The Lott–Customer Service Voice app
Conversation design of smart speaker apps for Australia's official lotteries.
My role: Conversation Designer & Project Manager (client project)
Credits: TalkVia tech team
Platform: Amazon Alexa & Google Assistant
Overview
Having just launched their first fully functional Android app, the Digital team at The Lott (now “The Lottery Corporation”) was keen to innovate by launching voice apps for Amazon Alexa and Google Assistant.
The main aim of these voice apps was to answer common user queries, such as the latest results and jackpot totals, and generally get people excited about playing games such as Powerball and Oz Lotto.
NOTE: Many images are low-res or blurred and cannot be expanded for confidentiality purposes.
Please contact me if you would like further information.
PROJECT PLANNING
I was both Project Manager and Conversation Designer for the project, so my initial focus was to map out a timeline with key milestones as the project had an aggressive proposed timeline of 8-10 weeks.
I updated the timeline on a weekly basis for full transparency to the client as to how we were tracking against the original schedule, and scheduled weekly calls with the key stakeholder.
I also created a Trello board for all tasks and questions, which also had links to shared resources for easy access by the client
RESEARCH & DISCOVERY
Some pre-work had already been completed by TalkVia prior to me joining the team, so I reviewed all of this early groundwork to help me understand the problem space before I began any further research of my own.
I carried out competitor analysis of other lottery smart speaker apps worldwide–especially Aus, US & UK–to compare functionality and identify any potential improvements that we might make in The Lott voice app.
The client provided detailed SEO data which I analysed to identify the most commonly asked user questions. I was then able to use those key user questions as a guideline for the MVP content and also to inform the intent training data.
IDEATION
Once I had identified the core user stories, I created a content map to visualise the core intents, and to highlight content functionality which had been considered but would not be included in the MVP.
This content map was simply indicative of the key use cases, but the voice experience was to be conversational rather than having a tree structure like an IVR.
As I turned my attention to the dialogue, I referenced the brand guidelines provided by the client to ensure the appropriate tone of voice in the conversation design. Persona and tone of voice details were kept fairly high-level at the client's request, as well as due to the tight project timeline. I focussed on the essentials from the brand guidelines, with examples of discourse markers and standard vocabulary.
I created several sample dialogues which I walked our client stakeholder through before starting work on more in-depth conversation design.
I drafted the initial design within the first few weeks of the project to ensure that we would be able to keep to the 8-10 week timeline. However client feedback was delayed significantly due to higher priority projects on the client-side. As a result it was 6 weeks before I received the first round of feedback, and a further 6 weeks before a subsequent round of feedback from another set of stakeholders.
DESIGN & DEVELOPMENT
Following the initial rounds of client feedback, I worked on iterating the design throughout the course of the next couple of months as further client feedback came in, very slowly. Build work happened almost in parallel to the design updates as–due to the nature of the TalkVia platform–dev changes can be made very quickly.
The app relied heavily on API calls to obtain lottery data to return to users, such as latest results or historical results. Subsequently there were a lot of slots and variables to be considered in some of the dialogue.
Surprisingly, complex flow diagrams weren't necessary as the app provided fairly straightforward Q&A responses and offered relevant follow-up questions, rather than interactions being transactional.
The only flow diagram that was necessary illustrated the disambiguation for any intents that would require clarification of slots, such as the lottery game brand, or the date of the results, to ensure the bot answer would correctly fulfil the user intent.
Throughout the course of the project, several features were removed from scope, to be considered for a Phase 2 project–an example of this is the ability for lottery ticket holders to check if their ticket had won.
The “check ticket” use case was initially sold to the business as a key function of the voice app but in reality it would have been a poor user experience, as users would have been required to speak their 24-digit ticket number to the voice app, and for this number to hopefully be recognised correctly and then be confirmed back to them. The flow would have been cumbersome and had significant potential for failure.
TESTING & VALIDATION
I had included time in the project schedule for user research, and arranged with the client stakeholder to carry out this research in collaboration with their in-house UX team. Unfortunately, due to the extensive delays with design feedback and approvals, the key stakeholder made the decision not to go ahead with the user research so as not to delay progress any further.
I was incredibly disappointed about this, but understood that there was now significant pressure to deliver.
After over five months, the voice apps were finally approved and we were able to commence our own internal QA and hand over to the client for their UAT. I was at least able to make some minor design improvements following feedback from these rounds of testing.
CONCLUSION
Our client was absolutely delighted when the voice app passed certification on Alexa and Google.
The Lott Digital and Marketing teams ran an excellent marketing launch and created a useful info page on their website, which is essential to help educate users about a brand’s smart speaker apps, where discoverability is a real issue.
I’m hoping that in a subsequent phase of work we can include enhanced functionality such as account-linking and also add branding to the voice app by using a VO talent to record the audios, instead of the standard TTS voices. This was one of the items that had been removed from the MVP scope, as the client needed to finalise internally who would be the Voice of the voice app.