RevLocal Uses Machine Learning and Predictive Analysis With Qlik To Drive Client Retention

• Read original article here

Formed in 2010 and headquartered in Granville, Ohio, RevLocal is a personalized digital marketing company for the small and midsize business (SMB) sector. With a focus on local search, review marketing, paid advertising and social media, it supports 7,500 clients from 60 different markets across the USA.

Need to capitalize on historical data

The SMB arena offers many lucrative opportunities for digital marketing so competition is intense and increasing year-on-year. To stand out from the crowd, RevLocal offers a service based on relationships with strategists who not only provide solutions but also help customers make best use of the tools to promote their business.

Despite this ‘white glove’ approach, the rate of customer churn is a constant challenge as Chief Data Officer, Ben Dean, explains: “Each month we would experience 4% to 5% customer churn. We were really good at bringing in new customers but not so good at retaining them because we had made the decision not to force contracts on our client base.”

RevLocal had a wealth of historical operational data and realized that it could use this to identify customers who were likely to leave. It could then change its approach to them and persuade them to stay.

Machine learning (ML) and predictive analytics were the answer and, after researching the market, RevLocal chose Qlik’s automated machine learning tool, AutoML.

“We wanted a solution that was relatively simple to use for people who had not had a lot of experience developing machine learning models,” explains Dean. “We needed training and ongoing support and that was what stood out with Qlik AutoML.”

RevLocal’s first machine learning project was to create a customer retention model. Datasets are fed into Qlik AutoML which highlights risk areas by identifying patterns and trends in customer behavior. To translate these predictions into business value, digital tasks are generated for marketing strategists to improve client relationships.

An initial test saw three months of historical data on just 331 clients fed into the model. It brought a 7% rise in retention rates.

“We see this promising result having an important effect because the numbers we are dealing with are so large,” adds Dean. “With over 7,500 clients, even if we just retain 5% or 6% it represents significant long-term value for the company.”

After eight months using the model, the business intelligence (BI) team has now input a year’s worth of historical data for 2,000 clients, amounting to four million rows and 30 columns.

“Overall, our retention averages are trending upward from where they were before using Qlik AutoML,” adds Dean. “We see behaviors changing, awareness increasing and a better customer experience for our newer clients.”

RevLocal has gained in-the-moment awareness which supports better decision-making as Senior Manager of Business Intelligence, Tiffany Ware explains: “This is helping us to understand our clients better. We used to have to make educated assumptions, but this gives us the opportunity to validate and explore areas that we may not currently have on our radar.”

This greater awareness helped to provide support and greater confidence in the company’s decision to move forward with six- and 12-month agreements. By retaining them for six to 12 months, they will have time to see value in what the company does and remain.

Having started with client retention, RevLocal is now using a similar dataset to create a client upsell model and plans to follow this with an employee retention model. The longer-term vision is to use machine learning to predict which prospects in the sales pipeline are more likely to become clients and spend more.

Dean concludes: “We were stumbling blind, trying to figure out what would stop churn and it was frustrating trying to find something that would move the needle. Now, machine learning has really given us direct clarity as to what will make a difference. Instead of guessing, we now know what will bring results.”