Avito Launches Trisigma: A New Platform for Automated A/B Testing of Online Services

Avito is introducing a new product aimed at businesses and developers, created by AvitoTech – Trisigma. This platform is designed for automating A/B testing and comprehensive analysis of digital services. Trisigma will assist companies in making informed decisions regarding the quality and effectiveness of new developments and updates to existing products, based on data analysis and experimental outcomes.

Originally created for internal use, Trisigma was first employed within Avito. Through its application, considerable experience has been gained. We recognized the need to share these advancements with the market as a full-fledged product. Soon, Avito plans to broaden its presence in the B2B solutions segment by launching additional products aimed at optimizing and automating development.

The platform allows for the automatic evaluation of new features across various audience segments prior to their widespread release. Companies can simultaneously conduct hundreds of experiments on different platforms and operating systems, as well as across various user segments. Features include flexible test configurations and the generation of detailed results reports.

Trisigma is capable of processing large volumes of data from digital services, calculating and organizing metrics and quality indicators of products. It integrates with existing client analytics systems (various databases, for instance, Clickhouse) to extract data for metric calculations. Furthermore, Trisigma visualizes results in an understandable format, simplifying analysis even for individuals who are not professional analysts.

The service supports a complete testing cycle, from formulating hypotheses and managing experiments to generating reports, with the human element responsible for drawing conclusions and making decisions. Additionally, Trisigma can handle high loads, allowing for the concurrent execution of numerous experiments within a single product, testing multiple new features on various audiences and analyzing their effects on user behavior. The platform is easily scalable and makes efficient use of computational resources.

Six technical capabilities of Trisigma:

1. **Group and weight configuration:** Trisigma allows for unlimited feature segmentation, with each group having customizable weights.
2. **Feature distribution across all devices:** Both web and mobile applications are supported.
3. **Handling a vast number of metrics and segments in experiments:** Tens of thousands of metrics can be seamlessly processed within a single experiment.
4. **Fully transparent experimentation:** Complete history of changes, comments, descriptions, and responsible personnel are maintained.
5. **Safe simultaneous launch of multiple experiments:** Layered architecture allows numerous experiments to run concurrently without conflict risks.
6. **User targeting:** Features can be distributed exclusively to specific users through targeted experiments and data analysis based solely on those who have interacted with the test feature.

Analysts’ time is saved as automation can free up to 90% of their work hours for more complex and creative tasks. The number of tests can increase without hiring additional staff, as the only limitation is the team’s capacity to generate new product ideas. Reliability of results improves by eliminating human error, and product development accelerates with faster implementation of new features.

Trisigma connects the testing of new features with the analysis of their impact on business metrics, enabling companies to gain a comprehensive view of their digital product’s development while avoiding losses from unsuccessful updates.

To learn more about Trisigma, visit here, and you can find the technical documentation of the platform at this link.

The need for an A/B testing platform arose at Avito back in 2016, during which two large-scale redesigns were executed without the benefit of A/B testing. Significant resources were invested in the redesigns, and no one wanted to delay launches, yet A/B testing proved costly. This backfired, leading to a noticeable decline in customer metrics after the releases.

In 2017, the year was earmarked for rapid A/B experiments, and releases began routinely undergoing A/B gates. Analysts conducted manual tests around the clock, highlighting that A/B analytics had become a bottleneck and the time had come to develop a platform, creating demand from the product team.

Key conclusions we reached:

— Without an A/B testing platform, product teams are burdened with manual labor, heavily relying on high-qualified analysts who may need an entire week for a single experiment.
— Implementing Trisigma eliminates these bottlenecks through automation, allowing the number of experiments to depend solely on the volume of deliverable ideas and hypotheses.
— The platform fosters the horizontal expansion of companies, enabling new product teams to emerge. Analysts can simultaneously run a dozen experiments and hundreds annually.

«Avito Tech aims to become a key player in the Russian market for automated A/B testing tools. The integration of Trisigma at Avito has raised the annual number of product experiments from several dozen to over 4,000. Based on our experience, A/B testing contributes an additional 10% to 20% growth in key metrics for Avito, such as revenue and customer numbers,» stated Ivan Guz, Avito’s Managing Partner for Business Development.

The Trisigma team has opened applications for service testing. Existing users include large companies in online education, fast food, and retail, with some already completing their platform integrations.

Trisigma will benefit businesses of all sizes, from startups to large corporations developing digital products for mass markets. However, it is especially valued by companies in e-commerce, banking, streaming services, and educational platforms.

Three signs it’s time to consider an A/B platform:

1. If the time for «simple» solutions has passed, and every product change brings risks and uncertainties.
2. If less risky decisions are needed to minimize uncertainty. A/B testing is an effective way to obtain data in a manageable format.
3. If you want to reduce the cost of conducting experiments, as the expenditure with an A/B platform is always lower. There are no risks associated with «releases without A/B.»

Our calculations suggest that independently developing a similar platform with basic functionalities would cost an average company several tens of millions of rubles and take around two years to complete. In addition, maintaining the platform’s infrastructure and team would require at least 50 million rubles.

In contrast, Trisigma is significantly more economical as it operates in the cloud, with all support and team functions provided by Avito. It does not require substantial technical resources from clients and is easy to scale. The precise cost of using Trisigma depends on the specific needs of each client, the number of tests conducted, and the volume of data processed. Fill out the application here, and we will provide a detailed estimate.

Another AvitoTech development is the Emcee tool for mobile autotests, which you can learn more about here.