Talon.One · Product Design · 2025–Present
A/B Test
Talon.One is a promotion engine that unifies promotions and loyalty programs into one strategy — turning customer rewards into measurable results. Disconnected discounts drain margins, and basic loyalty programs fall flat on their own. Experiments is the feature that lets customers test which promotion actually earns its keep, instead of guessing.
- Role
- Product Design
- Time
- 2025–Present
- With
- Talon.One
Ingredients
- Product Design
Experiments is the feature that lets customers test which promotion actually earns its keep, instead of guessing.
Context
Talon.One powers promotions and loyalty programs for some of the world's most-loved brands. But for a long time, the platform had no built-in way to test one campaign against another — customers had to decide which promotion to run on gut feel, or build a workaround themselves.
Problem
There was no reliable way for customers to test campaigns against each other inside the platform. Some found workarounds — running two campaigns in parallel and comparing the data manually — but reading that data was complex and often inaccurate. One especially costly side effect: depending on how a workaround was built, a shopper could see one promotion while logged in and a completely different one if they checked out as a guest. The inconsistency wasn't just messy — it actively undermined the customer experience the campaign was supposed to improve.
That gap is where the idea for Experiments came from: give customers a native, reliable way to A/B test promotions without duct-taping together external tools, segment attributes, and spreadsheets.
Approach — MVP
We started small. The first version of Experiments supported straightforward A/B testing: put one campaign against another with a basic 50/50 traffic split across the audience. Customers who already had an external allocation tool, like Braze, could keep using it and simply pass the assignment into Talon.One. For everyone else, Talon.One could handle the split itself, randomly and reliably assigning customers to a variant.
Keeping the MVP tightly scoped mattered. The goal wasn't to out-build every experimentation platform on day one — it was to solve the two problems customers actually had: no reliable way to compare campaigns, and no guarantee that a single customer would see a consistent experience throughout their journey.
Output — Beyond the MVP
Once Experiments was in front of beta users, we didn't stop there. Based on early feedback, we expanded the feature in several directions:
01
Goals
Customers can now attach a goal to each experiment, and AI-assisted analysis helps them read the results and determine whether that goal was actually met.
02
Flexible allocation
Traffic no longer has to be a flat 50/50 split — customers can set up a control group against a non-control group, or weight the split however their test requires.
03
Audience-based assignment
By combining Experiments with Talon.One's existing audience feature, customers can now allocate specific, pre-defined audiences to each variant — rather than relying purely on random assignment.
Together, these changes moved Experiments from a simple split test into a proper decision-making tool: customers can now ask a real question (which incentive converts best, and at what cost?) and get a confident, statistically grounded answer.
Slide 1 of 5
Reflections
Adoption of Experiments has been strong, and we're continuing active research as more customers bring it into their day-to-day promotion strategy. The bigger opportunity ahead is turning Experiments from a one-off testing tool into a continuous feedback loop — where every test run today makes the next campaign recommendation smarter. We're already scoping what comes next, including support for more than two variants at once and deeper segmentation of results.
You can read more about the feature and its full capabilities in the official announcement: