News: How Voice of Customer Data Can Spark Brand Loyalty
Predictive analytics—the finding and analyzing of historical patterns in historical and transactional data to identify future risks and opportunities—has been used by marketers for years to determine the best possible time, place, and messages with which to target prospects.
Until recently, Voice of Customer data (customer data derived surveys, social media and other sources that tap into customer perception) has been left out of the predictive mix. Now, however, some brands have been harnessing this rich resource to better predict, not just the who, when, and where, but the “why” of customer behavior.
In this article, Cody Bakken, Director of Analytics for InMoment, answers questions about how the resulting insights can help marketers, and the special insights Voice of Customer data can add to your understanding.
Predictive analytics has been around for a while. What does the addition of customer feedback and VoC data bring?
Bakken: True, marketing departments have been using predictive models for quite some time in a variety of ways—identifying buying trends as well as items customers will purchase together. Financial services firms have been applying them for even longer, forecasting stock prices, determining insurance rates, and so on. On the other hand, predictive analytics is new and relatively untapped in customer experience.
There’s a lot that customer experience data brings to predictive inquiries that other types of data can’t—particularly when it comes to identifying root cause. Predictive analyses using just the unstructured comments from surveys can uncover a treasure trove that can tell you not only how occurrences affect customer behavior, but— and probably most importantly—why they did. We can also build predictive models off of customer scores (NPS®, overall satisfaction, level of effort) with good results, but the truly unique modeling comes from using the rich data from unstructured customer comments paired with other contextual customer data.
Using this distinctly advantageous data set, our predictions can say: “This customer visited this certain store location, rated these elements such and such a score, and mentioned this reason for their ratings—therefore, they are highly likely to return, or not return, or call your contact center for support, or spend more on their next visit.” We can steer the analyses to target just about any desire business outcome.
And on top of the predicted outcome, those same customer comments provide detailed clues on what action can be taken to avoid or optimize an impending situation, depending on the situation. We are still in the early stages, but initial indications show massive potential.
With business outcomes being the ultimate predictive goal, which types of outcomes do you view as being helpful for marketers?
Bakken: One outcome we have worked on with marketers is loyalty and likelihood to return. With one client, we have implemented several predictive models that show which customers, based on post-transaction feedback, were on the track to become loyal and which ones were most likely just one-time shoppers. If we can take our experience data that we collect and tell a client, “Hey, this customer is likely to leave you” or “This customer is likely a repeat customer” and here’s why, they can take what that prediction tells you at a high strategic level and then you can also take it a step further and recommend action to best respond to that knowledge.
Another marking application is in the area of customer personas and demographics. With predictive analytics, we can slice the data along different lines and get more targeted information. For example, we were able to look at the behaviors of one brand’s first-time customers in contrast to existing customers and learn that, while over-the-top “wow” service is certainly appreciated, it has far less impact on the loyalty of first-time customers who are much more focused on understanding whether or not this brand will fit them functionally: Does it carry the products and promote the image they want?
Beyond marketing, what other types of outcomes are businesses typically looking to predict using customer experience data?
Bakken: One thing that nearly every company has across its departments and locations is a sales plan or a conversion plan. That’s something that the customer experience data can help predict, because if you know what’s going on in each location that is having a major impact on metrics, you can understand the direction you’re heading and how to improve your odds of hitting that sales plan. We’ve done some simpler predictions as well, like forecasting Net Promoter Scores and identifying the top drivers for that metric.
Another common outcome that is easily tied to ROI is reducing defection. Predictive analytics can tell you which customers are at highest risk to leave your brand. Pair that information with lifetime customer value and you have a specific “revenue at risk” number that can be extremely helpful in driving much-needed change.
We recently did a fairly unique model for a telecom company, where we used feedback from physical retail locations to predict which customers from which stores would be calling into their support center within a specific time period based on information gleaned from the retail interaction. These findings helped the company understand exactly how to retool its training program to reduce call volume in its contact centers?a major expense.
How do most organizations use predictive analytics?
Bakken: Use varies depending on what the business is trying to accomplishments. Some organizations have sophisticated analytics departments that want assistance building predictive models, and then the raw data so they can manipulate and explore it themselves. More often, however, companies turn to an outside expert to help. In this scenario, we start with the business objective and create a preliminary model based on that. We run a small test and vet with stakeholders to ensure the model is producing data that both rings true and can illuminate the objective. Once we’re confident that the models are producing reliable actionable insights, we’ll move to the next step of visualizing the findings.
Often these findings live inside executive dashboards that allow the company to track the outcome over time in a highly visual and intuitive manner. That’s always the goal. The model will then continuously run the data to provide a real-time view into the future of that outcome. And within a dashboard, you can segment the outcome by demographic, location, or whatever else is appropriate for the organization. It’s critical to periodically refine the models to ensure continued accuracy.
What’s next for predictive analytics in the customer experience arena?
Bakken: From a technical standpoint, we are always working to integrate more types of customer data and build new models. The more streams you can bring together, the more accurate the output will be. On top of that, we’re also working on the ability to run data through multiple models at once to identify the strongest predictors.
The ultimate goal for using predictive analytics to better understand customer experience is to surface predictions across well-defined customer journeys. Right now, most of the analysis is targeted at predicting along specific touch points or segments. In time, we will be able to see the direction a customer is headed across their journey. Instead of just focusing on whether they’re going to leave or not, we’ll see that they have taken certain actions, and based on those behaviors and activities, this is where they’re heading next—for the good or bad. This will give brands new, very customer-centric ways to engage. When done well, we’ll be able to help companies intervene to stop negative experiences, nurture relationships along, and reach out when customers want to engage?all based on what customers “tell” us through their data.