Conversation design is unexpectedly emerging as a top concern for brands, even though “conversations” don’t seem like something we could naturally have with a brand. AI, however, is hitting the market with such a speed and ferocity that Gartner estimates 85% of all customer service interactions will be automated by 2020.
These deployments only just begin with customer service, however. A majority of beauty consumers, for instance, are overwhelmed by the number of product choices available to them, and 74% of consumers say they would find “living profiles” valuable to curate experiences, offers and products they receive.
With these established trends gradually transforming consumer expectations, conversational AI and chatbots are proving to make two-way conversations with “brand” not an outlier, but the norm. Talking with a “brand”, a “brand AI” or a “conversational intelligence” can seem strange and depersonalized, but most consumers actually find it helpful and enjoyable when they do engage with it.
Brands may also find it strange to have a “machine” represent them, but in reality the deployment of Conversational AI allows brands to achieve a scalability, predictability and immediacy in consumer engagement that they couldn’t have achieved before.
Because Conversation Design will inevitably become something you have to consider in your broader marketing efforts, if not something altogether central, here is some context, plus a few tips and pointers, on how to get started and how to offer the best possible conversational experience.
What Makes Conversation Design Different from Everyday UX?
A typical User Experience (UX) encompasses all of a consumer’s interactions with a company, as well as its products or services. Conversation design can thus be considered a part of a user experience.
However, given the immediate responsiveness of things like conversational AI and the direct, predictable, two-way interface we are most accustomed to with messaging products, Conversation Design needs to be considered in pairing with traditional UX, or at least as a separate, interdependent user interface and user experience.
Obviously these aren’t the only limitations. There are solutions like conversational websites and landing pages on the market which make conversations the core of a user experience, and could conceivably be the entire way a use is meant to interact with a company.
While the possibility of this is attractive, conversation design today requires thinking mostly within a two-way messaging framework that complements a traditional websites and ecommerce or mobile experiences.
In developing this kind of experience, you then want to think about how conversation design can complement a traditional user experiences by offering the most value possible. With that in mind, the conversational experiences you built aren’t simply a “cherry on top” – they can be the “pièce de résistance” that enhances your entire user experience, as well as the results it generates.
Setting Goals for Conversation Design
Natural conversations are complex. When we converse with people, we don’t always know what they’re looking for, what they want or what their intentions are. When we think of “designing” a conversation, we’re almost inherently throwing our natural instincts, responsiveness and ability to interpret things like non-verbal cues out the window. To adjust to this, we need to think about setting goals.
So, what are those goals and why do we choose on goal, or multiple goals, over another?
Engagement can come in multiple measures. Whether it’s in the open or click-through rate of your chatbot, the minutes consumers spend with them, how many turns a conversation takes or how quickly consumers respond to prompts during the conversation.
Engagement is usually taken as a way to measure the health of the conversation you’re offering, and how interested your user is in continuing the conversation at any given time. Engagement can also be a driver of revenue, customer satisfaction and more, so it shouldn’t be underestimated in how it can contribute to other goals.
Brands can use conversational ai for marketing, commerce or customer service, but more and more brands are setting goals around generating insights on their consumers. If a brand isn’t accustomed to selling to a consumer on an early interaction, or would prefer they interact with a salesperson (think banks and automotive companies), the offering conversations, product discovery and guided selling through things like conversational AI can allow a conversation to generate unique consumer insight.
Often, the insights a brand generates can serve as a subordinate goal for engagement or revenue, but allow any particular conversation design to generate additional value by being leveraged for other marketing initiatives later on. This can also contribute to developing a holistic view of customers, which 42% of marketers define as a top priority.
Collecting emails is a simple and straightforward goal. Emails are a valuable commodity to all kinds of organizations. Whether in list-building, lead generation or fundraising, conversation design can serve its developer well in engaging consumers with the ultimate end goal being the receipt of an email, or even booking an appointment.
Everybody loves revenue. The bottom line is that “the bottom line” usually determines the success or failure of a particular conversational project. The nuance of conversation design is in how it helps generate substantial incremental revenue.
By engaging consumers, surfacing insight and providing things like guidance and recommendations – as well as enhancing how a user communicates with a brand – good conversation design can drive increases of 30% or more in average revenue per user when compared to traditional web experiences.
Percentage of Web Traffic:
Obviously, things like 30% sales lift per session are an attractive indicator for how effective conversation design can increase a web experience. While also reducing costs on things like serving customers, answering basic requests and providing unique utilities (or previous recommendations) to users, generating more web traffic to a conversation is critical.
While consumer behaviour and simply offering a compelling, sticky experience can do a lot for this, the web traffic you drive to a conversation can both determine its success and much of the broader overlap it has with your traditional UX practices.
Customer satisfaction is something every company is looking for. The old adage goes, it costs 5 times more to acquire a new customer than to keep an existing customer. While including conversational AI on advertising and marketing efforts can help equalize those differences, modern customer expectations dictate that you should offer a unified customer experience no matter where you are in your relationship. Customer satisfaction can thus be a crucial measure to improve, in concert with other goals of course.
A note on goals:
These goals aren’t unitary, end all, be all elements of a conversation. Each of these work together, serve as leading and lagging indicators for one another, and allow you to quantify different elements of your “chat funnel” and optimize your conversation design accordingly.
The Basics of Conversation Design
Your goals determine where conversation design can yield the most value. While unique alternatives could yield better results depending on your goals, conversation design is incredibly useful and adaptable, allowing your organization to develop stronger relationships with users and adapt those goals over time.
Particular to business cases of conversation design, it’s incredibly well-suited to meeting consumer expectations around personalization, connectedness, and privacy. 73% of consumers say that brands are struggling to meet their rising expectations for personalization, and conversation design is unique in its ability to create one-to-one interactions that offer guidance, recommendations, and responses that drive greater interest, loyalty and purchase volumes.
Hence, in thinking from your goals, you will be able to establish a use case or creative path to achieving those goals.
Using Goals to Establish a Use Case
When goals are clear, use cases can be established. For example, if a skincare brand wanted to increase revenue, they might want to use a bot to provide recommendations. Bot needs and user needs can be defined thereafter around providing that service toward the specific goal of revenue.
For a use case, here are a few examples of how different brands have used Conversation Design to achieve their goals:
Skincare brand uses a diagnostic to provide personalized product recommendations
A bank wants to book appointments for financial advisors, so it uses conversation design to find out what individual customers want to prioritize and book for
A grocery chain wants to encourage customers to buy more of their shopping list from store-owned brands, so it offers a conversational experience that gives them recipe suggestions and shopping lists based on their preferences when they’re starved for time
A telecom wants to make it easier for consumers to change their packages, access customer service or choose their next mobile phone. Conversation design can give them instant, automated access that also recommends them the most satisfying products while driving higher repeat revenue
A media company with various advertisers, sponsors, and product reviews wants to be more involved in the transactions consumers make. Conversation design can allow them to provide recommendations and ecommerce checkout, allowing the users to make easy purchasing decisions between brands and get a more satisfying experience, while the media brand increases revenue, stickiness, and authority with consumers.
Defining AI Needs and User Needs
Artificial intelligence has fewer limits as the days go on, but there are still technical limitations and efficiency requirements that determine what can be the most effective conversation design.
Human beings, when it comes to conversations, have few limits, but they do have a lot of preferences. Much as the classical marketing persona concept may evolve, a user persona is critical to understanding the wants, needs and conversation style that users will be both most interested in engaging in and most relevant to your goals.
What are your technical limitations?
Technical limitations with AI can be as follows:
- Ability to recognize unstructured or previously unloaded inputs through Natural Language Understanding and Processing
- Responsiveness, accuracy and clarity of consumer needs and wants, as well as product offerings and categories
- Ability of AI to categorize and easily access large numbers of products and tie them to traits, benefits or most relevant needs
- Limitations of AI around the scale at which products and traits, benefits or needs can be processed or understood. Choosing between 10 products, for instance, is far easier and more seamless in terms of conversation design than choosing between 100 or 1000.
While these aren’t necessarily limitations, these could be capabilities that are necessary to maximize the benefits brought forward by Conversation Design.
If your conversational interface is lacking hard AI capabilities like the ones mentioned above, then your design should be focused almost entirely on the user. The focus, in this case, can thus be things like: what are the most frequent and valuable questions or needs that they have, and what is the most efficient way to automatically address them based on copywriting and pre-programmed responses or conversation flows.
What are your user preferences?
User preferences can vary widely, and really understanding what your users want can come down to (1) their basic cultural outlook, values and most natural messaging style, and (2) what they find lacking or absent from a current user experience. Considering your use case is necessary as a matter of style to understanding your user preferences. Let’s think of our use cases above and how the preferences of users in those cases might differ.
Research has shown that, generally, people are more likely to identify with people who are similar to them. This can apply, too, to where they get advice and confidence in recommendations. Thus, in creating a persona for a conversational interface, you should aim to reflect the users or consumers that the persona will be talking to. While demographics and affinity groups can help determine this, a skincare brand may also want to present a poppy, fun-loving, charismatic presence that makes you feel as if you’re getting guidance and recommendations from a real-live, hip, lovable skincare expert.
Think “Let’s find you some makeup to dazzle” or “I’m going to find the perfect shade of red” instead of “Here is your best product”
Banks aren’t approachable by nature, and a lot of people take banking very seriously. Conversation design can reflect that but doesn’t need to be limited by it. For instance, offering a “virtual assistant” that is still friendly but with a limited personality can be a polite way to converse with consumers. Capital One offers Eno to its customers, which has a little more personality than a plain virtual assistant but doesn’t necessarily “blow the doors” off by being anything other than helpful. Perfect for what a bank’s customers expect, though the technology may be better applied to customer acquisition, rather than replicating functions that are already available elsewhere.
Think “As your virtual assistant, I’m here to offer help with anything you need” instead of “Wow, hey big spender, look at that last credit card bill”
Grocery Chain Recipes and Shopping Lists
Instead of using the traditional way to search for a recipe, somebody may benefit from a recipe chatbot if they are looking for new ideas, creative ways to eat healthier or fun recipes that take allergies or dietary constraints into account. The user base here could be multi-generational but may be middle-aged in that they’re looking for a new experience, or to accommodate the needs and constraints of their children and their children’s friends. In this case, a “friendly guide” persona may work best, avoiding the possibility of “getting in the way” of an experience by simply providing relevant, direct answers in a helpful way.
With that in mind, your copywriting can be polite or even fun but should be direct and not too familiar. You should also think about who the user is preparing food for, like something the kids will love, but will still be healthy and good for them.
Telecom Customer Service
Customer service needs to serve. The personality here is not determined as much by users as by goals. People may want to pay their phone bill or follow up on a complaint, and only after achieving those essential goals are they going to look for something like a new mobile plan or device. The age range and cultural outlook of these users are likely to be more varied (everybody’s got a phone right?), so tailoring the conversation style to a specific group will do more harm than good outside of limited circumstances.
Media Marketplace Recommendations
The marketplace that is offering recommendations could vary its conversation design depending on the users. Let’s think of two different use cases here: B2B software subscriptions and consumer electronics.
B2B software users may generally be younger employees in large organizations, or hip and aware entrepreneurs who are running small but fast-growing businesses. With this in mind, you can add a little bit of personality and flare in your conversation design.
For someone looking for consumer electronics, their age group may vary, but they may also be prototypical “early adopters” or looking to “nerd out” on various granular product details. Showing expertise and being familiar in style, the same way you might get treated by a retail associate somewhere like Best Buy (varying expertise notwithstanding), could be one of the better ways to offer a great experience and achieve the goals of a use case like this.
Think “This bad boy has all the specs you’re looking for and the best picture quality on the market” followed by the ability to request specs, instead of “55 inch display with UHD 4k QLED display and 4 USB ports, as you requested”
Copywriting and Programming Dialogue
User preferences have a great deal of influence on the copywriting style and the conversation flows that should be used. As you can see from the use cases and assumed user bases above, along with the associated examples, your copywriting style should reflect what users expect from the conversation (or what value it offers), and index towards phrasing or conversation designs that they are more likely to respond to.
Iterating your Conversation Design
In designing your conversations, you need to consider how to improve, so measuring the achievement of conversation-based goals is important. Much like how goals are set in website analytics tools, conversation goals can be based on numbers of turns, particular turns or benchmarks (like starting or completing a recommendation diagnostic), or anything else that can be considered a compact element of your whole conversational experience.
What Stops Users From Engaging in Conversations?
We need to have natural conversations. Not having that makes us want to quit. Natural Language Understanding (NLU) is essential to making a chatbot that’s as responsive as possible. Lack of responsiveness is usually the most common complaint among chatbot users because it makes them feel as if they’re not being understood in their own language, which diminishes the value of a chatbot experience and usually happens when not NLU-ready.
Conversational interfaces are still new, be more conservative and let more turns color the conversation. You’ll be surprised how much better this performs, because when we talk with someone, if they can’t keep their mouths shut, often we (as friends, party companions or conversation design users) would rather just leave – particularly if we have something better to do.
Sometimes you can stuff too much into a tiny space. Think of conversations as indefinitely incremental interactions, every increment getting you closer to your goal. Don’t be pushy towards it, just create a flow that the user can’t resist. Similar to the screenshot above, avoiding large blocks of text are essential to keep usability high. It’s a chatbot after all, not a novelbot.
Force people to talk with a bot
Sometimes people don’t want this, particularly in customer service situations. Consent and optionality is crucial to providing users a positive experience, don’t force a conversation on them at the cost of that.
Bots can and should be integrated with other customer information, where relevant. Requiring people to repeat themselves is never good.
Lack of Clarity
When we talk to a chatbot, it’s rarely for the sake of conversation. We have a particular expectation and a particular goal. This means that calls to action which bring us to a chatbot experience must be aligned with an outcome (e.g. we can’t promise free samples and then force people to buy something). It also means that plain-spokenness is preferable. Complex sentences will confuse users while keeping the “reading level” of a bot accessible will make it easier to use.
Marketing, Improving Conversion and Enhancing Customer Experience
While conversation design is a critical new discipline for which designers, marketers, and brands need to be sufficiently prepared, we’ve already been having conversations with users and consumers as long as we’ve practiced marketing. Implicitly, the repartee we have with customers – back and forth interactions crossing channels and media – is a conversation that simply takes place at a different pace, with different participants and a different sense of personalization.