7. August 2018

Click-Through VS NLU

Chatbots come in different shapes and forms. Some you are encouraged to have a real conversation with, while others are barely able to chat at all. 

Still, both approaches appear to have their pros and cons. Here, we would like to shed some light on the difference between chatbots based on click-through systems and NLU (Natural Language Understanding).

Click or Type?

Click-through bots offer users buttons to click on, which they respond to in a predefined manner. Essentially, the chatbot leads the users along a strict decision tree without much room to drift off course – if at all.
On the other hand, chatbots that allow complex input as if the user chatting with their best friend were not nearly as wide spread -probably because they are not typically built via a toolkit in one afternoon.
It is only natural that click-through-bots would overflood the market like they did in 2016. Users have been conditioned to write less coherent sentences and rely more heavily on typing in text speak or simply tapping pre-defined buttons. There is a great temptation to cater to that. However, the massive drop of active Facebook click-through-bots perhaps best illustrates how this approach can’t meet the expectations of the demanding user of today. Generally speaking, click-through-bots’ chat abilities are often restricted by the limitations of an app they are integrated in. For these sort of bots, the number of buttons to tap on may be regulated by their platform or the app: either design-wise (a small unchangeable Messenger size can only hold so much before it becomes convoluted), or rules-wise (Maybe the owner of your platform of choice is very stingy with click options for a myriad of reasons and only allows a fixed number of click options – devastating when you can’t add that necessary last fifth option).

The most obvious benefit of NLU-chatbots is the fact that users engage with them in a more natural manner. Free from having to follow a strict path, users write to their heart’s content and receive a satisfying answer. In theory. Should the NLU-chatbot be designed to take context into account, a problem can be solved step by step, without the user explaining the issue over and over again with each new message. Nevertheless, this opens up plenty room for errors as well. Unless the chatbot perfectly understands everything -from the most complex sentence, to the most ridiculous of idioms- users will unearth the chatbot’s limitations sooner or later. Maybe those limitations aren’t actually a big deal. A chatbot claiming to be a shop-assistant for fashionable watches doesn’t necessarily need to be able to answer “Will I need an umbrella tomorrow?” (even if it is likely that a good system can at least process the intent and politely admit „I’m sorry, I can’t answer this. However, what you WILL need is one of our fashionable watches“). But maybe you ask the chatbot to help you out with something it usually masters perfectly and is unable to do so at the moment. Considering the fact that there are many different ways of phrasing the same intent, the chatbot might have not anticipated you to word your request the way you did.

In that regard, click-through-bots offer a much safer alternative: The chatbot doesn’t have trouble understanding you, because it never understands you. The user clicks on one option and receives a pre-written answer. This might just be the perfect solution, right? Well, it would be if the click-through-bot has to deal with the perfect case its “on rails”-approach can cover. But what if the user needs to add an important detail not listed by the bot? Or if it’s a combination of problems? More often than not users will have to go through the entire conversation all over again because they misunderstood the cryptically short click option text and were lead to an irrelevant path. A “back”-button would be only a small comfort if you’re not certain where you went wrong. This results in an inconveniently long “conversation” with the uncertainty if it’s even worth all this trouble. The supposed advantage of simply clicking on buttons over writing long sentences doesn’t hold up when such circumstances are considered.

In comparison, chatbots powered by NLU triumph with an overall speedier experience. The user can jump right into whatever they want to talk about without following any linear paths. It may seem ironic but addressing a problem by writing a full sentence can save valuable time because you don’t have to participate in a game of elimination. Even if the chatbot won’t be able to help you, you will be told immediately. That is if your input was understood correctly.
On the other hand, click-through-bots can help focus overwhelmed users. Imagine an insurance bot that lets you file claims. It can be an emotionally overpowering experience and one might not have the patience to write in full sentences. You shouldn’t be denied service just because you are (literally) speechless.

The Best of Both Worlds

We at Kauz are well aware that click-through-bots and NLU-chatbots both have their advantages and disadvantages. This is why we favor a solution that combines the strengths of these two approaches. Users should be given the choice of a convenient linear conversational path in addition to the free form chat powerful NLU technology allows. Carla from the chocolate manufacturer Bilk operates very much like this. See for yourself how such a hybrid system can lead to much better user experience:

Visit Carla

News mail

Would you like to learn more?

Subscribe to our newsletter