Digital marketing teams have no shortage of data: impressions, opens, clicks, views and conversions. But when testing a new message, subject line, call to action, landing page or user journey, one question remains:

What actually works?

A higher number is not always enough. Teams need to understand whether one version truly performs better, how confident they can be in the result and whether the evidence is strong enough to act.

To address this challenge, LIBRA AI Technologies designed and developed an end-to-end A/B Testing Platform that turns campaign and behavioural data into practical decision support.

More than analytics: a complete software platform

The platform demonstrates LIBRA's ability to combine custom software engineering, behavioural data collection, data infrastructure, advanced analytics, Bayesian statistical analysis, and decision-support dashboards into a single solution.

Users can create and manage experiments, define two variations, configure conversion goals and monitor performance through a dedicated dashboard. Depending on the technical environment, behavioural events can be collected through a tracker or sent through backend APIs.

Under the hood, a modern web interface, a dedicated backend engine and a real-time data pipeline work together to deliver each experiment. Only anonymised behavioural events are ever collected, so every test is built to be GDPR-compliant from the start.

The platform can support experimentation across owned digital channels and applications, including:

  • websites and e-commerce platforms;
  • mobile applications;
  • web applications;
  • games and interactive experiences.

When a campaign runs through a third-party tool, such as an email marketing or survey platform, the relevant analytics can be exported from the platform and imported for analysis. The platform does not replace these tools; it provides a consistent framework for evaluating their A/B-test results.

Advanced analytics for better decisions

The platform uses Bayesian statistics to help teams move beyond a simple comparison of conversion rates.

What business questions can it answer?

  • Which campaign message should we scale?
  • Which version generates more conversions?
  • Is the observed improvement strong enough to act on?
  • What is the risk of choosing the wrong alternative?
  • Should we roll out, improve or reconsider the tested version?

This makes the analysis easier to interpret from a business perspective. Instead of only seeing that Version B generated more conversions, decision-makers can assess whether the evidence is strong enough to roll it out.

Proven in real deployments

The platform has been deployed across a mix of real digital environments as part of large-scale awareness campaigns in Greece, Spain, Germany, South Africa and Colombia (under the EU-funded CHOICE project). Those deploys include direct integrations on the e-fresh and CAAND websites, the Shrink Your Food Waste mobile app, and even the Hotspot Earth game on Steam, where A/B testing capabilities were built into the game via the platform's API to test different framings of in-game sustainability messaging. The platform was also used to analyse exported data from channels such as email marketing, WhatsApp, and Qualtrics surveys. Across both deployment types, the platform supported the same evidence-based decision process, whether events came from its own tracker or a partner tool.

In one large-scale email campaign for dm in Austria, an informational subject line was tested against an emotional one across two customer cohorts of roughly 1.15 million and 875,000 users. The emotional version won decisively in both: a 100% probability of outperforming the alternative, 0% expected loss, and a real conversion-rate uplift of 1.1 to 1.4 percentage points — the kind of clear, risk-quantified answer Bayesian testing is built to produce.

Conclusion

Data alone doesn't tell you what to do next; evidence does. LIBRA's A/B Testing Platform turns every test from a guess dressed up in numbers into a clear, defensible decision.

The real question isn't whether to test. It's whether you can trust the answer enough to act on it.

Test. Measure. Decide with evidence.

Watch the video to learn more about the platform.