Published in General

Reflections of A Practitioner: CX Maturity, Customer Expectations and the Use of Predictive Analytics

Following my four years leading the CX practice for, arguably, the best airline in the Americas, JetBlue, the transition to the “provider” side of things since last May has been very interesting. My experience has provided a unique perspective into the various levels of CX maturity in organizations across many different industries. It has also provided insight into the varying levels of investment, executive support, engaged cultures, talent, strategy, and, at the most basic level, the knowledge of what to do next for many programs. Although the variance in maturity is large, the collective recognition of the need for a great CX program seems to be widespread. But the paths to having excellent customer experience are as varied as maturity levels across companies.

In an effort to help our customers (and we at MaritzCX) better understand and articulate just where organizations reside along the CX maturity spectrum, and to provide objective guidance what to do next, we developed our CX maturity model CXEvolution. CXEvolution provides organizations an assessment to help them understand the progress in their CX efforts, and how they compare to peers and industry leaders. Since its creation, a significant number of CX professionals have taken the assessment – both from current clients organizations and from companies we do not currently work with.

The results so far have been fascinating. In addition to providing customers and organizations the information they are seeking, we have also been able to see a clear positive correlation between CX maturity, increased customer retention, and CX ROI. In fact, organizations at the highest rungs of maturity demonstrate customer retention and financial results of CX efforts nearly three times as large as those with less advanced programs. This alone has been a great story for practitioners to articulate as they seek investment for deeper CX initiatives.

Naturally, as we reviewed CXEvolution results with our customers and interested organizations, the “next step” needed was different for each. Some organizations struggled with executive leadership support, whilst others were very new to organized data structures and architecture. Having said that, we continued to find that CX practitioners, no matter where on the maturity spectrum, ultimately take action in two areas. I’m not talking about the “action” of creating a survey or performing text analytics – those are foundational elements that enable action. I’m talking about action that can directly impact customers and their experience.

The first real “action” concerns recovering or reactivating customers who have had a negative experience with a company or brand, or perhaps have had an experience creating the opportunity for “promoter activation.” Even giving customers the motivation to be part of a positive cross-sell or additional interactions with the brand can be part of service recovery. This action is more commonly known as having a closed-loop process within your CX program.

The second area of action is much more broad and, quite candidly, much harder. It is focused around designing and continually improving the customer experience. This action area is much more difficult than the first, because it goes far beyond the CX team and program. It is more than collecting data and sending it to appropriate representatives who interact with the customer; rather, it requires participation, support and engagement from adjacent lines of business and leaders. This is the part of the CX management job that requires much more art than science, and much more EQ than IQ. You as a CX practitioner need to be a great leader. You have to be influential within the organization. You have to know how to navigate the internal bureaucracy – and yes, it exists within every organization, more some than others. You need to build a level of trust and influence that will get all levels of the company involved in improving the customer experience. You have to change the company culture.

The differences between these two action areas are interesting as well. The first is much more “bottom-up” CX, where business rules guide actions for the frontline employees to engage or recover customers. The second is more “top down,” with some elements of bottom-up CX as you deploy localized action planning, etc. The first can be done fairly independently within the CX team, with some adjacency to the resources who are “closing the loop” with the customer, whether they be centralized in a service desk or distributed across the various lines of business. The second, however, requires much more from the organization – mainly engagement and support from leadership at all levels to enable much of the other CX attributes of maturity. Consider creating and cascading CX goals. Without leadership engagement, motivation will be obligatory and unnatural. This area of action is simply harder – and this is where CXEvolution helps so well. Is one more important or effective than another? Truthfully, they are both needed. You’ll find success in both directions, and having both in parallel makes each even more impactful.

Considering the comparison one more time, and this time from a data perspective, each of these “action” areas has different data needs. For bottom-up CX efforts, the greater the catchment of customer feedback, the better the opportunity to identify the most important individuals to engage. For the second, in contrast, more volume isn’t necessarily needed as much as the right volume of feedback volume and appropriate representation of various customer segments to ensure proper diagnostics, drivers, and areas of focus that can be weaved into strategic investments and organizational goal setting. Further, the very type of feedback collected for each can differ. The first, or bottom-up, efforts do not necessarily need all of the quantitative diagnostics. A simple understanding of how the customer feels as result of an experience, and whether or not that will motivate behaviors is enough to take action. Even more, these two data elements blended with CLTV (lifetime value) will expand the opportunity greatly. The second area of action requires much more rigor and appropriate survey structure to enable a proper diagnostic and representation of customers, drivers, etc.

Herein lies the challenge. As experiences continue to become more digitized and faster, customers are less and less motivated to spend ancillary time beyond the experience your products and services provide to give feedback. Don’t get me wrong, every brand has advocates that will play and some companies have many more advocates than others, but we’re continuing to see greater and greater response bias AND declining response rates in whole. In addition to changing customer expectations and discretionary time, the sheer barrage of emails and solicitations that we all receive from marketing campaigns, subscribed notifications, etc. have created significant customer fatigue.

In response, CX teams are doing the right things to mitigate both challenges. Surveys are becoming more touch-point based, more optimized for mobile, more embedded within the experience and even more conversational, like a simple SMS conversation with a friend. Even then, however, taking a survey for many has become just too much time in an on-demand consumer environment. Anecdotally, as I reflect on my own experiences, I can relate too. I’m a self-diagnosed Amazon addict, and a CX practitioner at the same time. Do I want to take a survey for them? Not really (well, maybe just to see what questions they are asking). My motivation to NOT answer their survey is partly because nothing went wrong in my purchase and delivery (there’s my bias), and secondly, If I do take the survey, then it feels like I’ve reduced the time-to-value that Amazon has so elegantly created for me. The survey itself may literally take longer than my search-to-buy experience entirely.

As mentioned, the efforts CX teams have employed to mitigate these challenges, at least for top-down efforts and more diagnostic based surveys, have been useful. Unfortunately, for bottom-up efforts, practitioners are limited to those who provide feedback. Reinforcing this point, I can recall my former CIO saying, “until we can anticipate customer sentiment or behavior, we’ll be subject and only focused (at an individual level) to those who raise their hand,” or more specifically, we’ll be in reactive mode. This thought resonated with me for quite some time, and eventually turned into a concept, partnership and product that I wish I had for myself back at JetBlue.


Here at MaritzCX, we talked about and dabbled with the idea of addressing this need for a fair amount of time; first by talking with customers, and eventually through performing tests with real customer data. Our concept wasn’t completely new by any means, but it certainly was one that hasn’t taken firm traction in CX. The goal in our mind was to leverage advanced statistical methods to enable our customers to predict customer sentiment, and ultimately behavior from that sentiment, at the individual customer level, beyond the respondent population to provide better visibility and engagement opportunities. We wanted to enable our client companies to understand, in real time, which of their customers needs attention, when they need attention, and what kind of attention they needed. If we could do this, we knew we could take the first area of action well beyond what it is today. We could enable proactive CX for our customers, and give them an avenue for CX program ROI well beyond many of the activities they are doing today.


Along with the concept comes the reality of actually making it work. It needed to be easy for our clients to engage without red tape, additional internal IT resources, involvement from supply chain, etc. It needed to carry SaaS-based flexibility as they have come to expect. At the same time, our customers and research also indicated that even if it were simple to activate without peripheral involvement, they would need guidance on not only HOW to do it, but WHAT to do at the same time. In short, we needed the solution to be simple in process and robust in scale for even the most complex needs, in both technology and supporting resources. Last, and certainly not least, we deliberated over the right technology to meet these needs. After much discussion, we decided the best direction for MaritzCX and our customers would be to integrate with a reputable technology and analytics partner who make this work their sole focus and proposition. We did our research and decided to partner up with PurePredictive, Inc., of Sandy, Utah. PurePredictive met all of our needs; they are completely SaaS based, have developed patented technology to deliver the most accurate predictive models, and delivers architecture that meets our security standards and demand for incredible scale. Needless to say, the concept and partnership have combined to create a pretty cool solution.


We know from our customers that they don’t have a data (volume) problem. There is no shortage of data. Unfortunately, for those who do have data challenges, it is due to disconnected data sources and disparate systems. One of the greatest peripheral benefits of putting together and growing a CX program is that you have to address this disconnected data problem in some fashion. As result, we have some of the richest datasets around by the time we’re ready to solicit feedback or join to unsolicited feedback sources.

The opportunity that arises from well-connected data is quite significant, specifically for bottom up CX efforts. Unfortunately, most of this connected data (which typically comprises of transactional, operational, relational, profile, demographic, and purchase data) is rarely leveraged in closed-loop practice. At least until now. With PredictionCX, we’ve successfully built highly-accurate models to predict customer sentiment by identifying the “CX recipes” from within this “big data” that produce behaviors, both desired and undesired. The result is pretty incredible. While most programs collect feedback from 5-10% of their customers, PredictionCX creates an environment where you can understand how virtually 100% of customers feel about you, or will behave as a result.

Using this concept come to life, the predictive models and customer-level predictions we’ve performed vary from organization to organization. Most often, customer recovery and focus on “detractors” is the first effort, which is no small opportunity in the majority of cases. The more-mature organizations are looking to predict specific behaviors around product cross-sell timing, propensity for very specific product use and behaviors, etc. As I mentioned, however, simply focusing on churn can yield incredible opportunities, even if you are simply changing the mix of customers who will make their way into the closed loop recovery practice.


That’s just the beginning, really. First and foremost, if we’ve accurately predicted a transactional or relational aggregate measure such as Net Promoter Score® for your customers, you’ll get a pretty good sense of your response bias between the total customer population and your respondent population – assuming this hasn’t already been done using customer-segment weighting. As a side note to this first result, it is VERY important to validate currently-used metrics (like NPS® or any other measure) to behaviors (like churn) through a proper linkage analysis or similar exercise. Just because you’ve predicted promoters and detractors, does that actually mean you’ve also predicted what they will do? This has always been a long debate amongst practitioners and I’ll tell you the linkage (or lack of linkage) has been pretty surprising. In some cases, Net Promoter and other KPIs serve very well as proxies to predict customer behavior. In other cases, it has been shown as a fairly useless measure. When it doesn’t serve well as proxy, we find a better measure or move directly to the behavior that needs prediction.

Secondly, and for the real benefit of PredictionCX, you’ll get a much bigger view of opportunities that exist. In a recent exercise we performed for a customer, we partnered to predict likelihood to churn. This prediction at the individual customer level, aligned with customer value segments, quickly identified nearly $70M of more easily-realized churn recovery. This wasn’t the total recovery opportunity, but rather a very specific segment of customers who demonstrated the greatest profitability, but unfortunately, the greatest propensity to churn. As well, the propensity to churn had very strong ties back to the experiences and products they consumed, providing a two-fold benefit; a very specific recovery opportunity, and insight/guidance on points of pain for this group of valuable customers.


As far as actually executing recovery, or taking action, this is done in two ways. The first is through more common one-to-one case management/closed loop process. The only difference is that we’re now funneling the most critical customers into the queue in addition to any who have provided feedback in good faith that it will not be tossed into the abyss. The second is slightly less common, but becoming more frequently used and considered important to email marketers. Rather than one-on-one, high-touch engagements through case management, we’re integrated with existing CRM and marketing platforms for outbound engagements following an experience or prediction that needs intervention. Even cooler, although these engagements are bulk in nature, they can also leverage the prediction results for individual personalization as our PredictionCX results includes the very recipe that generated that prediction. And as ancillary benefit, your email marketing team will have visibility and insight into customers who need to be temporarily suppressed for marketing message(s) given their sentiment or experience. Gosh, how badly I wished I had this while at JetBlue!

In the end, we’re very excited to deliver this new solution to our customers and more. It helps to further address the needs our customers have been communicating and continues to expand our unique and unmatched combination of industry leading CX SaaS technology and deep research, statistical and CX expertise. Here’s to your CX, and here’s to the ROI it will generate!


To learn more about PredictionCX, click here, and keep your eyes on the CX Café for further articles from great CX pros.