Chatbot UX: If every pixel matters, then also every letter matters
37signals once wrote “Copywriting is interface design. Great interfaces are written. If you think every pixel matters then you also need to think every letter matters.” What can we learn from usability best practices when designing conversational UI?
In 1997 Jackob Nielsen (the UX guru and co-founder of Nielsen Norman Group) posted a research article about how people read on the web. The research findings proved to be quite durable throughout the years and were repeatedly proven in everyday UX work.
The top 3 recommendations from that research were:
- one idea per paragraph (users will skip over any additional ideas if they are not caught by the first few words in the paragraph)
- use the inverted pyramid style, starting with the conclusion
- half the word count (or less) than conventional writing
So let’s see how these and other usability practices will make your chatbot messages more engaging and usable.
Convey one idea per message
When your bot does many things or needs to explain many options, you might be tempted to try to do it with a single message.
Here is an example from Hipmunk’s chatbot for flight and hotel booking.
When you try to do a flight search that doesn’t have any results, they present you with:
- Acknowledgment message that they understood your to and from airports.
- Message that the search could take a while
- No results message
- Suggestions message
So they are trying to tell me 4 important things with these two large messages. And as you can see from the screenshot they take pretty much the whole screen on my phone. At the end, it looks like a brick wall. And who wants to bang their head into a wall?
To make this more scannable, you can present each of these 4 important pieces of information with a separate message.
This way even if a user is disappointed that there are no results they will better see the hint to refine their search term.
But this example could be further enhanced by the other usability recommendations…
Say the important stuff in the beginning
Because people do not read every word you write (or your bot says), but only the first few word of each paragraph, you need to make sure that you say what’s important in these first words. Otherwise, the whole message could be ignored.
So in the example above Hipmunk bot’s cute exclamation “Whoa! This has got to be a first!” does not convey the important information first — that there are no matching flights. So the user could be misled about what just happened and the following suggestion to change the locations or dates.
Say it shorter — half the word count (or less) than conventional writing
In a conversational UI, large blocks of text are pretty uncommon. People just don’t chat that way. Especially on mobile where typing is slower and harder because of the form factor. Also, there are times that we even have 2–3 conversation going on at the same time in a single chat with a person. Interruption of thought is the norm.
And because it is typical for people to multitask or get distracted when using messaging, it is important to convey your message as quickly and as clearly as possible. Let’s look at a message that is too chatty.
The text above basically tells you that there will be 3 links below and you can click them. Which is completely unnecessary. It will make people read the same information twice and thus waste their time and undermine their confidence in your service. Not to mention that the “Customer Service” command is repeated once more in the bot extension menu.
Also, most messaging platforms have a limit of characters count per message (640 characters in Messanger for example). But these limits are far too high.
Try to constrain yourself into 100–120 characters (less than the Tweet limit) or just a few words. If you need to say something longer — split it into several messages. And for each message, of course, put the important information in the beginning.
Here is one more example how to be brief.
I was chatting with a chatbot and insisted on my reply to its question. So at some point in the conversation, it simply sent me an emoticon. Made me smile and conveyed the right message.
Read your messages out loud in one breath
Even if your followed writing best practices, it is always a good idea to test your bot’s messaging.
The simplest way to test it is to just read a conversation with your bot out loud. This way you’ll immediately feel where the structure isn’t right, or where it sounds superficial. For longer messages try to read them in one breath. If you can’t — it’s too long.
The other way is to catch a colleague or a friend and try to do the chatbot conversation but in a real dialogue. You’ll quickly see where your messaging is not clear enough, or you need to repeat it several times to be understood. This way you’ll also be able to capture the real language that people use when conversing with your bot.
Monitor real usage of your bot
And, of course, keep monitoring the conversations your bot is having with people. You will be able to see trends in where people get stuck or experience difficulties.
As a conclusion here is a quote from Signals vs Noise (the people behind Basecamp) blog post. It says sums very well why copywriting is so important for user experience:
Copywriting is interface design. Great interfaces are written. If you think every pixel matters then you also need to think every letter matters.
Good writing is good design. It’s a rare exception where words don’t accompany design. Icons with names, form fields with examples, buttons with labels, step by step instructions in a process. Clearly explaining your refund policy is interface design.
Do not skimp on writing. Pay attention to what you are asking people to read. Read what you write out loud. Don’t use seven words when four will do.
Write good interfaces.
I can only add — write good conversational UIs. Using Darvin.ai.
