https://hgs.cx/wp-content/uploads/2023/01/blog-banner-CDP-1.webp 580 1650 Syeda Rushda https://hgs.cx/wp-content/uploads/2021/10/HGS-Logo_Blue-DK-Grey.svg Syeda Rushda2023-01-23 02:56:352023-01-23 02:56:45How to measure the performance of a Customer Data Platform
Building better chatbots:
7 key considerations
Implementing a chatbot is definitely not a once-and-done affair. Monitoring users’ evolving needs, creating a vision, iterating, and leveraging analytics ensures your chatbot reaches its maximum potential.
According to Gartner, “In two years, the anticipated most valuable technologies revolve around the customer, through digital self-service platforms and understanding customer behavior through analytics.…It is crucial that leaders understand how customers interact with digital channels in order to contain customers within them, and improve their overall customer experience.”
HGS’s director of intelligent automation, Venkata Jagan Saka, couldn’t agree with Gartner’s guidance more — particularly when it comes to chatbots. Saka has observed that, “Delivering a human-centered chatbot experience requires organizations to articulate a vision and define an analytics strategy as part of the chatbot implementation.”
While working with some of the world’s biggest brands, Saka has collected a field-tested list of 7 key considerations for building the best chatbot possible. Customer success managers may find these useful as they implement agent-assist chatbot tools or work to outshine the competition with chat options.
HGS’s recommendations for implementing and maintaining a chatbot include the following.
1. Gain clarity on the people most likely to interact with the chatbot.
Carefully defining the audience for chatbot should be your first concern.
Are chatbot users internal (employees or partners) or external (customers or prospects)? What problems or questions are they trying to address? What device(s) will they be using? How quickly do they need the answer? Where do they most often look for answers already? Do they use other contact channels? What are their demographics?
Creating personas that encapsulate who your users are (as well the anticipated contextual/environmental factors they face) will assist those involved in creating interfaces, programming, and scripting — and receiving transfers when chatbots fail.
2. Think carefully about what type of chatbot to implement.
There are two main types of chatbots: decision-tree chatbots and AI conversational chatbots.
Decision-tree chatbots have clearly defined branches (or user paths) and use structured data. They are useful for addressing very specific, and limited, requirements or questions. Anecdotally, 70‒80% of the chatbots used today are of the decision-tree type.
An AI conversational chatbot, however, captures and provides more unstructured data (content) and is more useful when you cannot anticipate what question will be asked or what information the user needs. An AI conversational chatbot is useful when the customer is more emotional or needs to feel like more like a human being is responding.
If you’re new to chatbots, or looking for a quick win, decision-tree chatbots can be more economical and faster to implement than AI conversational chatbots. Gaining experience with decision-tree chatbots and then switching to an AI conversational chatbot later on as users — and the organization in general — grow more accustomed or dependent on the channel is a strategy well worth considering.
3. Create a short- and long-term vision for your chatbots.
Customer success managers see better chatbot success with roadmaps, goals, and targets.
Your short-term vision for the first 6 months, for example, may include launching a simple decision-tree chatbot for a specific web page, addressing 3 of the most common contact center problems, setting up a dashboard and reporting mechanism that is useful for one or two main departments, reducing call volumes by 5‒10%, and aiming for a 7/10 customer satisfaction score.
Your long-term, 12‒24 month time horizon vision may include expanding the scope of the chatbot to address 5 other call types/problems, reducing the number of transfers to agents, reducing call volumes by another 10%, experimenting with conversational AI chatbots, and providing in-channel end-to-end resolution.
Creating a short- and long-term vision helps to prevent you from getting side-tracked and ensure you are maximizing chatbot potential.
4. Pinpoint specific problems that you can easily solve with a chatbot.
If you’re implementing a chatbot for the first time, it’s best to keep the scope small and simple.
Using frequently asked questions with answers that can be pulled easily from an existing database or knowledgebase is a great place to start.
Answering questions related to such things as tracking number, delivery scheduling, order/payment/claims status, hours of operation, or appointment-booking, for example, are low-hanging fruit. Implementing chatbots for these types of inquiries makes set up, testing, and measurement more straightforward, and it means your staff are able to focus on more complex concerns.
Analyzing incoming inquiries by call volume and call duration may also provide clues on what issues can be effectively addressed by chat — or by updating your website, sales team, or IVR.
5. Issue a survey for every chatbot conversation, and use that data to iterate.
Developing a chatbot is definitely not a once-and-done affair. Getting customer feedback is essential to continually improve chatbot effectiveness.
Two short and simple questions may yield enough information to help you check whether your chatbot is helping or hindering: “Was your question or issue resolved?” and “How was your experience?”.
It can take 3‒6 months to get the details for even a non-AI-enhanced chatbot right. This iteration and adjustment period can be much longer, or ongoing, with an AI conversational chatbot. Using A/B testing techniques methodically to determine the best interface, language used, options and decision trees, transfer points, popup timing, etc. is sure-fire way to build a better chatbot.
6. Actively monitor and analyze chatbot effectiveness.
Your competitors are planning to use — or are already using — analytics and dashboards designed to monitor chatbot satisfaction and effectiveness.
Some of the most effective metrics include response times, chat duration, most common issues, call transfer rate to live agent, languages and geographies, customer type, customer satisfaction, and issue/inquiry resolution rate. Analytics tools that show cross-channel patterns and trends can be particularly effective for increasing and enhancing chatbot use over time.
Pay particular attention to chat time duration, as most chats should be resolved in minutes. If a customer is chatting, or trying to chat, for 15‒20 minutes, there are definitely opportunities for improvement. Chatbots are intended to help people get the right answers, fast, and at their convenience 24/7.
7. Equip staff to handle customer chatbot transfers and escalations.
Agent training should be incorporated into all chatbot development plans.
Share the chatbot purpose, branching, boundaries/limitations, answers, and related database contents with agents so they have a sense of what areas they will be responsible for covering. Step-by-step instructions and tools should be made available on how to ensure a seamless customer experience despite cross-channel transfers.
Staff receiving transfers from chatbots absolutely must have access to the complete chatbot interaction, a sense of the customer sentiment/mood based on natural language processing (NLP) analysis, and a few moments mid-transfer to read the chat prior to engaging. NLP is a particularly useful tool to incorporate with even with the most straightforward chatbots, as the language used by the customer —combined with survey results — will reveal whether they are satisfied with the customer service and the brand in general.
Delivering a human-centered chatbot experience while maximizing chatbot efficiency and potential requires organizations to define analytics strategies and articulate a clear roadmap as part of the chatbot implementation. HGS’s chatbot expert, Saka, notes that — “The short-term and long-term vision of implementing and adopting chatbots is very important. Running analytics regularly to determine whether customers are engaging with chatbots, whether their problem is effectively resolved during the chatbot conversation and what mode of channel they’re using to access the chatbot is critical for long-term success.”
Venkata Jagan Saka, Director of Intelligent Automation, HGS
Recent blog posts: