We recently hosted a webinar and demonstration to show how Intelligent Exploration and the Virtualitics AI Platform can help solve a classic business problem: customer churn. Just one of many practical applications, this example helps show how AI-guided exploration gets to the heart of the matter without requiring specialized data science techniques. Read on for an overview with video highlights, or watch the full webinar here.

Whether you’re producing physical products, or software, or providing continuing services, customer churn is probably on your radar. And the statistics on customer churn and the financial impact on businesses are plentiful:

  • A variety of sources claim that it’s anywhere between 5 to 25 times more expensive to get a new customer than it is to keep a current one.
  • It’s estimated that improving your retention rates by as little as 5% could raise profits by 25% to 95%.

Here’s the bottom line: companies that are prioritizing customer experience have been shown to be up to 60% more profitable than their counterparts that don’t. Customer experience and retention are clearly linked…but not every customer has the same needs.

That’s where your data comes into play. Your data can tell you what keeps customers coming back, and it can help you find the link between customers who leave. It just takes a little Intelligent Exploration to help you create the right customer experience for your organization.

We took a large, complex dataset, used the embedded AI algorithms to analyze it,  and visualized it in 3D to clearly illustrate customer segmentation.

7 Steps To Target Customer Churn

Rebecca Woody, our Solutions Engineer, outlined seven steps that she used to identify the source of customer churn with our customer, helping them make the leap from raw data to impactful strategies:

1. Explore the obvious drivers.

There is value in the things you already know. But we don’t want to take those things at face value, and we don’t want to assume those are the only factors impacting your churn rates. Acknowledge what you know, then be willing to look beyond that.

2. Use Smart Mapping to explore possibilities that aren’t fixed.

Look at variables that you actually have some control over, because those are the things that lead to actionable output. Smart Mapping uses AI to surface a target’s key drivers from the list of user-defined options. After identifying the obvious drivers, Rebecca removed them and used Smart Mapping to go a level deeper.

3. Extract a Network Graph to identify customer communities.

Using that same data that you explored in tables and Smart Mapping, quickly create an explorable network graph using our patented Network Extractor. While other analytic techniques look at the data in the aggregate, network graphs show every node (in this case, customer), which allows us to spot patterns.

4. Identify high-churn communities.

Look for the clusters of customers that churned the most, and then use the AI Insight tool to surface the variables that define them. Understanding what defines those communities gives us the information we need to act on.

Intelligent Exploration finds and segments our customers into communities that share key characteristics—then our AI platform allows us to focus in on those communities that also experience high churn.

5. Determine which high-churn communities are worth keeping.

You want to make sure that the customers you keep are the ones that will actually give you a return on investment. Reducing customer churn does have a lot of potential for ROI, but that still doesn’t mean there aren’t some customers who aren’t worth the investment.

6. Define a retention program targeted to resolve causes of churn for that specific community.

When you understand who you’re targeting and what it is you need to do to keep them, you can actually put a program in place. For example, if the data is showing that training and certification are critical to retention, maybe it makes sense to heavily discount that certification, or even give it away.

7. Calculate the KPIs for the program to track progress.

Proving the ROI of your data science and advanced analytics team illustrates that you are providing good value to the business, keeps momentum in place, and provides fuel for future projects.

We invite you to watch this powerful demonstration in action so that you can see just how easy it is to turn complex data into compelling, informative analyses that drive successful strategies. 

Caitlin Bigsby, Head of Product Marketing at Virtualitics, sat down with Manny Sevillano, Virtualitics’ Head of Product, and Aakash Indurkhya, Co-head of AI to discuss how organizations can move from experimenting with AI to getting actual practical use out of it. We’ve broken out this engaging and informative conversation into bite-sized chapters to make it easier for you to watch.

Video 1: Introduction

A quick overview of the current state of barriers to the practical adoption of AI.

Video 2: Why is there such a gap between the business and data science?

Data science function is meant to support the business but it continues to be so disconnected. Understanding why can help understand how to change that.

Video 3: How has the gap between the business and data science teams hindered the adoption of AI?

Closing the gap between the business and the data science functions is critical to an organization’s AI strategy.

Video 4: How can visualizations build trust in AI?

No code AI has opened up all kinds of possibilities, but it’s made the role of visualizations in the process and the consumption of AI more important than ever.

Video 5: What kind of visualizations are best suited for explaining AI?

The fact is there’s no one single visualization that explains AI, but looking at your data the way the AI does is a critical step to successful AI development and deployment.

Video 6: What’s explainable AI?

What’s the opposite of black box AI? It’s AI that tells–and shows–the story. Learn a little more about how Virtualitics’ network extractor, XAI, and multidimensional visualizations help you discover the opportunities in your data.

Video 7: How do you make everyday adoption of AI easy?

What’s the pathway to successful adoption of AI applications when people are naturally distrustful of automation? It’s a combination of design, change management, and the right technology.

Video 8: How do visualizations help guide organizations to create more practical AI?

Exploration can feel very daunting–undirected, unguided, very open-ended. Visualizations help you find your direction, and provide the common language to share your discoveries with others.

Video 9: Validating AI recommendations

Building trust in AI means providing the opportunity to validate the output. It’s about making better decisions–not following blindly.

Video 10: Pathway to practical AI

Getting AI into production has proved to be challenging–it’s not just the skills and the tech stack, although that’s part of it. It’s about choosing the right opportunity and executing it in partnership with the business.

Video 11: What’s the role of people in an AI-enabled workplace?

There are more roles for people in a world that’s really optimizing AI and automation, but it’s important to consider what those roles look like, and what those people need.

Q/A Session

Video 12: Does Virtualitics help with data preparation?

But what about the data? Preparing data is one of the core pieces of any analytics initiative. Learn how Virtualitics helps during data preparation for AI projects.

Video 13: How does Virtualitics generate these visual networks?

Network graphs are usually generated using carefully, and tediously, prepared files that dictate the relationships. Virtualitics’ patented network extractor actually finds and extracts the networks in your data for you.

Video 14: Can you remove bias from AI?

We’re committed to better AI, and part of that is keeping biases from distorting the output. We’re all hamstrung by the biases that are in our data today, but visualizing your data early in your AI process can help put you on the right track.

Many organizations have already been looking to AI to drive business growth and value — a McKinsey survey found that 63% of respondents report revenue increases from AI adoption within their company. And AI’s role is only going to become more important. In fact, it’s estimated that AI will contribute nearly $16 trillion to the global economy by 2030

However, even though businesses recognize the importance of AI, they still have trouble implementing it. Half of all AI projects fail, and of the companies that successfully launch projects, very few of them are scaling AI across the entire organization. This happens due to several reasons: a lack of trust between the business and data science teams; AI being used in one-off tests that are cut off from the business; and not enough data scientists to meet growing demand, resulting in a lack of time to develop effective AI models (and communicate these to the rest of the company). 

In our recent webinar I sat down with Manny Sevillano, the Head of Product at Virtualitics, and Aakash Indurkhya, the Co-Head of AI at Virtualitics, to tackle these challenges head-on and discussed best practices for addressing this AI gap between data scientists and business teams. Read on for five ways to help AI gain traction within your organization. 

Find a Common Language

For data scientists, it can be easy to get hung up on optimization and improving accuracy, to the point that they lose sight of how AI is going to impact the bottom line. If your organization is too focused on the discovery and exploration aspects of AI, you run the risk of leaving important stakeholders behind. This is why a road map is crucial, and why your organization’s vision for AI has to be communicated to both data scientists and the business people. 

“You need a common language for people to talk about that AI vision and the problems that they’re trying to solve,” said Aakash. 

You can also achieve this by broadening the access to the tools that only data scientists typically use — giving the analysts and business people on your team access to easy-to-use AI tools can bring everyone onto the same page. 

Make Engagement a Priority

Part of this disconnect between data and business teams stems from the simple fact that one team uses AI more than the other. “Practice makes perfect,” said Manny. “If you have a community that’s using AI on a daily basis, then you’re inherently building trust and confidence. If you go through enough feedback loops, you’re going to get meaningful value from something.” 

Because data scientists are heavily involved in the day-to-day training and implementation of AI models, they can clearly see the value of AI within the organization. But that kind of engagement is siloed — in order to get business people on board, you also need to find ways to engage them in the process from the start, so that they also understand how AI can drive business value.

Invest in Visualization

Think about how you learn…well, anything. It’s always more effective to retain information when there’s a visual aid. “Most people are highly visual thinkers,” said Aakash. “If you can’t see what’s going on, then you have no ability to really trust it.” 

To that end, visualization can go a long way in closing that gap between data scientists and business people in your organization. Visualization acts as the bridge between human and machine — especially as data gets more complex, you need something to help people understand what they’re looking at. 

For example, in Virtualitics’ AI platform, one of the visualizations you can use is a network extractor, which takes a tabular dataset (e.g., CSVs) and transforms it into a 3D network. As a result, you can see natural clusters of different aspects of your data with similarities, as well as the relationships to other clusters. The interactivity of this visualization is key, too, as you can zoom in and out, click into any node, and automatically generate explanations, which helps tremendously with building understanding and engagement.

Focus on AI Exploration 

In order to take your AI to successful production, you first need to focus on AI exploration. But exploration can be a sticking point in organizations because of how daunting and unguided it feels — there’s often so much data to sift through that it’s hard to know where to start. This opaqueness can lead to business teams downplaying the importance of AI or relegating it to one-off experiments that have no real business impact. 

AI exploration is such a critical piece,” said Manuel. “Making that much easier and more accessible with visualization means people can actually use exploration as a tool to understand every step of the incremental AI process.” 

You can think of it like going to a doctor’s office — if they try to prescribe a treatment but can’t explain why they’re doing it, then you’re not going to trust the doctor. Similarly, you’ll want to use visualization to provide much-needed context around AI exploration and help non-data teams understand how proper exploration can lead to more successful AI deployment.

Don’t Forget the Human Element 

Ultimately, people are at the core of your business, and positioning AI in a way that loops them in is the most practical path to true AI adoption. You want to make sure you don’t neglect the human element in your race to deploy AI — your teams play a crucial role in double checking the AI models and providing the strategic context necessary to ensure the AI is following your business’ overarching mission. 

“Humans are still the stakeholders,” said Aakash. “AI is being built to help humans, so we should be thinking of AI as the assistant.” 

By making sure people play an active role in the AI processes, you can create a more effective environment for deployment. You want someone who’s consuming data from the AI to be able to provide good feedback, which allows the AI to iterate and learn, and makes it more likely to be successful. 

Make Your AI Easy and Actionable

AI doesn’t have to be a black box that nobody quite trusts. By following these best practices, you can bridge the gap between your data science and business teams, making it easier to deploy AI successfully and drive impactful business value. For even more tips, be sure to watch our on-demand webinar.