Jeremy Martin | November 26, 2018

How to Improve Customer Service with Data

It shouldn’t come as a surprise that every year U.S. companies lose more than $62 billion due to poor customer service. Nor should it shock you that, after having a positive experience with a company, 77% of customers say they would recommend it to a friend. Basically, it's obvious that appreciation, prompt service, authenticity, and personalization can all help to reduce churn.

But what’s less apparent is how leveraging customer data can lead to happier customers, greater client loyalty, and bigger profits. So in this post we’ll talk about ways you can work with your data to improve customer service, starting with getting a holistic view of your customer.

 

A 360-degree View of the Customer

Most organizations use a wide range of systems such as Salesforce, NetSuite, Microsoft Dynamics, Marketo, HubSpot, Zendesk, Freshdesk, HelpScout, Shopify and more.

Yet while each of these SaaS applications fosters greater productivity, when data gets siloed within them, customer service struggles to get a 360-degree view of the customer.

Unifying data from different sources is also complex and time-consuming. Each application has its own unique structure. And 80% of an analytics project is spent on data prep and ingestion.

But regardless of what team the customer interacts with—product management, the support organization, or point of sale—their information should flow to the same central repository. So how do you do that?

Tip: Connect data sources to a data warehouse. In that data warehouse you would have data such as customer and company names, email addresses, purchase history, ticket data, orders, all of which will help you to personalize your communications, directly market to customers, and follow up with customers if there’s a problem with their order. Plus, such a central repository can help you to build a customer health score. 

 

Customer Health Score

With data unified in a warehouse, it’s now possible to take multiple dimensions of customer data metrics and classify them into a health score you can view from any business intelligence dashboard.

A customer health score consolidates customer information—product usage, account growth, time as a customer, renewals, upsells, support data, invoice history, onboarding data, etc.—to speed up, scale, and prioritize communication.

Such tactics are useful for serving the 75% of customers who believe it takes too long to reach a live agent. It can also be a means of forecasting for customer success operations. But how should you use a customer health score? 

Tip: Use a customer health score dashboard as an early warning system for your customer service agents. Green should denote that a customer is getting value from your products and services, that your engagement is effective; red that something requires immediate attention; yellow, a state in the middle. Take a look at the example below:

 

 

Account-Based Roll-Ups

For B2B organizations, account-based roll-ups provide insights by aggregating data at an account or company level. Organized at an account level, data has more meaning than when it’s scattered across dozens, or even hundreds, of individual records on leads, tickets, or activities.

Rolling up ticket and company information by account is extremely valuable because it lets you understand an entire account’s support experience with your business. You could answer questions like:

  • What is the average number of tickets per account per month?
  • What are the Company and Contact properties for each ticket profile?
  • What is your combined satisfaction rating?
  • What is the average number of tickets or conversations for a new contact

Account-based roll-ups are also beneficial for viewing data about leads by account. You could questions such as: 

  • How much have you invested in targeting a specific account?
  • How much pipeline revenue was generated by that account per campaign?
  • How many interactions were needed before the account was converted to a customer?
  • What was the medium (e.g. phone, email, chat, etc.) for these interactions?
  • What was the optimal sequence of interactions that most effectively led to the acquisition of a new customer? 

Tip: Account based marketing (ABM) is a targeting strategy where companies market to accounts (different people at one company) versus individual users. We recommend you gather data from many sources—CRMs, ERPs, marketing automation systems—to personalize your inbound marketing content and send relevant, helpful information rather than transactional, spammy sales messages. With accurate contact data about the buying personas at your target accounts, you can better execute your ABM strategy. 

 

Churn & Retention Metrics

According to the Harvard Business Review, acquiring a customer is 5 to 25 times more expensive than keeping a current one. Given this fact, every organization should focus on keeping churn as low as possible. 

Luckily, if you use a CRM to collect data about contacts and actively track and segment contact variables, you can identify the most valuable customers over time and increase customer loyalty by providing customized products and services. 

When used in tandem with CRM data, finance and product usage data can then help you gauge a customer’s health and intervene when surfacing signs of distress (see customer health score).

Tip: Regularly measure the number of customers you lose (and gain) over time. But you should also track the reasons why customers are leaving, whether the churn was involuntary (e.g., a customer goes out of business or gets acquired) or voluntary (e.g., customer dislikes the product and chooses to cancel or picks a competitor). 

 

Online Chat & Conversation History Data

If online chat is the digital equivalent of a salesperson, then the conversation between the customer and sales agent is a valuable trove of data. You can quickly discover why certain customers need assistance, the best ways to close a sale, and reply fast enough to not lose business to your competitors. Like email, live chat lets you answer questions. Unlike email, however, live chat lets you engage customers while they’re still on your site and ready to make a decision—not hours after they’ve left.

Besides supporting your customers in real-time, support apps can also give you a conversation history of previous chats. And many integrate with CRMs like Salesforce, eCommerce apps like Shopify, and of course with Zendesk's helpdesk. You can see who has sent a message, clicked the chat button (but hasn’t sent a message yet), activated a trigger, browsed your site, or gotten stuck at the checkout page. From here, you can initiate chats with customers or view detailed information about a specific customer. Speed things up with shortcut canned messages, or automatically start a chat with a customer when they hit a specific page.

Tip: To maintain a consistent experience for your customers, your CRM should talk to your help desk (and vice versa) through an API integration. This contextual data helps your sales team make better recommendations for current and potential customers. For example, the Help Scout + HubSpot integration automatically adds Conversation activity to your HubSpot timeline, so your support and sales can stay aligned without adding a step to their process. We recommend integrating your chat support data with your CRM to create a more holistic customer experience.

 

Automated Monitoring & Conversation Scoring

Your customer service reps should use the right language, with a thorough knowledge about the product and how to speak to the customer for a given situation. Now, you could use data such as average call and hold time to score conversations, but automated monitoring and scoring calls, chats, and emails offer a more accurate assessment.

Tip: implement a platform that can transcribe content and turn it into searchable, easily categorized text and structured data. Such a platform will let you flag calls that contain desired or undesired language, combined with other measurements such as silence, agitation, or transactional measures. It will also help you to correct support agent behaviors before they become habits, provide positive feedback, train new reps, keep top performers, talk customers through difficult situations, and cater to agents’ strengths.

 

Sentiment Analysis

But what if you want to study the conversations themselves to extract trends and best practices? For this, there’s Sentiment Analysis. Sometimes referred to as Opinion Mining, Sentiment Analysis is a field within Natural Language Processing (NLP) that seeks to identify and extract opinions within a body of text.

Such texts might consist of emails, case notes, sales records, social media posts, or support tickets. Whatever your “corpus”, you would be trying to analyze customer conversations to garner feedback data recorded in your helpdesk, CRM, or sales system. 

By picking out keywords, you can then assess trends in what your customers think about your products, services, brands, and any other data that might help to identify trends in your customer service data. 

Sentiment analysis can also deliver high levels of granularity about either the satisfaction or frustration your customers feel based on the language being used. In other words, you could slice and dice the data into categories such as very positive, positive, neutral, negative, and very negative so that the analysis works much like a net promoter score (NPS) or customer health score.

Tip: Based on what you learn from your findings, you might then decide to train your sales team differently, or modify the script for your support agents. Perhaps there are very important pieces of product feedback that you could pass along to the PM, who can resolve those issues so as to indirectly lighten your support requests and reduce churn for the business.

The catch, as you might have guessed, is that sentiment analysis requires a unified data source, which you can get by connecting SaaS applications such as Zendesk, Freshdesk, Jira, and Salesforce to tools like Fusion.

When all your customer data are structured in a cloud data warehouse, you can then feed that data to train machine learning and statistical algorithms like Naïve Bayes, Logistic Regression, Support Vector Machines, or Neural Networks.

Now this may seem like a lot of work. But it’s actually less work than a manual audit of web chats, emails, phone calls, and social media threads. And in light of the degree of subjectivity involved, sentiment analysis is still 70-80% accurate by most standards.

 

Chatbots

Chatbots are a touchy subject. Some worry they might replace human support operatives and customer satisfaction reps entirely. Others think they sound too rudimentary and foster miscommunication. 

Whether you love them or hate them, chatbots are here to stay. Thanks to dramatic advancements in machine learning and natural language processing, today’s chatbots are smarter, more responsive, and more useful than ever.

Still, in order to make them work properly you need more than just a machine learning layer. You need the data store, too. That data store is what allows you to teach the machines the difference between “good” and “bad” words, since bots only know what they’re programmed to understand. 

 

With the right data store, it’s then possible to improve your customer experience via a chatbot. This is very important for short-staffed companies. Bots can free up overtaxed agents and provide 24/7 support, answer questions in under 2 seconds, resolve customer disputes, and prevent escalations. They can analyze logs of previous conversations, keeping the customer from waiting until a support rep is available or having to repeat themselves after being rerouted.

Tip: If used properly, chatbots can be incredibly useful. However, analytics can only teach chatbots so much. You still need data to be interpreted by real people who can spot miscommunications, uninformed support agents, and whether customers’ expectations were met given the context of questions, industry patterns, and customer reactions. As always, don’t be so overly focused on automation and data as to eclipse what’s important. Otherwise agents will be too focused on closing tickets instead of learning the nuances of their customer conversations. Remember that humans are still much better at speaking with one another than computers can utilize natural language processing (NLP) to convert natural language into machine language in order to find its meaning. Support agents are still better equipped at providing expertise for their ever-changing products. Use automation to save support time, such as suggesting to whom to route specific tickets, or which article may be useful, given the topic of the support ticket, for answering the customer’s inquiry.

Want to use customer data to improve your customer service? Try Fusion for free today.

Subscribe Here!