In our new blog series, we asked our trbo colleagues what their favorite feature in the trbo tool is. Our Senior Client Success Manager Alejandro Guizar picked out a very powerful tool right away and gives us some insights on how this feature works.
1. What trbo feature are you most excited to implement?
I am a big fan of our Multi-Armed Bandit Test.
In the context of A/B testing or marketing campaigns, the ‘multi-armed bandit’ refers to trying multiple variants of an ad or offer at the same time to see which one works best. Unlike traditional A/B testing, where you divide resources evenly among the variants and wait for a statistically significant difference between the groups, our Multi-Armed Bandit test relies on an adaptive strategy where resources are reallocated to the successful variants according to the results.
2. Can you describe the feature of the Multi-Armed Bandit Test in a bit more detail? Where can it be included on a website and what opportunities does it give retailers?
The feature is used to test different variants against each other. But unlike the multi-variant test, this test is not about finding an absolute winner among the variants. The goal is to show visitors different variants so that each user gets the best variant for them.
A few examples of what can be tested with this campaign:
- Product recommendations with different logics, e.g. ‘top sellers’ vs. ‘recommended products’ vs. ‘often bought together’.
- CTAs with different texts, different colors, at variable positions.
- Different banners at the top of the home page.
- Different versions of the checkout, e.g. multi-step vs single page.
3. Warum gefällt dir genau diese Maßnahme so gut?
I like the dynamic aspect of the test. Let’s say the campaign aims to test CTAs with different texts and colors. The system plays out the different variants until the user interacts with one of them. So the more the user interacts with the different CTAs, the better it is to predict which of the CTAs the user prefers. This could be considered personalization: the user gets the variant they prefer to interact with. But it is not static. If the user stops interacting with that variant, the system will show the other options again until a new favorite is identified.
4. An online retailer wants to implement the feature on his website. What advice would you give him before implementing it?
Here’s what I would recommend:
- Make sure you’re ready to run the test for a while.
- If the distribution settings are flexible, the system will have more room to make more accurate predictions.
It is not necessary to have many variations. However, to create a purposeful experience for the user, having multiple options to choose from is beneficial.
5. Ist die Option ein All-Time-Klassiker oder wird der Impact deiner Meinung nach bisher unterschätzt?
It all depends on what the customer actually wants to achieve. But I think the more customers use it, the more ideas are developed for the many uses there are. So it has great potential to become a classic.
6. Für welche Branchen lohnt sich der Einsatz ganz besonders? Kannst du uns ein paar Ergebnisse nennen?
The application possibilities are endless, and therefore it can be used in all industries.
For example, a customer in the fashion industry has tested product recommendations on the homepage. ‘Top sellers’ vs. ‘Recently viewed products’ vs. ‘Recommended products’. The results show that ‘Top Sellers’ was the variant with the best conversion rate and AOV. But the ‘last seen products’ variation generated many more clicks than the other variations. However, we are talking about average figures here, which apply to the entire period. This is because all variations resulted in clicks and conversions during the test period. This proves what the test is trying to achieve. Every user reacts differently! Some users convert when they see the best-selling products, while others do better when they see their most recently visited products on the home page.
If the client had stopped the Multi-Armed Bandit test and set the ‘top seller’ to 100% (for all users), it is likely that many users who respond better to ‘recently viewed’ or ‘product recommendations’ would not have purchased anything at all.
Conclusion: different variations therefore give users more leeway to respond to something they prefer, rather than always being confronted with generic elements. This dynamic aspect of the Multi-Armed Bandit test can greatly improve the user experience.
Want to learn about our trbo tool and its many uses? We’ll be happy to show you how your webshop can benefit as well. Get a free Demo-Termin now!