The path to data-driven customer service

As well as providing 24/7/365 customer support, driving efficiencies and increasing first contact resolution, a chatbots’ data collection and analysis capabilities are often overlooked or not fully understood. Effectively utilising the data captured by bots can improve your bottom line through increased  revenues; ASOS for example  increased orders by 300% using Messenger chatbots, and got a 250% return on spend while reaching 3.5x more people. Let’s look a little deeper into how AI-powered chatbots can help us develop a more data driven customer service.


ASOS increased orders by 300% using Messenger chatbots, and got a 250% return on spend while reaching 3.5x more people. Facebook Business


Profile Customers

The initial stages of most chatbot interactions typically mean some degree of customer profiling. A series of automated questions can generate useful data you should catch, analyse and leverage to  gain insight for future marketing and sales activity.

Profiling data can then be used to create custom ‘groups’ of people who share similar characteristics. Our microservice, ‘Swarm’ does exactly that (think MailChimp but for bots)! Once a ‘swarm’ has been created, you can send highly targeted and messages to relevant users at relevant times. Pretty cool huh?


1–800-Flowers: reported that more than 70% of its Messenger orders derived from new customers! QuyTech


Cross Selling and Upselling

According to Medium, a customer that has made a purchase is 33% more likely to buy again. This creates a great  upsell and cross sell opportunity post purchase. Not only are you now able to purchase through chatbots, there are also a number of post purchase capabilities that can help you upsell and cross sell. For example, once a customer has purchased an item through the bot, their data will be processed and profiled  and assigned key character/personality traits. This then allows you to effectively and efficiently retarget them with products that may go hand in hand with the product they have just purchased or notify them when a similar products come to market that you deem relevant as per your upsell algorithm.

Automating post purchase notifications has never been easier and the creation and distribution of personalised campaigns to relevant buyers at relevant times can go live at any point during the marketing or sales process. So, how does this work? Let us tell you a little bit more about our 'Propensity to Purchase' algorithm …


Marriott: Use of chatbot services at the company has grown 85% month over month since the technology was launched via Facebook Messenger. Mobile Marketer


Propensity to Purchase Algorithm

 

Our Propensity to purchase algorithm is a culmination of three core indicators; sentiment analysis, trigger/ keyword analysis,and past purchase history.

 

Sentiment analysis

 

It’s important to firstly identify whether a user has the intent to purchase something that you’re offering. Do they have a positive opinion of your company? Does their behaviour suggest that they are open to being a customer of yours?

Sentiment analysis is basically the process of computationally identifying and categorising the opinions expressed by a user in order to determine whether the users attitude towards a particular topic, in this case your company or product is positive, negative, or neutral. This insight can determine when is a good time to target this user.

 

Trigger Words

 

As well as utilising natural language processing, tagging keywords is part of a chatbots’ core functionality. Say for example you are a sports club and a fan asks your chatbot ‘what time is the game?’, this signifies that the user is either interested in buying a ticket which provides upsell opportunities or already has a ticket which provides cross selling opportunities.

There are a number of avenues that you can go down in relation to tagging trigger words. It’s important to ensure that whatever a user is enquiring about, the data is effectively assessed and used to optimise sales activity.  

 

Past Purchase History

 

Traditional propensity to buy models score customers based on their past purchase history and although history is not always an accurate predictor of the future, it’s an important area to consider when it comes to the propensity to purchase algorithm.

Data is king and if your chatbot is integrated with your existing CRM system, you are able to gain a huge amount of insight into what a customers buying patterns are, what they tend to purchase, how often they purchase from you plus much, much more. Say for example a particular user has bought three pairs of red shoes over the past twelve months and you have decided to run a promotion on red shoes, the  past purchase data for this user will signify that this particular promotion would be of interest to them.

Following this identification process, an automated promotional campaign can then go out to all users deemed appropriate as per the propensity to purchase algorithm at an optimised time with a personalised message.

In overlaying these three core datasets as well as utilising predictive analytics, you are able to effectively score each user based on their chatbot interactions and past purchase history. This results in a far more insight and data driven perspective which consequently leads to a higher number of engagements and conversions.

To learn more about our analytical capabilities and how you can take a more data driven approach to customer service whilst saving money, increasing sales opportunities and improving customer experience, book a demo today.