Adoption of chatbot from small to big business is without a doubt, increasing at an incredible pace. Based on the survey done by Business insider, more than 80% of the business want chatbot to be part of them by 2020. You can read here to read more about it.
There are many ways to utilise chatbot from guiding the customer throughout the journey, sending content, follow up and even helping to qualify a lead.
At Monocal, we utilise a chatbot to qualify a lead, educate and try to make a sale. Since our adoption of Chatbot on January 2019, we managed to maintain our 6 sales figure by reducing 50% of our headcount (from 10 to 5).
Although it is not a straight forward process, we want to share 7 tips on how we achieve this result.
Chatbot Using Proven Sales Script
Depending on the purpose of the chatbot, it all starts with the flow (discussed later) and the script. Chatbot is garbage in, garbage out. Without proven sales script, in our case, the sales script has been vetted multiple time and it was actually heavily used by our salesperson. The script itself has generated 6 figures sale.
Proven in this case, multiple salespeople use that script generated more or less similar amount of sale. Therefore, the sale result is pretty much predictable if we use the script.
Once we have confidence with the script, it will be deployed on our chatbot and intensive monitoring will be don in the early phase.
Clear Workflow & Responsibilities
Why would we need a clear workflow & responsibilities? In our case, we have defined the certain task for our salesperson to execute. Each of the lead coming through our bot will be automatically assigned to our salesperson in sequence. It is distributed evenly and this also increases the visibility and responsibilities.
One of the workflows that we have defined is when the bot is having a problem to understand the lead question or statement, our salesperson will intervene the chat and resume chatting with the prospect.
With this workflow, it helps to reduce the bot bounce rate. Bot bounce rate in this case simply a term that I use to represent whereby lead tries to restart the bot by using the word “Reset” or “Restart” or even starting to feel upset with an unanswered question from them.
In Chatfuel, you have “Live Chat” which you can offer to the prospect an option to engage with the operator or salesperson.
Integration with Artificial Intelligence (AI) or Neuro-linguistic Programming (NLP) Engine
The most feedback that I receive when I met the business owners, they have their own reason on why they are late to adopt chatbot. The most common answer that I received is chatbot is not suitable for their business, it is too difficult, the prospect doesn’t like to engage with chatbot and many others.
While I didn’t disagree with that, however, there is no solid proof of experience that says that. Almost every single business owner think that their business is unique. However, as I always mentioned to most of them, it’s time to experiment with chatbot and see if it going to work.
In our situation, we are using Dialogflow as our NLP engine. What is Dialoflow you may ask?
Give users new ways to interact with your product by building engaging voice and text-based conversational interfaces, such as voice apps and chatbots, powered by AI. Connect with users on your website, mobile app, the Google Assistant, Amazon Alexa, Facebook Messenger, and other popular platforms and devices.source : Dialogflow.com
There are stark differences between Chatbot that uses button and Chatbot backed by AI. The button-esque is straight forward and based on our finding, a niche that focuses on median users of 45 years old and above tend to abandon chatbot interaction that uses the button.
Is this a robot?
Well, that is the most favourite question when we use the button on the early stage of our bot development. However, when we start to adopt with AI, our conversation flow become better and subsequently, our chatbot engagement is high than ever.
The process of nurturing AI doesn’t stop although we deploy to our LIVE page. It’s still happening right now. Months after we take it LIVE!
Below is the analytics on Dialogflow page that shows there is still 25% intent that the bot itself not able to determine at this time. It just needs more training.
Removing the Bottleneck
Bottleneck in chatbot sense is whereby a bunch of users stuck at certain flow, repeatedly. When designing chatbot, all the countermeasure need to be put in place. Although they are a lot of chatbot analytics out there such as Botanalytics or Recast.Ai, it might not be appropriate for a startup or small to medium business users.
These countermeasures need to be designed way before we start building the chatbot blocks. Without the proper design, to remove the bottle neck might be tricky when you do not have a quick look at where you bot stuck at.
When building small scale chatbot with less than 20-30 blocks, it is easy to determine where the users stuck – you chatbot builder might even remember where is the bottleneck.
However, when your chatbot adoption become larger and complex, the bottleneck is less visible. Couple with a huge influx of users who did engage with your chatbot daily, you need a smarter way to view these bottleneck. It does feel like finding a needle in a haystack but it is a task need to be designed in advance. (business owners, that is not your task)
So how do you exactly remove bottleneck? In general, below are our tips:
- To have proper documentation and bot schematics
- To have a marker in place
- To have a reporting mechanism about the marker
We are not going to discuss removing the bottleneck in details but, you get the idea.
Systematic Approach for Bot Building
When chatbot becoming the first line of the attack, you want your chatbot becoming the best attacker right? You don’t’ want them to slack off, able to answer properly if not, react properly. With that in mind, the systematic approach for bot building needs to be executed.
It always starts with “Why?”.
Why do we want to build the chatbot in the first place? What is the purpose of the chatbot? What it is supposed to do? What is the role of chatbot? What is the KPI of the chatbot that you or your builders have on that chatbot? How do you measure the success of the chatbot?
It seems trivial when these are the question that you are going to answer even before you start engaging bot builders.
Otherwise, there will be no room for improvement, you won’t know either the chatbot is actually helping and increasing your metrics or sale.
There will be no plan or action item to improve the chatbot.
The chatbot that we build has a clear purpose.
To qualify and segment the lead.
Anything else is a bonus. Why? Because there is a time where the user decides to purchase with our chatbot. Although that is not the intention in the first place, we love a sale. I know you do. Otherwise, you won’t be reading this.
From there, we decide what will be the scope of chatbot. The scope is important because there are a lot of fancy stuff that chatbot can do such as integration with CRM to track the lead and even to track the most profitable ads copy.
The scope is important to be established because you want chatbot to do exactly what you need to do first before you start adding sugar and spices.
Then we discuss about the KPI of the chatbot. How exactly we know the chatbot that we build actually met the KPI or even exceed the KPI?
In our case, back to the purpose of our chatbot to qualify and segment the lead.
We grouped our lead into 3 categories – COLD, MEDIUM, and HOT. What we found out is when a COLD lead is passed on to the sales team, the conversion may take up from 10 days up to 3 months before they were converted to a customer. Meanwhile for medium, it usually between 3 days – 10 days.
For HOT lead, we can convert them less than 2 days.
However, as the majority of the population, we have over 56% of the lead is categories as COLD, roughly 33% is MEDIUM lead and the rest of it is HOT.
What about you? Did you have a systematic and replicable approach in bot building? If yes, we love to know more about it!
Bot Control – Internal, External and Bridge
We would like to define the chatbot control as to how you lead your user to do a certain task and bring them towards your destination (in this case, block or group of blocks)
The task can be as clicking the button, answering the question or asking them to purchase from your chatbot.
Internal control means how do you redirect them to a certain block or flow of your chatbot. In ChatFuel, you have “redirect to blocks” function that you can further control, which of the user will be redirected to the specific blocks. You can use the attribute filter to further define these.
Once internal control has been put in place, it needed to be tested further to ensure, those specific users are actually will go to the desired block. Over the past few months, we’ve encountered several mistakes that we did such as the routing of the users did not properly test that lead to the assumption of our routing going to work.
Once we go live, that routing is not actually working or accurate as it seems. Big lesson learn there.
Tips : Click on “See people matching the filter” to verify further who will be affected by your routing.
External control means how do you control who engaged with your chatbot either via ads when they first engaged or when they reengaged with your chatbot.
For example, Sam has already engaged with your chatbot and looking for an option or listing to purchase a house. Sam already told you which area he is looking for and the budget that he has in mind.
On the first engagement, Sam did not complete the questionnaire and eventually, he returns again to your chatbot and the first thing to his mind is he already told your chatbot about his preference.
As a person, when you have Sam as a lead, you pretty much have his details so when he returns to you for, you already know what he’s looking for right?
For external control mechanism, your chatbot needs to continue where Sam left otherwise, it becomes so frustrating for Sam when he needs to repeat all his preference, again. You don’t want to redo it. Same with your customers too.
Back to the systematic approach, this is where it is important to map out all of your chatbot functionalities and plan this control mechanism properly.
Bridging in chatbot sense is the chatbot connected to external or 3rd party apps via API or any other mechanism. For example you want your chatbot users to save as a contact in CRM and you need to start tracking what happen with the user.
There are literally hundreds of CRM out there but we personally preferred Hubspot. Two reasons. It’s free tier and extensive API that allows us to do our automation.
If you want to store your chatbot attribute or simple analytics, you can start store your user last block visit into Google Sheet and do the analysis.
You might seen the banner below:
Our preference in this case would be Integromat.
The visual builder really helps a lot to visualize when we start connecting 3rd party apps.
So how you can start this bridge?
In ChatFuel you have a JSON module that allows you to connect with tools like Zappier and Integromat then you start to process your instruction further inside the Zappier and Integromat.
Once you get the grasp concept of JSON and connecting it to the 3rd party app, your user experience will become better. Or.. you can let us do it for you.
Well, that is pretty much about it on how we are adopting the framework when we build our chatbot for the client. Let me recap.
- It all started with proven sales script (if your chatbot focus on sale).
- Clear workflow and responsibilities in term of supporting chatbot.
- Integration with AI (DialogFlow)
- Removing bottleneck
- A systematic approach for bot building
- Bot control mechanism
I hope that this post helps you to understand the approach that we take. There is no right or wrong approach but we believe that simplicity and doing the right thing even on the first try is the key to deliver a top-notch experience to the client and the user.
Let me know your thought on this and if you have any other method that you adopt, feel free to share with us on the comment section below.