Drinkaware
Chatbot

An interactive tool to provide an engaging, personalised experience for users looking for information and support with alcohol.

Overview

The Challenge

The Drinkaware website is a useful resource for various user groups looking for information and advice around alcohol, however due to the breadth of information available on the site it can at times be challenging for users to navigate.

In addition to this, there is currently too much guesswork determining what our users really want or need from the Drinkaware service. 

The Goal

Providing Drinkaware users a service that provides a more personalised and engaging experience that can quickly direct them to the information and advice they need.

Gain further understanding into the specific needs, intents, and motivations of the diverse Drinkaware user groups.

In the first three months post launch, the Drinkaware chatbot engaged
10,216
unique users.

Product launch: September 2021

Project timeline: 6 months

Role: UX Design, Research & User Testing

Deliverables: Research findings presentation decks, User Testing plan and report, UAT report


Developed with ScreenMedia

Background

I joined the project in April 2021, prior to which a series of initial user research had already taken place collaboratively between the Drinkaware and ScreenMedia product teams.

Two primary personas had been chosen as the main target audiences for the initial development of the chatbot - users who are ‘trying to change’ and users ‘considering change’ based on Drinkaware’s ‘State of Mind’ user profiles.

Prioritisation Workshops

Establishing what to prioritise for MVP chatbot development to meet initial user needs and business requirements.

What features do we prioritise?

Through a series of workshops with the ScreenMedia and Drinkaware product teams, a core set of 4 key features were identified for the initial product development and trial. These were considered to provide the most value to users while balancing the business requirements.

Users would be able to take a ‘risk assessment’ of their drinking level

1

Users could get quick links to Drinkaware’s existing tools

2

Users could ask for and receive general info about alcohol and drinking

3

Users could sign up to receive monthly tips on managing their alcohol consumption

4

What personality do we give our chatbot?

Initial challenges were faced in establishing the ‘personality’ and tone-of-voice of the Drinkaware chatbot, with some key questions raised:

Would Drinkaware users react best to a casual tone, or more formal and factual?

Will users understand that it’s an AI bot and not a human they are talking to?

Do we give the chatbot a name and ‘identity’?

I conducted research into existing chatbot conversation design and ‘personalities’ to determine what, if any, standards already existed around crafting a personality for a chatbot.

This initial research showed that the naming, personality and language of chatbots can vary greatly and is largely dependent on the overall business communication style and industry.

This research was presented to the Drinkaware UX, Drinkaware Communications and ScreenMedia team and a workshop was held to decide the direction of the chatbot language style and tone of voice.

The main consideration that needed to be taken into account was the content of the Drinkaware website, and the reasons users visit to our website. Given the serious nature of the content, the chatbot needed to match this tone, striking a balance between conversational and formal communication styles while avoiding ‘technical jargon’ and maintaining a friendly and supportive tone.

Along these same lines the decision was made not to give the chatbot a personalised ‘name’, in order to try and avoid ambiguity in who users of the chatbot are talking to.

Prototype Development

The development of the Drinkaware chatbot prototype was handed over to ScreenMedia, who created a usable prototype in Botmock for testing and review by the Drinkaware UX team.

Putting it to the test….

Once the Botmock prototype was developed from ScreenMedia, I developed a user testing plan around the existing programmed journeys to establish the user experience around:

Chatbot tone of voice & conversational language

1

Visual design and brand representation

2

Ease of use and technical performance

3

Clarity of CTAs and directional guides

4

Trustworthiness and perception in the eyes of the user

5

Possibilities to differentiate between different user groups and provide personalised content

6

Positioning of the chatbot within the Drinkaware digital platform and possibilities for deployment in third-party platforms

7

As the prototype of the chatbot was in an early stage of development with limited functionality, the user testing was conducted as remote 1-2-1 sessions that guided participants through a sequence of tasks while prompted to share their experience and respond to additional follow up questions.

Four remote user testing sessions were completed over the space of a week. Overall response to the chatbot was positive, with test participants validating a number of design decisions including;

  • the generic naming of the chatbot

  • the conversational language and tone of voice of the chatbot

  • the use of infographic presentation of the self-assessment test results

  • the use of chips to give quick responses

  • the overall visual design of the chatbot

While no major errors were experienced by users in completing the tasks, a number of small ‘errors’ and dislikes were noted to be addressed before moving forward with further development.

While a few of these errors could be contributed to limitations with the prototype, there were some key improvements to pick up with chatbot developers to improve the user experience including;

The speed at which messages from the chatbot appeared

The length of chatbot messages and size on the screen

Allowing for a more direct user journey to finding information if users want to skip the chatbot ‘menu’ of options

Notes from our user testing were handed back to ScreenMedia to incorporate into further technical development of the chatbot. The UI design of the chatbot was completed by ScreenMedia in line with existing Drinkaware Design Guidelines.

A final round of UAT testing was completed with internal and external Drinkaware stakeholders to pick up and amend any final error prior to launch.

Chatbot Launch

The Drinkaware chatbot was launched online at drinkaware.co.uk early September 2021.

Initial Results

  • In the first two months, 78.5% of users indicated they were satisfied with their experience using the chatbot

  • Within three months the chatbot engaged over 10k unique users

Next Steps

Analysis of chatbot statistics and direct user feedback through the chatbot’s initial three months post-launch indicated a number of opportunities to address in the next round of development, as well as ‘Phase 2’ of feature development. This work commenced mid-December 2021.