arrow flex-icongithublightnings linkedinlock target triangle-icon twitter

Marks & Spencer: Natural Language Routing

Louisa
August 29, 2018

Challenge

The modern day call centre has taken some significant technological advances over the past few years and multinational retailer Marks & Spencer want to stay ahead of the curve.

Whilst IVR used to be limited to DTMF, recent advancements in natural language processing technology mean that businesses can offer faster and simpler ways to engage customers.

To ensure they can offer best-in-class customer experiences, M&S asked DVELP to help replace a legacy phone system which routes 11 million calls to stores and contact centre per annum, with an AI-powered solution to improve the accuracy, efficiency and scalability of routing.

The build had to maintain high-availability and uptime during open hours and an MVP had to be ready for production deployment in 30 days.

Solution

DVELP built a flexible and scalable solution using technology from Google, Twilio and Looker. The solution, driven by AI-powered interactions and intelligent routing, allows M&S to be more responsive to customer needs and gives unprecedented visibility to the business.

The application converts speech to text using Twilio Programmable Voice and Google Speech-to-Text. The phrases or ‘utterances’ from the customer are interpreted as intents by DialogFlow and used to route calls to the correct department or present self-serve opportunities to the customer.

DialogFlow, an extremely powerful tool for building conversational customer interactions sits at the heart of the application. The product allows great flexibility and it was crucial this was maintained. The application and DialogFlow remained loosely coupled in order to reduce the engineering time required to build out new flows or interactions. This was achieved by creating a dictionary of actions, such as ‘continue’, or ‘hangup’ that could be added to intents for the application to consume. The approach allows the team to build out rich flows directly through the DialogFlow UI, with little or no engineering time.

The data (inbound/outbound numbers, time/date, responses from DialogFlow, routing endpoints, etc.) is fed into Looker, which gives rich real-time data visualisation and helps M&S be proactive about trending customer concerns and provide insights into key metrics that were previously obscured by legacy systems. The introduction of real-time reporting allowed M&S to remove previously cumbersome and inaccurate processes, such as Reason For Calling (RFC) codes. The increased visibility not only reduced average call handling time, but gave new insight into customer behaviour.

Result

The application and associated use cases continue to evolve and grow, but following the initial period of production deployment it’s clear that the application delivered huge value on a number of planes:

  • Customer experience
  • Operational efficiency
  • A springboard for iteration

Customer Experience

The solution has lead to increased accuracy and efficiency of customer routing. It was and continues to be a firm belief that self-serve opportunities play a vital role in improving the service, but also that often a person is best suited to speak with a customer directly. A primary goal of the programme was to clearly differentiate these interactions and provide a platform with the flexibility to deliver on them.

In the months since deployment, we have seen a 95% engagement rate from customers and have increased routing accuracy by over 70% in comparison to the legacy DTMF IVR.

Operational Efficiency

Inaccurate routing, limited self-serve opportunities and lack of visibility typically culminate in the requirement for expensive and inefficient for human intervention. The legacy landscape of this project was no different. The valuable time of M&S employees was being consumed by manning phones to re-route or complete menial tasks in the wrap up of calls.

The net effect of DVELP’s solution is that more time is spent on interactions with a higher value both to customers and the business.

A Springboard for Iteration

Unlike many traditional telephony implementations, where obsolescence begins on the day of launch this project is at the beginning of it’s journey. With only a handful of self-serve opportunities in place and data analysis in its infancy, it’s clear there is exponential value to be added to the customer experience.

The stats

  • 11 million calls a year
  • 70% increased routing accuracy over rigid DTMF IVR
  • 95% utterances translated into actionable intents
  • 95% caller response/engagement
  • 10 seconds saved in Contact Centre Average Handling Time due to automated ‘Reason For Call’

Articles

By using this website you agree to our cookie policy
x