apply the configuration.

The See Detailed Analysis button takes you to the expanded analysis mode.

---

Detail Analysis

Detailed analysis is designed for an in-depth study of optimization results. Here you will find tools to understand which parameters actually affect the metric and how they interact with each other.

Configuration Table

A table with all tested configurations and their results for each metric. Allows direct comparison of any two variants and sorting configurations by the required indicator.

Overall Parameter Impact

Summary rating of parameters by their degree of influence on the target metric. Shows which parameters have the greatest effect and which have practically no effect on the result. This helps prioritize planning for subsequent optimizations.

Impact Analysis

Graphs showing the distribution of metric values for each configuration. Each graph displays:

• Median — central value of the result.
• Deviation — spread of values around the median (confidence interval).

The narrower the confidence interval and the higher the median, the more reliable and significant the result of this configuration.

Correlation Analysis

Correlation matrix between parameters and metrics. Allows understanding if there is a strong link between a specific parameter and a change in the metric. High correlation means the parameter significantly affects the result.

Parameter Interactions

Analysis of interactions between parameters: how a combination of several parameters affects the metric. Sometimes two parameters separately have no effect, but together significantly change the result — this section helps discover such dependencies.

Metric Trade-off (Pareto Front)

The Pareto front graph shows configurations that are simultaneously optimal for several metrics. This is especially useful when you need to find a balance between two competing indicators — for example, between ARPU and Ad Impressions per User (more ads give more revenue but reduce user experience).

Configurations on the Pareto front line are the best available compromises: it is no longer possible to improve one metric without worsening another.

---

Optimization Functions

Three additional functions are available in the optimization menu:

Duplicate

Creates an exact copy of the optimization with all settings: parameters, metrics, and configurations.

Important before running a copy

Before running a duplicated optimization, be sure to change the parameter values or variant ranges. If you run a copy with the same values, data from the two optimizations will overlap — this will distort the results of both. Use duplication to quickly create the next iteration based on already verified hypotheses.

Print Study

Generates a PDF report with optimization results: graphs, tables, and key conclusions. Convenient for sharing results with the team or management.

Export to CSV

Exports all optimization data to a CSV file. Use this function if you want to conduct your own analysis in Excel, Google Sheets, or any other data analysis tool.

---

What’s next?

After completing the optimization, two scenarios are possible:

1. Apply the best configuration — click Apply Parameters on the recommended configuration card. New parameter values will become standard for all users.
2. Start the next optimization — use the Duplicate function and test new hypotheses based on the results of the current study.