October 27, 2016 | Written by: Scott Stockwell
Categorized: Blog | Marketing | Retail
Share this post:
Spencer Morris, senior data scientist at InMoment. InMoment have an ‘experience hub’ that champions the voice of the customer. The company handles 2750,000 customer experiences every day. Forrester have reported that ‘improving the customer experience is the surest way to win, serve and retain customers’.
Spencer Morris, SVP Data Insight at InMoment explains the power of cognitive for customer insight
Customers’ experience now defines brands
The perception customers have of the culmination of their experiences over time with all aspects of your brand. It could be in-store or on-line. Through a mobile app or a conversation with a friend or co-worker. Brands are no longer control of the brand experience – that’s now with customers. Gartner declared that “customer experience is the new competitive battlefield”. It’s 5X more expensive to recruit a customer than retain one.
The power of feedback data
Performing data-analysis on unstructured data such as phone-call surveys has been very challenging in the past. Much insight would have been available but was not possible. Satisfaction scores are all well and good, but they don’t give direction on what could work better. Mining this from unstructured survey such as open-ended feedback form questions has great potential.
InMoment partnered with IBM and was the first company to use Watson’s enterprise level cognitive capabilities. The company used Watson to build text-analytics models for all sorts of industries using the text feedback and unstructured feedback that the companies gathered.
Finding the sentiment in feedback
In a comment such as “The teller Diane, checked me deposit slip for my name, but I like it when I’m called by my first name, everything at the drive up went really quickly.”
This gives six tags for analysis:
- speed of service
- banking supplies
- mentions name
The statements are also broken down into sentiment level phrases. Is the customer happy, sad or neutral? And what level of sentiment is this – strong, medium, weak etc. Being able to do this in real-time can provide great insight into the customer experience.
In each industry, the talk is different. In different regions, the language is different. The same word in different sentences can have different meanings. Being able to identify the true intent of customers’ words is key to defining sentiment accurately. Removing text incorrectly categorised is also important.
A feedback conversation – not a satisfaction number
Text analytics, emotion analysis, tagging – in a tool? Rather than providing a text box for a customer to provide feedback. Giving the customer a slider to show how satisfied they are starts off the conversation. Natural language response to the first piece of feedback elicits greater depth of feedback from the customer. Feedback becomes a conversation and not a satisfaction number. As the customer provides more feedback, the tool can use techniques such as multiple choice and prompted questions to get to the real customer feedback.
InMonent have build these feedback conversations for multiple industries. They’ve seen feedback increase 33%. Word count and length has gone up 40%. With more feedback, brands have more data to mine insight from.
A data analyst available 24-7-365
Using Watson, InMoment are now able to include predictive analytics into a feedback conversation. They bring structured and un-structured data together and run text analytics on it. On the unstructured data they run text analytics, sentiment analytics and categorization. They bring the two together using machine learning to have a predictive conversation.
- Churn – greater insight can give more confident indication of customers who will churn.
- Sales goals – brands can make savings retaining customers as it’s 5X more expensive to recruit than retain
- Problem resolution – cognitive capabilities can now identify customers with problems before the call the contact centre enabling brands to reach out proactively and save costs using more effective channels – earlier.
Adding in the context of location, time, weather and other data adds new levels for insight. For example, are negative sentiments confined to specific regions requiring remedial intervention. Are they due to a particular touch-point such as coupon redemption, or speed of drive-thru response? As the analysis is done in real-time, alerts are now possible as soon as trends are identified.
In summary, being able to analyse structured and un-structured data in real-time can enable brands to have feedback conversations with customers. Doing this in real-time with the context of other data such as geography, time and weather provides greater context. Lastly, creating alerts can allow brands to act on issues before customers churn, saving money and increasing brand value.
Spencer Morris, SVP Data Insight, InMoment