Conversation Design is the Future of UX
Explore how to design conversations that build better relationships with customers at scale.
Think about the last time you got customer support through a company’s app or website or talked to Alexa or Siri on your phone or smart speaker. Did you navigate through an automated voice or text menu? That was a conversation, and it was informed by conversation design (CxD, for short).
Your impression of the brand you were dealing with was no doubt influenced by how well they designed that conversation. I believe conversation is the next wave of user experience. In my role as the Conversation Design Principal at Salesforce, I empower businesses to create conversational experiences that build user trust.
Every conversation is an opportunity for relationship building, whether to win a new customer or strengthen bonds with a longtime fan of your brand. Successful companies leverage conversation design to minimize pain points and build customer trust. The stakes are high for good conversation design. A U.S.-based survey found that nearly 73% of participants would not use a company’s bot again after a bad experience with it.
On the flip side, poor conversation design can erode trust — fast! It is a sin borne out of a lack of awareness and intention. I mean, why would a company design conversations that alienate their customers? What brand would intentionally build a racially insensitive chatbot? These horrible experiences happen due to a combination of ignorance and language assumptions that ultimately lead to poor design.
So, what is conversation design? Well, it’s not graphical UX nor conversational UI, which are grounded in the visuals of user interfaces and organizing text in a visual space. Rather, CxD is all about the words, emojis, characters, tone, and overall language a user interacts with in a turn-taking pattern.
For example, the copy on a website is copywriting, not conversation design, because the user does not take turns with the words; they’re reading static text without typing anything back. However, a user does interact with a chatbot who is having a conversation via SMS (short message service); this is conversation design. CxD is about being strategic and systematic about what you say and how you say it. Luckily, we have decades of research in linguistics to show us that conversations exhibit clear patterns. Patterns and insights we can leverage in our conversations today.
Let’s explore how conversations are at the heart of relationships. And expand your expertise by learning to create user experiences that build better relationships between your company, your customers, and the rest of the world (also known as Relationship Design).
Here are 5 tips for designing better, more inclusive conversations:
Decide on a conversational style
Depending on what your brand is doing and who you’re talking to, you can determine if your conversational style is light and fun or serious and professional. Let’s consider designing a chatbot for contact tracing COVID-19. Considering the seriousness of the health news your bot is discussing, you’d want to communicate in a caring but credible way with users. This may not be the case with a bot that helps you order fresh fruit for delivery. The appropriate persona for a bot like that could be your friendly seller at your local farmer’s market; relaxed and happy.
Conversational style isn’t particularly special — it’s just the way you talk or chat. In the same way that users expect a webpage to visually convey your organization’s brand, users also expect your bot to express brand style through the language it uses.
People develop their personal conversational style by having conversations with their surrounding community — family, friends, coworkers. Sociolinguist Deborah Tannen discovered in 1984 that, ultimately, what one group of people consider “proper” or “normal” in how to have conversation may be seen as frustrating by others. For instance, overlapping speech is not considered an interruption in some communities in New York, while the style in places like California tends towards waiting for a pause in the conversation before chiming in yourself. Tweaking your bot’s response delay can result in a manner that leans more towards considerateness or involvement — it’s all down to the style that best serves your users, your brand, and your purpose!
Be descriptive
As a conversation designer, you should always be inclusive, observant, and descriptive. Users represent their social identity through their language, which means language is alive, and the rules are always changing. When you let your preconceived notions about how language “should” work, rather than designing from a descriptive, observatory stance of how language works in practice, you run the risk of alienating users. It becomes a usability issue at best and an equity issue at worst.
Remember that:
- Language is a practice. It’s constantly changing. No one person sits somewhere and writes a rule that we all then have to follow. We negotiate the practices of our language as a society.
- When folks refer to “proper grammar” or if something is “grammatically correct,” they’re actually forcing their ideas of socially acceptable conversation on people. There is no right or wrong way to have a conversation.
- Spelling is made up. It rarely represents language in a fashion that is accurate to the sounds we actually produce. There is no such thing as a misspelling because, again, language is a practice, it evolves. Be open to all spelling variations and account for them when you train the bot to recognize user inputs.
Be purposefully marked
I got into conversational linguistics in college because of a breakup. I remember it was my sophomore year, and I was dating this guy. We would communicate over instant messaging all the time in a really casual way. We didn’t capitalize the beginning of our sentences or have much punctuation at all. As time went on, he started adding more formality to his messages with stuff like capitalizing people, places, and things and including a period at the end of his sentence. And I was like, gee, I feel like he’s being distant. And then, lo and behold, we broke up.
At the time it was a major heartbreak, so I was determined to figure it out. Of course, as a linguistic student, I wanted to analyze the language so I found a class on campus called Text and Talk, which I thought was going to be about how to analyze text messages. It turns out it was about six years too early for that kind of research. But it did show me the principles of how to analyze discourse and how conversation works.
In analyzing the text messages with my ex, I found that our baseline or unmarked way of communicating was informal, which could be mapped to closeness. And so when he introduced more formality, that inherently created distance for two reasons. First, it was atypical or marked. More specifically, he was intentionally deviating from our usual way of chatting, thus creating a distant, cold, or even angry conversational style. Second, by shifting his tone and invoking more formal written conventions, he was sending me a meta message that our relationship wasn’t as close as it once was.
If you deviate, even unintentionally, from the baseline style of your conversation design, it will send a meta message to your customer about the state of the interaction and relationship at hand. Once you pick your baseline pattern of communicating with your customer, stick to it, and if you decide to deviate from it, do so purposefully.
Start with a verb
If I had a dollar for every time I heard someone working on a bot say, “but what about the topic? Can the bot stay on topic? What topics will the bot cover?” Unfortunately, the topic is only one half of the whole. Conversation is also about action. Without considering what the user is trying to DO, we won’t ever actually be able to complete tasks at scale for them.
Imagine trying to learn a different language. Let’s use yogurt as the example here (why not?). You wouldn’t want to just learn how to talk ABOUT yogurt. You’d want to know how to BUY it, CRITICIZE it, PRAISE it. In this example, yogurt is the topic, and the verbs convey actions in relation to a topic.
In the context of CxD, users will always be coming to a bot to solve some problem (they’re probably not there to hang out). So, to help them solve their problem quickly if, say, a user wants to speak to a representative, a prompt simply saying “representative” is incomplete and not helpful. However, prompting them to “talk to a representative” is a clear directive.
To get started, look for opportunities to activate language in your daily life. Next time you use an app or chatbot, notice if they use verbs to start a conversation? If not, what verbs would you use in those situations?
Use discourse markers
Don’t get intimidated by the word “discourse” here; it just means written or spoken communication. In this case, we’re talking about words that are often mistaken for filler words, like “so,” “uh,” and “oh.” Although they may seem superfluous, discourse markers are actually one of the pillars of conversational language. They’re really important in creating a natural flow to the conversation.
Discourse markers show the cohesion between different sentences and ideas and how they relate to each other (Schiffrin, 1987). You can think of them as conjunctions; for example, using “and” means you’re building upon an idea; whereas, “yet” means you’re comparing two things. Similar to omitting these connector words, when we don’t use discourse markers, bots can seem rigid and even confusing. But when you include discourse markers, you design an experience that is more akin to the natural conversations we have as humans. And by creating a natural conversational experience, we can better strengthen a relationship with a customer.
They work well, especially for voice interactions. And I use them more sparingly in chat interactions to avoid crowding the message. But I never use these forbidden discourse markers — hmm, uh, umm — because they signify cognitive processing, which we all know machines are incapable of.
Conversation is the currency with which we negotiate and build relationships with people. It’s fundamentally a collaboration, an achievement built with your customer. By leveraging CxD, you can craft back-and-forth exchanges that are more engaging, informative, and ultimately meaningful in the customer’s overall experience of your brand. And, you can reach a broader audience of people by celebrating their dialect or conversational style in your designs. Conversation design is the future of UX because it naturally fosters closer relationships in a way that no other digital experience can.
Want to learn more?
Check out Salesforce’s Conversation Design Module and our Relationship Design Module on Trailhead.
And here are some articles and books that can inform your CxD practice!
New DigitasLBi research shows more than 1 in 3 Americans are willing to make purchases via chatbot from Digitas.
Discourse Markers by Deborah Shiffrin (Cambridge University Press, 1987).
What a father and pre-verbal child can teach us about good conversation by Greg Bennett (UX Collective, 2019).
Conversational style: Analyzing talk among friends by Deborah Schiffrin (Oxford University Press, 1984).
Cooperative Overlap in Conversation by Richard Nordquist (ThoughtCo, 2020).
—
Many thanks to Margaret Seelie, Madeline Davis, and Adam Doti, without whom this article wouldn’t have been possible, and to Noah Kravitz for bringing the conversation design Trailhead module to life. My deepest gratitude to Deborah Tannen for being my north star in linguistics and conversation design—I hope to have done justice by your life-changing research and am honored to introduce it to a whole new audience. This piece is dedicated in memory of Deborah Schiffrin, who, at every turn, always encouraged me to pursue my wildest research ideas.
Learn more about Salesforce Design at www.salesforce.com/design.
Follow us on Twitter at @SalesforceUX.
Check out the Salesforce Lightning Design System.
