Chatbot is a technology, a channel and an interactive navigation system to help humans access information and trigger task models.  It is therefore, not any different from a mobile app or a website when it comes to providing a personal experience to every user.

This blog is not about chatbots versus apps versus websites.   Certain tasks are best suited for chatbots and others are better suited for apps.

Chatbots need to be convincing to the users.  Here are some challenges and how they can be overcome.

Putting technology before experience

Chatbot is an exciting technology but, technology should never get in the way of solving a problem or providing a memorable experience to users.

First and foremost, chatbots should solve a business problem.  People want to “book a flight”, “return an item”, “play a playlist”.  No one says, “I am going to book flight using a bot”.

Chatbots that can get a task done with less resistance compared to traditional ways will have much higher engagement.

Further, chatbot is channel.  It may work great for a certain segment of the population while others may insist on continuing to use AOL.  Artificial Intelligence and speech-to-text-to-intent based chatbots get most love but the most successful bots are simple and solve real world problems.

Lack of focus causes frustration

Many chatbots do not have a clear focus, are designed as know-it-all, and end up being frustrating to users

The best chatbots have a clear focus and set the expectations clearly and upfront with the users.

“I can help you pick the right makeup for your skin”.

“I can help you with your past orders and returns”

Hyper-focused chatbots on the other extreme just do one thing, but do it amazingly well.

Focused chatbots are perceived to be less frustrating and people tend to converse within the scope of the chatbot.  Having a clear and personalized on-boarding for users is helpful in setting expectations.

Minimal access to data sources

Chatbots cannot personalize with meaningful information and/or keep conversations engaging if they do not have access to data sources in a consumable manner

I wrote about Meaningful Personalization in Messaging & Conversational Interfaces recently where we used an example:

Your package with Philips Sonicare and 3 more items, will be delivered tomorrow by 7PM via UPS Ground. Would you like to see UPS tracking details, set up a return or see accessories like travel charger or wall mount?

This conversation in this example is concise,  it is keeping the user engaged with leading questions, and also upselling additional items – a win-win for both user and marketer. However, to be able to provide such an experience, the chatbot needs to have access to the order management system, along with shipping and carrier details.

 

Having access to all pertinent integrations and information to personalize a conversation is key to chatbot’s success with individual users.  For example, a chatbot that is not integrated with payment systems cannot accept payments over chat.

Marketers create user profiles based on usage, behavioral, psychographic, geographic, and attitudinal data for behavioral intelligence driven marketing.  Chatbots should have access to same user profiles, but in a chatbot consumable way.  One known way is to tag users when creating profiles and use tags during chats.  Expert and novice photographers, for instance,  can be offered personalized experiences.  Users who prefer to pay with PayPal could be offered that choice first.

Lack of context and traits

Trouble associating and understanding context and a user’s traits

In any intelligent system, context can be derived from

  • the current conversation (for example, Julia may be interested in a telephoto lens)
  • previous conversations (for example Julia owns a Nikon camera body) and
  • user traits, usage and behavior (for example, Julia is a pro-consumer and an avid photographer derived from search and purchase history)

Context from previous conversations is very useful for personalizing answers.

Julia:  “I’m looking for a telephoto lens”

Chatbot:  “Are you looking for the telephoto lens for your Nikon D5 camera?”

This would immediately filter out incompatible products.

Further, understanding the technology limitations of NLP-based, speech-to-text-to and NLP-based chat bots should not be ignored.

Up-selling is much easier when the context is well understood.  A pro consumer photographer is more likely to buy a protective filter for their telephoto lens.

Unable to use feedback

Not being able to use explicit and implicit feedback

The basis of personalization is the ability to personalize experiences based on based on both explicit and implicit feedback.

Customization, configuration and answers to direct questions can be thought of as explicit feedback.  Implicit feedback plays an even more important role in making chatbot conversations personalized for users.

Examples of implicit feedback are time pent reading a blog or observing scrolling activity on suggested products

Unable to parse through feedback

Not being able to apply appropriate weightage to feedback and use it subsequently in the conversation

A typical chatbot would typically get a lot of implicit and explicit feedback.  Being able to intelligently apply the feedback with correct weightage is crucial to building a chatbot user profile.

Lack of fallback strategy

Not having a fallback strategy when the chatbot cannot help a user

Again the chatbot is not the product it is a channel.  Chatbots learn over time and cannot be perfect on day one.  Chatbots should have a multi-level fall back strategy.  For example,

  • Switch to a user interface to ask multiple choice questions
  • Request a real human being to assist
  • Offer to search based on keywords and known context

 

Chatbot systems, and systems in general, that have access to data sources, are focused on saving a problem, care about a user experiences, can utilize user’s traits and context fully will always be more engaging and convincing to users.

 

Posted by Dickey Singh

Dickey Singh is the CEO and co-founder at Pyze and has over two decades of experience in mobile, Big Data and SaaS. He started Pyze to help app publishers engage, retain and grow their mobile users using automation. https://twitter.com/DickeySingh Get Pyze: https://pyze.com