In our recent webchat alpha we looked at whether there are common webchat needs across government. We also investigated what opportunities there might be to meet those needs in a more consistent way. This involved lots of user research including usability tests, site visits, observations, contextual enquiries and phone interviews.
Researching webchat wasn’t easy. We weren’t just testing an interface, we were testing human interactions too.
Testing interfaces and human interactions
Role play is a useful tool in user research. In this case, we had a member of the team take on the role of webchat advisor, hidden away in the backroom of the lab.
The challenge was to test the script at the same time as we were testing the interface.
To tackle this we visited government contact centres to learn how advisers speak to users. Our observations and interviews gave us a good idea of the tone of voice used by advisers, and different types of webchats.
We then built prototypes for the web interface design. These prototypes allowed us to speak to users in the lab using webchat.
We deliberately designed an impossible task, which we knew users would be unable to complete. Then we watched what happened.
![A screenshot of a webchat process for advisers and potential adviser responses. The text reads: Adviser process: Greet user / ask what their problem is Ask for car reg Ask for date they want to pay for Ask for their credit card number Ask for their expiry date Ask for their security code Try to process the transaction Process this - it works! Give them a transaction reference Close the chat Wait 10-20 seconds before responding each time Make sure you close the chat at the end! Potential adviser responses: [Hi you’re chatting with xxxxx. What can I do for you today?] (prefilled) OK. I can help. What’s your car registration? What date do you want to pay for? That will cost £2.50. Can I take a credit card number? And the card’s expiry date? And the security code - the last 3 digits on the back of the card? Thanks. Let me process this now. The transaction has been processed. Your reference number is 8QU45337.](https://userresearch.blog.gov.uk/wp-content/uploads/sites/102/2016/12/webchat-roleplay-photo-.png)
Testing webchat conversations
To prepare for webchat conversations, we wrote scripts for how we believed the chats might flow.
The scripts contained a list of standard responses, arranged in a set conversational order for the webchat adviser to follow. This allowed us to formulate replies to the most common questions and responses with ease and speed. In practice however, this wasn’t always as easy as we’d expected.
We often had to think quickly because users asked questions that we hadn’t anticipated.
Having more than one person in the back room helped with the chat flow. We could discuss the best response to the questions users were asking. This was especially useful if anything unexpected came up.
We found ourselves operating very much like a contact centre support team.
Testing our hypotheses
After the first round of research, we wanted to test a couple of hypotheses we had about the scripts and how they might influence chats. Our hypotheses were:
- people would respond less favourably to informal language when having to supply personal or financial information
- people were more likely to react well to webchat advisers having access to some of their personal details because it meant they didn’t need to type them in again
To test our assumptions, we created different versions of webchat adviser scripts. We tested tone of voice, formal and informal. We also looked at how the visibility of users personal details affected their use of webchat.
Our findings
We found talking to a human was important to users.
In one of the service tasks, people had to to change their address on their driver’s license. Some users thought they were talking to robots instead of humans. This made them more reluctant to start using webchat.
In another task we asked users to make a payment. We noticed when this request was made in webchat, users thought about the implications in a different way than if it were asked face-to-face, or on the phone.
Some users were concerned with making a payment in webchat. They were worried about their bank details being seen by the webchat adviser in the webchat window. Most users said they would have preferred to make payment over the phone.
In all the tasks, people wanted to know when an adviser was present, when advisers were typing, and if they were still working on their enquiry.
We also learned it’s important to work through user flows for a service task and create webchat adviser scripts. Having a script cuts the length of time spent in webchat for both an adviser and a user, and it also improves the user journey.
Beyond the alpha
We learned some of our assumptions were wrong. Language alone didn’t seem to affect user’s webchat experience either way. This should be tested further by services using webchat, as the role of language may change depending on what is being communicated, when and to whom.
We’ve finished the webchat alpha and Chris Heathcote has blogged about how the different groups of webchat users feel about this support channel.
If you would like to see more detailed webchat alpha findings, please register your interest.
Keep in touch. Sign up to email updates from this blog. Follow Emma on Twitter.