Why You Need to Hire a Decision Scientist.

Late last year, I was going over some data insights with our customer experience (CX) team when we learned our agents were spending too much time in the chat tool manually answering standard questions and repeatedly collecting the same information from customers.

  • “What product would you like to discuss?”
  • “What is your account number?”
  • “What’s the phone number on the account?”

This flagged a bigger issue for us. We’d been collecting mountains of data on customer interactions but hadn’t spent any energy on automating and streamlining our customer service efforts so that our precious customer service agents could spend more time on the nuanced, complex issues rather than spend valuable time on customer information gathering.

To begin, we focused on a critical customer experience area that we felt would benefit from immediate improvement: account closures.

Sometimes a customer will want to close their account and because I was working in the financial services industry, regulations required us to ask specific questions and gather certain bits of information. That said, we also needed to account for various individual circumstances and differences.

Some organizations might have taken this as an invitation to rigidly standardize and force the user flow to match the regulations. Instead, we wanted to keep things personalized and build on natural behavior versus forcing our customers into a one-size-fits-all process. Think NLP, surveys, bots, customer data and databases. 

Data and decision science . . . you’re up.

Why You Need to Hire a Decision Scientist

Decision scientists are data scientists who specialize in combining analytics, data science and business strategy to help leaders make data-informed business decisions. Tinkering with tools and business processes might traditionally sit with technology and ops teams but the intricate knowledge of data trends and insights held by data professionals has an important place in improving business processes. This is the next logical step for organizations that want to leverage data and decision scientists to drive real impact for their organization.

Tinkering with tools and business processes might traditionally sit with technology and ops teams but the intricate knowledge of data trends and insights held by data professionals has an important place in improving business processes. This is the next logical step for data and decision scientists who want to drive real impact for their organization. 

Here’s how our startup tapped into the decision science toolkit to improve a business process, increase customer service capacity and deliver an automated (yet personalized) account closure experience for customers — all while saving hundreds of thousands of dollars worth of staff time. 

More From Chris DowsettWhat Is Selection Effect, and How Can I Avoid It?

 

How to Use Data to Streamline Your Business

Heading in, we knew the majority of account closures didn’t need to involve a customer service support person but there were a few considerations to our product that meant we couldn’t just implement a “close account” button.

Our products were in financial services, which are highly regulated. For example, customers must explicitly instruct us to close their account via email, phone or chat. Fortunately, most customers choose the chat function to reach out — which queued us up nicely for some automation.

We also have to consider the fact that the number of customers closing their account was a lot smaller than the total volume of customers using chat to ask us questions, change their account or sign up for a new product. Account closures were prevalent enough to warrant automation but not one of the top five reasons customers chat with us. This meant we  didn’t want “close account” to be one of the first options in the chat window as that would take valuable real estate for an option that wouldn’t apply to most chat conversations.

Finally, we had to account for different circumstances; people close their accounts for lots of different reasons. Some customers have a single product with us and want to cease using that product while others want to close their entire account. Meanwhile, some might have multiple products but want to close a specific combination of products. Then, of course, there are closures by way of a complaint which we treat differently under financial regulations and require an entirely different flow.

Whew. Data and tools to the rescue!

Wait . . . Back Up. What’s a Decision Scientist Again?Data Science vs. Decision Science: What’s the Difference?

 

Decision Science Use Case: Automatic Account Closures

With all these considerations in mind, we turned to data knowledge and tools to see if we could improve the account closure experience while driving some business efficiencies. 

As a data team, we were well positioned to add value here because they have a lot of nuanced information, collected across a variety of sources and tools. This means we also have insights into unique integrations between tools and how to tap into the flow of data from one tool to another (like form integrations with databases and customer experience tools, in this example).

First, we looked at all past customer service conversations in our InterCom chat tool. We mined all the language customers used in past conversations when they wanted to close a product or account (e.g. close, withdraw, shut down, etc.). Using that language, we built a natural language library of terms customers used to close their account. 

We paired that with emotion and sentiment tools to draw out negative or positive words. This is a critical step because, as I mentioned above, federal regulations require negative or complaint-driven closures to gather different information, so an account closure request with a negative comment has to move into a different closure flow.  

We then set up our closure language library and emotion tagging database to interact directly with InterCom. When a customer uses a closure word or phrase, it eventually triggered an InterCom chatbot response for account closures. Something like  “Hi, it looks like you’re trying to close your account, is that right?”

Next up, we gathered detailed information from the customer for record-keeping and legal obligations. 

At this point in the optimization, we ran into limitations with the InterCom chat bot; it only had so many functions. For example, we could only add a limited number of answer-based rules in InterCom and we couldn’t have dynamic, personalized calculations happening in the background to direct the customer conversation.

We needed a more flexible, structured data collection tool in the closure user flow. So we brought in our survey tool, Typeform, which provided a lot of flexibility in creating rules and user flows based on a customer’s answer and allowed us to use almost any input (free form, keyword, multiple choice answers and boolean) to direct customer conversations. As a result, we had the ability to make the conversation feel more personalized.

Typeform embeds and integrates directly into InterCom as well as our customer databases so it became the foundation of the closure user flow within the InterCom chat window. The Intercom-Typeform integration gave us the ability to direct the customer through the user flow and ask them for information based on their individual circumstance. We combined pre-loaded customer data in Typeform pulled from our database along with some fancy survey skip logic to deliver a dynamic, personalized experience — all without any human input on our end.

This made for a personal closure experience and ensured we were making it as easy as possible for the customer. The Typeform survey template also meant we could collect data in a structured way and auto-write all the customer answers to our databases, which became crucial for the next steps.

At the end of their closure user flow, we provided each customer with a personalized time frame within which their products would close based on their Typeform answers and a background dynamic database calculation. We also used the Typeform responses to auto-send emails or administration tickets to different customer service teams if the closure required more work. For example, customer A might need a change of bank account before closure whereas customer B’s account might need to go through a fraud check. 

As you might expect, most closures are now automatically actioned. The Typeform answers automatically trigger actions in our database and closure processes. Any comments or other bits of information we gather (such as competitive products or reasons for closure) are added to our customer feedback databases for tracking purposes.

More on Customer Success From Built In ExpertsWant to Serve Your Customers Better? Give ’Em a Nudge.

 

Should You Automate CX?

We’ve seen a lot of success by using data to improve the account closure process. 

First of all, we saved hundreds of customer service hours. We removed the need for a customer service agent to be involved in 95 percent of account closures, which freed them up to work on more complex customer service chats. This has increased the capacity of the customer service team, reduced response times and saved the business hundreds of thousands of dollars in staff time. 

We also increased our data insights around account closures by implementing a systematic way of understanding why people closed their accounts, what competitive products they chose instead all while adding more signals to our churn predictors. This has given us much more feedback for future product builds that should help reduce customer churn. Additional signals mean we’ve enhanced our ability to know if a customer is going to churn — and hopefully prevent them from doing so.

Last but not least, implementations like this have encouraged other teams to think about how data might be able to help them automate their own processes. Operations, for example, engaged the decision science team to rethink their fraud detection workflows. Finance  has reached out to the decision science team to use data to improve processes around payment of invoices and vendor work. Marketing took inspiration from this activity to move toward one-on-one marketing with folks at risk of churn. All of this adds up to efficiency gains, resource savings and the ability of our startup to scale and serve more customers.

More Career Advice for Data ScientistsUX and Data Science Teams Need to Work Together

 

How Decision Scientists Add Business Value

Regardless of the process in question, data and decision science teams needn’t stop at producing insights and data recommendations. Instead of outsourcing the use of insights to other teams, I’d encourage data professionals to take their impact a step further by reaching out to colleagues and becoming a central part of groups driving the use of their data insights.

As data professionals, we can capitalize our knowledge of various data tools in partnership with our ability to connect the dots across various data gathering tools and how a business might bring those tools together in a better way. The benefits of our knowledge and insights can be enormous — better outcomes for customers, more efficient delivery of services, more customer-driven product development and the ability to scale precious resources to attract and serve more customers.