Carlos Carvalho, Senior Consultant junokai
Conversational Design – do you speak Bot? Part 3
This expert tip is about how to use bot dialogues to map well-designed conversation control that works efficiently but is also perceived as pleasant by the user.
Turn-Taking
As already presented in part 2 of the expert tip on Conversational Design, users perceive balanced conversational portions as positive. To maintain or control this balance, there are a variety of conversation control tools.
Example 1:
User: “Hello XXX, what services do you offer?”
Bot: “I can give you general information about junokai or organize an individual meeting with one of our consultants.”
In theory, this concludes the dialogue. However, with the addition of a query, you can request further engagement from the customer and thus keep them in the dialogue.
Example 2:
User: “Hello XXX, what services do you offer?”
Bot: “I can give you general information about junokai or organize an individual meeting with one of our consultants. Are any of the items of interest to you or are you looking for something else?”
Such queries are also helpful to signal to the user that he should now say or write something if he is unsure. In this example, the question about “something else” is also important for identifying possible further points of use for a bot where no service is currently available via the bot.
Barge-In
This element is only relevant for voice bots and can be loosely translated as “falling into words”. Although this is generally perceived as negative, rude, or disrespectful in human-to-human dialogue, it has the opposite effect in bot-to-human communication.
Example 3:
User: “Hello XXX, which Italian restaurants are open in my region right now?”
Bot: “There are ten Italian restaurants open in your region. The ‘Da Carlo’, the ‘Sicilia’, the ‘Da Gianni’, the…”
User: “Gianni sounds good – book me a table for four in 30 minutes, please.”
We give the user with a barge-in option the possibility to jump in within the bot’s response and thus interrupt the bot. This saves him from having to name seven more restaurants in the example and thus saves him time. Now, you might think that having the barge-in option permanently running is a good idea, but it also has disadvantages. With this option activated, the voice bot permanently “pricks up” its ears and background noise such as children, TV or noise can then interrupt the bot unintentionally.
Barge-In should be used specifically for the following elements:
- introductions
- implied confirmations
- options
- decisions
Barge-in shouldn’t be used:
- for prompts when a user needs to hear the entire text
- important instructions
- legal guidelines (e.g., information on call recording)
Tapering
The aim here is to reduce and vary recurring information to such an extent that the user still receives all the important data.
Example 4 (without tapering):
User: “How much does the San Remo bicycle model cost?”
Bot: “The bike model San Remo costs 1,000 euros plus 19%, i.e., 190 euros VAT and 50 euros for shipping, so in total 1240 euros.”
User: “And how much does the Monaco bike model cost?”
Bot: “The bike model Monaco costs 1,200 euros plus 19%, i.e., 228 euros VAT and 50 euros for shipping, so in total 1,478 euros.”
User: “And how much does the Roma bike model cost?”
Bot: “The Roma bike model costs 800 euros plus 19%, i.e., 152 euros VAT and 50 euros for shipping, so in total 1,002 euros.”
Example 5 (with tapering):
User: “How much does the San Remo bicycle model cost?”
Bot: “The bike model San Remo costs 1,000 euros plus 19%, i.e., 190 euros VAT and 50 euros for shipping, so in total 1,240 euros.”
User: “And what is the cost of the bicycle model Monaco?”
Bot: “The bike model Monaco costs 1,200 euros. Including VAT and shipping, it is then 1,478 euros.”
User: “And how much is the Roma bike model?”
Bot: “The Roma bicycle model costs 1,002 euros all inclusive.”
Example 5 is much more pleasant and comes much closer to the human way of communicating.
Jenga-Technique
In the well-known game Jenga, the object is to remove these elements one by one from a wooden tower built of many individual elements without the tower toppling over. This technique is similar in the context of conversational design. Background: Our brain needs about 250 milliseconds to process a word, i.e., to read and process 100 words, 25 seconds are needed. In this case, that means that by saving 50 words, I save 12.5 seconds of “talk time” with the bot. Although this time is not costly for customer service, it means extra time for the customer which they are happy to save if the desired result is still achieved.
Example 6 (before Jenga):
“When you write something in an informative text, you often put a lot of thought into it. Then before you know it, you’ve written something that is way too long and could have been much shorter. Therefore, it is always a good idea to take a step back and look at what you have written first and then see what words can be eliminated. In most cases, you can cut about 50 percent of the words without losing the meaning. Those extra words are completely unnecessary to understand the content.”
Example 7 (after Jenga):
“If you write something that is way too long you can save 50% of the words. These words are superfluous.”
In this example, there is more than a 50% savings and yet the content is intact.
Discourse Marker
Discourse markers are used to explain the effects of data on users in the context of conversational design and to obtain this information from them.
Example 8 (without Discourse Marker):
User: “Please add the jeans to my cart.”
Bot: “I need size information.”
Example 9 (with Discourse Marker):
User: “Please add the jeans to my cart.”
Bot: “I need your size, otherwise I can’t complete the order. I will be happy to make a suggestion for this style based on your previous orders. This way you will get the right size and we avoid unnecessary costs for returns and save the environment at the same time.”
Example 10 (without Discourse Marker):
User: “Please book me a train ticket from Berlin to Munich for tomorrow at 8 a.m.”
Bot: “OK, done. Do you want a seat reservation?”
Example 11 (with Discourse Marker):
User: “Book me a train ticket from Berlin to Munich for 8 a.m. tomorrow, please.”
Bot: “OK done. I recommend making a seat reservation, as this train is already very busy and a free seat will probably not be available tomorrow. I’m happy to search for available seats for you now.”
As you can easily see, examples 9 and 11 seem much friendlier and more convincing to use the service mentioned.
Conclusion: Conversational techniques improves customer experience and usage rates
Conversational design dialogues using the conversation control techniques that we have discussed in this article. It is not mandatory to use each of the techniques in every prompt or dialogue – but a healthy mix of different elements as well as testing by non-involved dialogue testers is advisable. In the next expert tip on conversational design, we will look at the types of errors in bot dialogues and how we or the bot can best deal with them.
About junokai:
junokai is a consulting company based in Berlin that supports clients from different industries in all areas of customer service. junokai was founded in 2013 by experienced managers with extensive professional expertise in sales and marketing, and customer service. The company’s strategic pillar is its operational experience and focus on the areas of customer experience (CX), customer service and sales. For more information on junokai, please visit: www.junokai.de
