Chatbots — mapping the user journey.
The understanding of a digital experience as steps in a journey along a desired path is a great metaphor for both designing digital products and services, and communicating this experience to other members of the product team.
The user journey map developed from this design process is a core UX artifact and is deeply satisfying to build (we know UX designers love a good map).
User journeys are a less useful tool for designing and communicating natural language based, non-deterministic chat experiences. The concept of an ‘optimal’ path through a chatbot experience is sufficient to design simple deterministic chatbots, but once you give users the ability to enter free-form text, they develop the expectation that the chatbot will ‘understand’ them, which can easily knock them right off the path you’ve put all that elbow grease into paving.
The beginning of the design process for a very simple chatbot could look something like this:
The user starts the conversation (the point at where they start depends on what they did last time they were there and/or what we know about them). They can perform one of two different actions, or if the chatbot doesn’t fully understand them, it needs to get further detail from them.
So, if we ignore the complexity in start location, the user journey map looks something like this — users start and either go straight to a specific intent — e.g. Register me for your service — or we clarify what they mean before sending them to the right intent.
When the intent requires multiple pieces of information, for example, providing a name, email address and password, the journey expands:
In this case, each of those individual elements still needs an, ‘I’m sorry, I don’t understand what you mean’, or a ‘this isn’t really an email address’ response.
The user might in fact just give you their name right at the start, without first clarifying that they are trying to register for the service. With this in mind, we need the ability to move through the registration process no matter the order in which we receive key pieces of information, such as a name, email address and password.
And what about cases where a particular field, such as an email address is used in both registration for the service and registration for a newsletter? We need to be able to, as soon as we identify the change in intent, move people between the two types of interactions.
Since the majority of chatbots will have more than two intents, you can see how a map that explores all options is not the most effective tool to communicate with copywriters and developers.
So what do you do when you need to design the most robust and enjoyable chatbot experience possible, and create the necessary creative briefs and specifications for the rest of the chatbot production team?
A good first step for you, as the designer, is still to create a simple, ideal user journey map. But thereafter, you need to start looking at new tools.
We need to think in terms of bi-directional micro-experiences, and their necessary guards and conditions. We need to identify entry points and end states and make sure that both onboarding and offboarding are satisfying across multiple interactions.
Chatbot UX design is a new field and we need new maps to navigate the territory. My next post will explore some of the new tools, methodologies and visualisations that I’ve found useful designing experiences in this new space.
