Evolv

Introduction

Evolv’s solution revolves around an AI-driven SaaS platform that allows launching multiple UX experiments at a scale never possible before. Evolv started in 2019 and is the result of a technology spin off from Sentient Technologies who spent the better part of decade as an AI research and development firm. That team incubated some innovative algorithms which are now included as a building block in Evolv’s AI-driven optimization platform. Initially, we provided our service to Sentient Technologies; after Sentient sold the majority of its business to Cognizant, Evolv was born from the intellectual property of the offering previously called Sentient Ascend.

Maintaining and upgrading such a robust system requires significant development resources. Since the Redwerk team had already worked with Sentient Technologies and was familiar with the platform’s business logic, Evolv decided to continue our partnership.

Evolv works closely with their clients to define their optimization strategy, generate a constant source of new ideas to be explored, and leverage their proprietary AI to serve progressively better customer experiences and find the best customer journey. Evolv leverages evolutionary algorithms (a branch of AI that uses the principles of natural selection) to determine which combination of different improvements being tested makes up the winning design. An Evolv project represents the long-term optimization effort of a given user journey.

Evolv uses an ensemble of machine learning algorithms and statistical models to gain an in-depth understanding of user behavior and variant performance. Evolv also uses their own UX research to identify areas for improvement and outline multiple changes and their variations that can positively impact conversion. This approach allows Evolv to validate more with orders of magnitude, less traffic, and more robust results.

From there, the UX improvements are coded and tested in Web Editor and then exported to Manager for the actual release. Once the experiment is launched, each variant’s performance is evaluated in relation to the control candidate (original version) in the standard A/B testing mode, with the worst-performing candidates automatically excluded. Once the platform has aggregated a sufficient amount of traffic to move to the evolution phase – EVO – it starts combining variants from different variables using an active learning approach. Because the platform reacts to customer feedback in real-time, it continuously trains its machine learning model to predict the best-performing combinations within a short time. Having detailed statistics on each combination, the client clearly sees what UX changes are driving results toward the target KPI.

Challenge

At the initial stage of our cooperation, we were tasked mostly with coding UX improvements designed by Evolv’s growth team. One time we were involved in the ideation process, which entailed thinking through variable combinations for the experiments to be run.

Witnessing our commitment, expertise, and timely delivery, Evolv decided to expand the scope of the services we provide and fully engaged our team in developing its core product – a pioneering optimization platform consisting of Web Editor and Manager. Web Editor is Evolv’s brand new desktop app that allows coding UX improvements without the client’s source code as well as previewing, editing, and testing the implemented changes, among other handy features. Manager is Evolv’s web platform with advanced testing settings that enables the launch of the experiment itself.

The platform’s complex architecture coupled with extended business logic posed some technical challenges, yet we managed to dive deep into the development phase within a short time. Moreover, we have limited access to some of the platform’s parts; therefore, incomplete understanding of how different services communicate data also created some obstacles. For these reasons, getting to the heart of certain issues demanded more effort at the very beginning of the project.

Solution

We joined Evolv’s development team when they were just starting to re-architect their platform, which allowed us to get to know it step by step. At the initial development stages, we asked as many questions as possible, perhaps even overcommunicating, to gain the most exhaustive understanding of the established business logic. Evolv, in their turn, provided us with timely and comprehensive answers, doing their part to streamline the communication process and synchronize both teams. This joint effort turned out to be quite fruitful as now we can complete most of the frontend tasks assigned to us with maximum independence.

At present, the Redwerk delivery team is focused mainly on continuous upgrading of Evolv’s platform, which consists of Web Editor and Manager. We are proud to know that most of the frontend tasks get delegated to us, and here are just some of the frontend parts we have delivered by far:

• Manager: Show Total Unique Visitors vs Total Unique Converted Visitors – displays all the data on visitors for the entire experiment as well as for each combination of variants.
• Manager: Variable Listing – Tree View – variables are displayed in a tree view instead of a list with the option to group them and check performance data (if available) for each variable.
• Manager: UI – Project Listing – displays project lists with the options to view project history, performance, visitors, goal events.
• Web Editor – QA Configuration – the user has the ability to select one or more variants from any context that matches the current preview URL and apply the variant configuration to the current URL in the Preview panel.
• Web Editor: Cross Browser QA – possibility to create a copyable URL that will help to preview the currently selected variants in the browser.
• Web Editor: Screenshots for variants – allows taking and saving a screenshot of a variant combination currently loaded in the QA Mode.
• Web Editor: Login UI for Web Editor.