It's been few years since I embarked on a journey to build a platform which can understand customer conversations accurately and at scale. The journey gave me the opportunity to meet various brands across insurance, retail, telecom, technology, logistics, banks, hospitality sectors and understand their challenges faced in today’s hyper-competitive consumer markets when it comes to customer satisfaction.
One positive theme common with most of the brands was understanding the importance of customer feedback to improve customer satisfaction. It's no surprise why customer ratings, reviews and Net Promoter Score (NPS) have become so popular and brands have accumulated lakes of customer feedback data.
My first obvious question to the brands during those meetings was "What are you doing with this feedback data?". Answers ranged from "nothing", "nothing but we plan to...", "we track the score", "we have an analyst to go through the feedback ", "ad hoc analysis " etc etc. Scores like NPS and ratings only tell that there is some problem but they fail to tell you What is the issue?, Why is that an issue?, Is the problem getting worse? How many customers experience those issues etc etc.
Reading and understanding customer feedback can uncover answers to these questions but its extremely difficult. One of the UK's biggest retailer collects approximately 20,000 NPS surveys per week but they struggle to systematically analyze all these surveys due to limited resources. It would take 1 analyst more than a week to read and make sense from those 20k+ surveys and then the challenges of human subjectivity and bias. Can you employ more analysts? But humans don't scale.....
"How I would tag and analyze the sentiment of our customer surveys would depend on my state of mind that day" - Analyst (leading motor insurance company)
"I can't agree on relevant tags and sentiment with my colleague, it is so subjective" - Analyst (leading UK retailer)
Due to these challenges more and more companies are realizing the importance of technology in automating the analysis of customer feedback data.
Can a machine understand customer conversations accurately and at scale? Can a machine understand the language your customers speak ? Have a look at 3 different customer feedback talking about the same topic i.e insurance certificate.
"i received insurance certificate very quickly"
"request to email policy was not followed up on until the third request."
"very easy to set up online and and get document through email straightaway"
As a consumer, we know that we use different terminology when we are going through the process of buying insurance to that we use when we are in the bank or when we are shopping for clothes. When I buy insurance I get quotes and I buy a policy and when something bad happens I make a claim, not words I use when doing other forms of shopping. When I am talking to my bank I am discussing my current account or my savings account, whether I am in credit or have an overdraft or if I want a new financial product f from them, The point is that I use the vocabulary and language that is recognized for the different engagements I have with the companies I do business with.
The added challenge here is that I also may use the sector specific language in different ways through different channels depending upon my experience. I may leave feedback if requested through the formal feedback channels using the language that the business dictates to me. However, if things do not go as planned I may leave feedback through a whole range of channels (social, email, phone call), whichever one achieves the result I want. Here I will use the language that I feel will get my point across best and the sentiment in my language might be obvious to me but not necessarily to a computer that is reading or listening to it.
This difference in language between industry sectors and in some cases between companies in industry sectors makes it very difficult for most computers to effectively interpret the language used by customers giving feedback. Which is why you cannot have a one solution fits all when it comes to interpreting your customer feedback and sentiment.
To get the most accurate view of your customer feedback it is important to have a solution that can interpret the specific language of your business, the industry language and in some cases the specific business terminology that you and your customers are likely to use regularly. An accurate interpretation of your customer feedback with insight will help to identify which parts of your customer journey are working and which ones are not.
Think about how much of your industry language is specific to your business or your sector and how this might be used to train your customer feedback analysis solution to do a better job. This is just what SentiSum does. Get in touch if you want an AI system which understands the language of your customer.