How Orange delivers
the right offer at the right time with AI

Orange France is the leading telecom provider in France, bringing together the Orange brand (23 million unique visitors) and Sosh (23 million unique visitors). The majority of traffic on these sites comes from mobile users, accounting for 53% of visitors for Orange and 68% for Sosh.

Since 2017, their experimentation has gained momentum with the launch of a dedicated CRO expertise center, supporting teams and promoting a culture of experimentation.

Cutting-edge tools for continuous improvement at Orange France

• Kameleoon — To conduct all experiments and personalization initiatives, including on SPA pages.
• Contentsquare — To detect issues and analyze test results in depth.

Additional tools like GA4 (for data collection) and Qualtrics (formerly Get Feedback, for gathering customer insights) also enhance their efforts.

How Orange iterates A/B tests 
with Kameleoon x Contentsquare‍

‍Orange follows a methodical approach, systematically analyzing test results to optimize and iterate efficiently.

‍For example, after testing two different phrasings for an exclusive customer offer in their personal space, the team used Contentsquare to dig deeper. The result? 1 in 5 visitors didn’t even notice the offer.

To address this, a new test was launched to assess the impact of placing the offer higher on the page. The objective? Ensure that greater visibility doesn’t negatively impact interactions with other links.

By combining Contentsquare and Kameleoon, Orange accelerates the process from insight to action, a crucial factor in optimizing conversions.

Orange personalizes the visitor experience in real time with AI

Feeding Kameleoon’s AI with behavioral data

‍Orange offers a €10/month discount on its "Livebox Fibre" plan but wants to reserve it only for hesitant visitors. This approach preserves the experience for uninterested visitors, and avoids giving discounts to already-convinced buyers, protecting profit margins.

To achieve this, Kameleoon’s AI analyzes visitor behavior on the Orange site over several weeks. Once trained, the algorithm can identify three audience types, and predict conversion probability in real time.

‍Validating the accuracy of predictive targeting

‍Once the model was calibrated, Orange tested the effectiveness of AI-driven personalization through a split test with three variations:

• ‍Original version – No promo code displayed (control group).
• ‍Variation 1 – A promo code is shown to all visitors by default.
• ‍Variation 2 – A promo code is displayed only to visitors identified in real time as hesitant by AI.

‍This test aimed to measure the impact and relevance of AI-based personalization on conversions.

‍AI helps Orange protect its margins by targeting only hesitant visitors

‍Without AI, 68% of purchases included a promo code. With AI, only 25% of purchases used a promo code. By reducing unnecessary discounting, Orange preserved its margins, ensuring that only truly hesitant buyers received the incentive.

At the end of the test, the results showed a +11.6% increase in conversion rate vs. original (no promo code) and a +7.6% increase in conversion rate vs. Variation 1 (promo offered to everyone).

The future of experimentation at Orange

‍This first AI-driven personalization experience marks the beginning of a broader strategy to offer tailor-made experiences to each Orange user. The company now plans to go further by integrating product recommendation modules.

With 5+ years of CRO expertise, the Orange CRO team has reached a level of maturity that allows it to tackle even bigger challenges. Next step? Expanding AI experimentation to the Orange and MySosh apps.