{"id":17776,"date":"2019-05-24T10:30:19","date_gmt":"2019-05-24T09:30:19","guid":{"rendered":"https:\/\/marvel7077.wpengine.com\/?p=17776"},"modified":"2020-12-14T11:30:41","modified_gmt":"2020-12-14T11:30:41","slug":"i-wish-id-known-designing-conversational-uis","status":"publish","type":"post","link":"https:\/\/shopjessicabuckley.com\/blog\/i-wish-id-known-designing-conversational-uis\/","title":{"rendered":"What I Wish I\u2019d Known About Designing Conversational UIs"},"content":{"rendered":"

If there\u2019s one lesson that has been validated in every conversational UI product I\u2019ve worked on, it\u2019s that building the interaction is just the beginning. Testing and iterating are the most time-consuming and headache-inducing (but worth the effort!) parts.<\/p>\n

Here is what I wish I\u2019d known about the design process:<\/p>\n

1. Build with a clear objective.<\/h2>\n
\"\"<\/a>

This is a perfect example of defining a simple objective. The three steps to the objective can be determined later. (Credit: User Testing\u00a0Blog<\/a>)<\/p><\/div>\n

Remember: a bot excellent in one thing is infinitely better than one that is mediocre in three things, so be picky with the objectives. Unlike mobile apps, users stay on one screen, so there isn\u2019t as much leeway to note a change in context by switching states.<\/p>\n

The goal of your chatbot should be apparent throughout the decision process \u2014 from the copy to technical integration options. These objectives will come from discovering who the users are, their lifestyle, and pain points.<\/p>\n

2. Guide users towards this goal, through copy more explicit and straightforward than you may be comfortable with.<\/h2>\n
\"\"<\/a>

This user task flow for web contains the decision points applicable to chatbot products as well. (Credit: Nielsen Norman\u00a0Group<\/a>)<\/p><\/div>\n

Define the different steps in a user journey, and how the bot should be programmed to respond at each of these points.<\/p>\n

Pro-tip from a UX Writer: Be minimal with your copy length and simple with your language choice.<\/strong><\/p>\n

<\/div>
<\/svg><\/div>

\"Make the time required to comprehend each sentence ridiculously short, and the time and effort required to take action even shorter.\"<\/p><\/span><\/blockquote>\n

Try different phrasing to see which one is the clearest for users.<\/p>\n

For example, instead of \u201cI don\u2019t understand,\u201d how about \u201cGet help\u201d so it\u2019s actionable<\/strong>?<\/p>\n

You may find that the first option is preferable because the second option requires the user to guess what they should do with the bot. Instead of making guesses yourself, try different options and ask your users to think out loud.<\/p>\n

3. Experiment with different questions about the users.<\/h2>\n

In my experience, a bot persona<\/a> that truly fulfill your user\u2019s needs requires both what I call \u201cflat data<\/strong>\u201d (age, gender, occupation, etc.) and dynamic data <\/strong>that are situation-specific. The latter type of data could be: their work routine, the thing they hate doing the most at work, etc.<\/p>\n

\"\"<\/a>

Knowing the user\u2019s typical schedule allows you to design a product that complements<\/strong> this schedule. (Credit: bananaworks<\/a>)<\/p><\/div>\n

Knowing users\u2019 schedules and habits in a specific context will help you design a conversation that they could enjoy<\/strong> having, not just endure for productivity\u2019s sake.<\/p>\n

For example, if you know closing out tickets is the IT technicians\u2019 least favorite part of their job, you could have some humor when they\u2019ve gone through all their tickets:<\/p>\n

???? \u201cHooray, it\u2019s only downhill from here!\u201d<\/strong><\/p>\n

When they start with their first ticket, you probably don\u2019t want to have this kind of straightforward humor.<\/p>\n

I was inspired to incorporate questions about the various situations users get placed in by the various \u201cday in the life of\u201d videos from YouTube creators. As you follow along their day, you get a sense of things like what they look forward to doing, what they put off until the very end, and how they wrap up their day. In an ideal world, you would be able to \u201cshadow\u201d your user to understand their day-to-day life.<\/p>\n

With this kind of information, you can design a bot that complements your users\u2019 lifestyle, not just a sliver of their day.<\/p>\n

<\/div>
<\/svg><\/div>

\"A functional interaction only requires efficiency of task completion. A pleasant interaction requires efficiency and having desirable emotions along the way.\"<\/p><\/span><\/blockquote>\n

4. Gather KPI\u2019s that will actually allow you to make changes.<\/h2>\n

Like mobile app products, data points such as the percentage of returning users don\u2019t do much good without knowing which task had the highest rate of completion and which had the lowest.<\/p>\n

However, because conversational UI relies on technology such as natural language processing, it requires that you also have data such as:<\/p>\n