Ascend by Evolv: Artificial Intelligence‐Based Massively Multivariate Conversion Rate Optimization

Abstract

Conversion rate optimization (CRO) means designing an e-commerce web interface so that as many users as possible take a desired action such as registering for an account, requesting a contact, or making a purchase. Such design is usually done by hand, evaluating one change at a time through A/B testing, evaluating all combinations of two or three variables through multivariate testing, or evaluating multiple variables independently. Traditional CRO is thus limited to a small fraction of the design space only, and often misses important interactions between the design variables. This article describes Ascend by Evolv,1 an automatic CRO system that uses evolutionary search to discover effective web interfaces given a human-designed search space. Design candidates are evaluated in parallel online with real users, making it possible to discover and use interactions between the design elements that are difficult to identify otherwise. A commercial product since September 2016, Ascend has been applied to numerous web interfaces across industries and search space sizes, with up to fourfold improvements over human design. Ascend can therefore be seen as massively multivariate CRO made possible by artificial intelligence.

Creative AI Through Evolutionary Computation

“…The first example is Ascend by Evolv, an actual commercial application of evolutionary computation on conversion optimization, i.e. on designing web interfaces to make it more likely that a user will take the desired action on the page, such as signing them up, buying something, or requesting for more information [23,25].…”

Data-driven software design with Constraint Oriented Multi-variate Bandit Optimization (COMBO)

“…The most similar work to ours is by Hill et al (2017) at Amazon where the problem formulation of multi-variate multi-armed bandits (see next section) is first formulated and addressed with a solution. There is also related work on search-based methods (Iitsuka and Matsuo 2015;Miikkulainen et al 2017;Tamburrelli and Margara 2014) and hybrids between search-based and bandit optimization (Miikkulainen et al 2018;Ros et al 2017).…”

“…If the implementation cannot be broadly applied to many parts of the software product this investment might not pay dividends. Previously, data-driven optimization algorithms have only been applied to simple software (Hill et al 2017;Miikkulainen et al 2018) with a flat structure in the decision variables, such as colors, layouts, texts, and so on. Software with more complex behaviours cannot be directly optimized with these techniques.…”

Enhancing Evolutionary Conversion Rate Optimization via Multi-Armed Bandit Algorithms

“…In evolutionary CRO (Miikkulainen et al 2017a;Miikkulainen et al 2018), each genome represents a web interface design. The search space is pre-defined by the web designer.…”

“…Although this approach leads to impressive improvements over human design (Miikkulainen et al 2017a;Miikkulainen et al 2018), several open issues remain in evolutionary CRO. First, candidate designs are expensive to evaluate, and traffic is often wasted on bad designs.…”