Customer analytics for CX: Protect, innovate, serve – OR PERISH

0
305 views

Share on LinkedIn

Customer analytics – we’ve wired em like lab rats – now what?

If you’re starting a CX career, trust me, you’re lucky.  You have mountains of data, fast machines, channels galore, advanced algorithms, and amazing customer analytics to work with.  But this wasn’t always the case.  In days gone by, data was harder to come by, expensive, and time consuming to process.  And though statistical modeling was around, customer analytics hadn’t come of age.

Today, customers are increasingly willing to trade their privacy for vanity, points, and free apps.  As such, whether they realize it or not, they are surveilled, surveyed, tracked, and analyzed to death.  In the process, they relinquish device data, location, digital behavior footprints, social sentiment, and transaction history.  Because of stiff competition among cloud providers, the cost per byte of customer data stored and processed has dropped precipitously.  The cost to store 1 GB of data is as low as 3 cents[i].  Processing it is much cheaper too.  So, don’t complain there isn’t enough data and tooling to crunch it all.

The least we can do as CX professionals entrusted with this data is to protect it, use it judiciously, and with it endeavor to return value-added services and a better customer experience:

  • When a consumer surrenders their location, reward them with great nearby offers
  • Pay attention when consumers shop online and why they went there
  • Recognize important journey moments and provide value.  The only way to do this effectively is to master customer analytics

Without the right data at the right time, you won’t know Jack

And your job is to know Jack very well – and Jill too; and every other customer you serve.  The prerequisite to any effective customer analytics is the right data.  Yet wrangling and operationalizing data so it’s fit for purpose is a daunting assignment.  A huge mistake is embarking on a mega project to build a central data warehouse that likely becomes a jack of many data trades and a master of none.

It sounds like a reasonable goal to build a 360-customer view.  But the trouble is this project invariably holds up otherwise well-conceived and well-scoped customer analytics efforts.  Further, once the 360 view is finally ready, the data is often stale and still requires data wrangling and redesign for a specific purpose.

Customer data will always be spread out.  The recently announced open data initiative (ODI)[ii] by Adobe, Microsoft, and SAP (three of the biggest companies on earth working together on this problem) is further confirmation of that, and that:

  • There is no single superior solution available now
  • There are many data sources that house valuable data, and we’ve seen rationalization of these sources in the last 20 years, and critical masses of this data now sit with a handful of big companies

However, beware because:

  • The announced effort is still theoretical and unproven
  • The initiative is not oriented around very specific outcomes and decisions that will be made by the consuming applications
  • Although there’s value in data sharing, a practical monolithic data model may be an oxymoron
  • Rarely do any so called “Open Initiatives” result in something that is truly open, free, and well-protected

End users of an experience platform (like Adobe), a CRM platform (like Microsoft), or a big data platform (like SAP) will likely find the open data model either too nebulous or incomplete, and thus of limited value.  Each platform has dozens of applications, and each application has different data and performance requirements.

That said, there is no denying that data equals power.  Consequently, consider two simple questions:

  1. Are you fundamentally a data controller or a data processor?
  2. If you’re a data processor, how will you get access to key data?

Data controllers are responsible for collecting, curating, and distributing data (if they chose to do so).  Since they own data they control its rental price.  Conversely, data processors rely on a data supply like a vehicle needs fuel.  Data processing services are subject to serious competition, price pressure, commoditization, and potential replacement.  In other words, they aren’t sticky unless they’re generating deep insights that continuously prove worth.

The lesson here is that those who own the data will have the power to set prices and control who gets access to it.  Thus, plan your strategy around the data you need, and carefully protect the data you control.  In most cases, first party data is the most valuable data.  For that reason, guard it, curate it, and distribute it with your customers’ best interests in mind.

Why customer analytics matter

Math, science, data, statistics: sounds messy, complex, and doesn’t exactly conjure up images of simple, elegant, and enjoyable experiences.  A snippet of python code; multi-variate testing; Monte Carlo simulations; all are intimidating and cause eyes to glaze over and mere mortals to tremble.  Many avoid these like the plague or tune them out entirely, lest they trigger recurring nightmares (like one where you’ve failed to study for a looming high school math test).

But avoiding them isn’t advisable.  Your very survival may depend on gathering data, protecting it, and automating customer analytics.  Its algorithms that extract insights from data are the cornerstones to differentiation and delivering great customer experience.

Why getting customer analytics right matters so much was extensively explored in a 2016 McKinsey study[iii].  It showed, for instance, that firms extensively using customer analytics outperformed competitors in their market by 115%.  Summed up in a few words: predict customers’ behavior and use those scores as triggers and ranking mechanisms to provide relevancy and value.  Do this and you’ll drive up revenue, NPS, and ROI.  It simply boils down to that.

When you use data and math to understand customers, you’re doing it to better predict what customers want (their intent), why they want it, and when they need it.  The more you guess right and automate, the more they’ll derive value from interacting with your business.  And though clever content, UIs, copy, and packaging matter, what you present, when you present it, and how you fine-tune that presentation needs to be based on match, science, customer analytics and data.   Check out this article I posted earlier this year: “AI-based Promotions: Welcome to the Creative Machine,” for more on how the lines between creativity and machinery are blurring for marketers and CX professionals.

Gut calls are for cave dwellers

A small business owner that delivers a kid glove customer experience is special.  Remembering names, birthdays, children, past purchases and product preferences, key life events, and so forth, matters immensely.  Conversing with customers and recalling what they need and why is invaluable information for service-oriented businesspeople.  But simply put, human memory can’t scale in any modern business environment.  Try remembering 500 things for a single customer and it’s obvious why you need machines to help.

But the challenges don’t stop there.  Now try making meaningful individual decisions and recommendations for 10 million customers.  For example, decisions like when to proactively reach out, what to serve up as an ice breaker, and how to pivot decisions as context is gathered – all while working within business constraints.  It’s impossible without the right 1:1 real-time solution.  And short of having this solution, many essentially fit recommendations to vast sets of customers (calling them segments).  They employ human judgement or rudimentary rules to treat them, such as:

  • Promoting a new service to gold tier customers
  • Offering a product to all customers who bought something similar in the last 3 months
  • Advertising holiday products to customers in the top two deciles

These all have something in common:  they are based on simple customer segmentation models, push products, and use simple rules or judgement calls.   And those calls are usually sub-optimal.

Focus on the right data and solution for the job

When the perfect mix of data and tools are effectively combined into a functional platform, they can work in harmony to quantitatively surface next-best-actions.  That can drive value and enhance CX.

Figure 1 depicts a reference technology / capability stack for CX:

customer analytics stack

Figure 1: CX stack

Concentrate on these principle areas:

  1. Data processing areas (orange) such as event processing, where highly valuable contextual data can be fed in
  2. Critical capabilities needed for customer analytics (yellow) like predictive analytics, where the solution calculates propensity to respond
  3. Value creation areas (green), where customers authenticate in owned channels, opt-in to tracking and profiling, and are served recommendations

There are many other boxes in figure 1.  Some are emerging technologies such as NLP (Natural Language Processing) and NLG (Natural Language Generation) and still need time to mature.  Others are useful for offline and retrospective analysis, such as Dashboards, BI (Business Intelligence) and Reports.  Yet the highlighted areas are where real-time data coupled with dynamic customer analytics and adaptive machine learning act in concert to deliver immediate decisions to customers’ moments of truth.

Conceptually depicted in figure 2, best next actions bubble to the top by surviving the scrutiny of each layer.  Guiding principles & operating rules (deep blue) are configured and underpin all strategies.  For instance, some offers will expire or reach caps.  Others will be invalid in certain jurisdictions or not applicable to certain customers.

Customer behavior & context data (orange) are fed into propensity or intent detection models.  Customer analytics (yellow) are used to calculate things like intent and “P” – propensity to respond to given offers.  “V” – Value, is assigned to each offer/action in the form of price, revenue, or margin. Finally, arbitration methods & resulting next-best-actions emerge and are delivered to the interaction point.  If customers aren’t interested, the system learns with each interaction and models are updated in real-time.

NBA framework

Figure 2: Guiding principles and resulting next-best-actions

Serving relevant and timely recommendations that are economically viable is vital to business success.

Predictive customer analytics at work

Here are just a few cool examples of stochastic customer analytics and next-best-action in real use today:

  • Based on a given product’s useful life, and information about its owner’s activity, predict when a replacement will be needed, and calculate propensity for interest in an upgrade. For example, an individual who owns a 6-year-old smartphone (and repeatedly runs out of memory) should score high on propensity to respond to a new device offer.
  • Using a customer’s product consumption patterns, calculate when refills or replenishment is required (such as printer ink or pet supplies). Trigger offers once replacement need is apparent.
  • Use sentiment analysis to monitor a customer’s self-service interaction, and if too much time goes by and/or the customer’s voice (or other activity) indicate rising frustration, escalate the interaction to a human.

Use the wisdom of data, human and machine crowds

Customer service is the act of caring for and meeting customer’s needs by delivering professional, prompt, polite, personalized, and useful consultation and assistance. As a brand, your ability to provide consistently high-quality service depends on many things, like well-trained agents, and flawless machine-learning systems that crunch, understand, and orchestrate customer analytics.

Great customer service ratings are never the result of one data source, a few humans, or one massive machine.  Instead, high customer satisfaction happens because many machines coupled with empathetic human teams perform an elegant dance that ensures massive amounts of customer data, algorithms, and delivery mechanisms work in concert around customer-centric goals.

Conclusion

Presently, customer data and customer analytics are plentiful.  Yet quantity does not guarantee quality.  And even quality, without speed, can still mean missed opportunity.  So, it takes quality and speed to win.  As such, you need a CX platform that can process immense quantities of disparate data, filter signal from noise, learn automatically, and trigger actions when the iron is hot.

Great cooks, artists, scientists, and inventors share many traits.  One that’s timeless is they never start from scratch.  They learn from the wins and sins of the past, pushing up the state-of-the-art in their respective fields.

Those committed to delivering better CX to customers must do the same.  Focus on meaningful data that can fuel insights and champion analytics that will surface those insights.  Take immediate and continuous actions based on those insights.  And unremittingly learn and improve with each insight and recommendation delivered.

[i] Forbes, https://www.forbes.com/sites/tomcoughlin/2016/07/24/the-costs-of-storage/#509ac6333239, July 2016

[ii] Destination CRM, https://www.destinationcrm.com/Articles/ReadArticle.aspx?ArticleID=128276, October 2018

[iii] McKinsey, https://www.mckinsey.com/business-functions/marketing-and-sales/our-insights/why-customer-analytics-matter, October  2016

LEAVE A REPLY

Please enter your comment!
Please enter your name here