Cazoo chatbot

My role (solo designer): Service design • UX flow • Iterative UX design • Prototyping • Usability testing • UI design

Project overview

Redefining the chatbot capabilities to support with pre-sale customer queries and elevate pressure on the CS team

The Solution (What)

YEXT integrated Chatbot to serve content directly from the support site

The problem (Why)

Containment rate (the % of interactions that don't reach a human agent)

23.8%

Total transfer requests for pre-order

60%

Transfer requests over time

43.34%

Reduce pre-sale chatbot transfers to CS agents by increasing the containment rate

The Cazoo CS team was handling a high volume of pre-sale customer queries that could have been self-served online. Many customers were transferring to an agent from the "I'm looking to buy a Cazoo car" dialogue, despite extensive information already available on the site.

Each transfer to an agent incurred additional cost and time, while CS agents were already stretched managing after-sales cases individually. The containment rate for this pre-sale dialogue was just 23.8%.

The discovery (How)

Key problems

Initially, it was essential to identify:

Dialogues generating the most unsatisfactory results
Dialogues frequently triggering transfer requests
User messages in the chat prior to transfers
Gaps in the UX flow that hinder self-service

Working closely with the content designer, we analysed chat transcripts to identify opportunities to expand content and investigated the causes behind spikes in bot confusion.

Top queries

Analysis of the data revealed that low satisfaction and transfer requests were mainly linked to queries about 'viewing a car', 'part exchange', 'finance', 'valuation' and 'features and specs', topics for which help content was already available on the support site.

Lack of content

Chatbot transcripts revealed many content gaps that, if addressed, could enable the bot serve relevant responses. However, updating these manually would be time-consuming.

UX flow issues

The existing UX flow responded to most queries with a dialogue and a single article link, but the article often didn’t fully answer the user’s question. Additionally, features like the captcha in the finance flow lacked sub-categories or relevant support content, leaving users wanting to speak to an agent.

Design approach

These findings prompted me to:

Ways to surface more help articles to improve self-service
Ways to provide effective bot responses at key points on user input captcha, like finance, to increase containment
Adjust the UX flow to allow users to select alternative options instead of following a linear path

To understand best practices, I reviewed how other e-commerce sites like Intercom, Squarespace, Zoom, Samsung, Apple, and Tidio used bots to serve responses. A common pattern emerged that bots frequently surfaced help content and support links to guide users and encourage self-service.

The existing Salesforce chatbot had limited content-serving capabilities. Therefore working collaboratively with engineers, we explored integrating YEXT, used for the site search as a possible solution to maximise the bot's responses.

The solution (What)

Hypothesis: Integrating the chatbot with Yext to serve content directly from the support site will enable it to handle most pre-delivery queries, improving containment rates.

I designed a prototype to test the hypothesis, and usability testing confirmed that customers were willing to self-serve by receiving help content from the bot before contacting an agent.

Key improvements included:

Enhanced UX flow with multiple recommended support articles
Better bot understanding of user input captchas for a smoother experience
Option for customers to return to the main menu to aid navigation and self-service
UI updates consistent with the brand and design system

The next iteration is now in discovery looking at how we can embed support within the chatbot so customers can access help content straight away, which should aid in reducing the CS contacts further.

The iterated solution - YEXT integrated chatbot

This new solution was simplified as the thinist slice and launched as an A/B test so the impact can be measured and compared with the current solution.

Success metrics

(30 day period)

contact deflections (containment)

250

CS cost saving of

-£875

Containment increased by

10%

Kirti Devi Ram