A favorite Boy Scouts mantra about the importance of being prepared is that it’s better to have and not need than need and not have. In the same vein as it relates to customer experience (CX) is the importance of having real-time capabilities for delivering a next-best action. While not every next-best action needs to be in real time, the decision that generates a next-best action must be based on accurate, real-time data. In other words, a real-time view of the customer during an ongoing, fluid customer journey is critical in deciding if a real-time response is – or is not – the next-best action for a specific moment during that journey.
What is a Next-Best Action?
To better understand the importance of a real-time engine in delivering a next-best action, let us first examine what we mean by both next-best action and real time. A next-best action is, at its core, meeting the customer where the customer is with an action, an offer or a piece of information that precisely matches the customer’s desires and inclinations. The concept inverts the traditional “marketing-out” approach which is to guide the customer along a path to purchase. It does so by striving to innately understand the customer. Selling a product becomes secondary to a brand or organization making the customer recognize that the brand understands them, which is accomplished by a recognition that the customer is in control of a dynamic, customer-driven journey. A next-best action, then, is an action that is perfectly aligned with a customer’s aspirations in the moment of the journey that it is delivered.
A next-best action may or may not directly drive a brand’s measurable goals – whether those are to sell a product, enhance CLV, reduce churn, etc. A brand that knows a customer is interested in hiking, for instance, might display an image of a mountaintop sunrise during a web visit by a customer who just bought hiking boots, and that aspirational message could be the next-best action for that customer in that moment. But if the reason for the visit is to start a return, a mountaintop image might introduce friction. A next-best action must recognize this context, meaning that it should be at the nexus between what the brand thinks is the next action to take for the customer – move along to purchase, loyalty, retention – and what the customer wants.
Part of the context for a next-best action is to respond in the channel where the customer is operating. If a customer is on the mobile app, a next-best action will likely be in the mobile app as the customer journey unfolds on that channel. A next-best-action during an active engagement must also feel conversational. From the customer’s perspective, it is a fluid, unforced and relevant interaction that feels natural within the context of the engagement.
The Meaning of Real Time
Making a next-best action part of a natural conversation sometimes means the real time arrow must stay in the quiver. In an abandoned shopping cart situation, a customer who receives an instantaneous message about an abandoned cart with an offer to come back and complete the transaction may be annoyed, feeling the brand respects neither their desires nor proper boundaries.
Conceptually, a real-time framework is easy to understand. It may be as instantaneous as the brand can possibly make it: The customer arrives on a homepage, and the mountaintop sunrise image loads within milliseconds. Or, for an abandoned shopping cart, perhaps the next-best action is to wait 10 minutes or even longer depending on a customer’s next signal. In between those extremes, another real-time construct is a human-to-human conversation, such as a customer dialing up the call center, or interacting with a store associate or a front desk clerk. The overarching consideration in any interaction with the customer is that real time is whatever time is required to ensure that the next-best action is frictionless, relevant and organic to the dynamic, customer-driven journey.
With this goal in mind, it becomes clear why having pristine, real-time data is vital for orchestrating real-time decisions into the right touchpoints – and why every next-best action depends on having this real-time understanding. Consider, for example, the constraints imposed with a one-minute data refresh. For an online session, the images, text, banners or pop-ups that a brand shows a customer that are even 10 seconds or so behind the dynamic journey may be irrelevant. Perhaps they’re close to the customer’s aspirations at that moment, but even being slightly off indicates to the customer that that the brand does not fully understand them. Or consider the shopping cart example, where real time may mean waiting for another signal from the customer before the brand responds. If the brand is one minute behind the customer, that response may not be optimized for the customer’s follow-up signal, whatever it may be. If placing another item in the cart warrants an instantaneous response, one that is a minute or two later may not be relevant or may seem unnatural. And in a human conversation, it’s easy to see that friction is a likely result if the brand’s agent lacks a real-time understanding of the customer journey.
To ensure decisions are organic, smooth and in a cadence that matches the journey, a real-time capability must be infused with intelligence to make next-best action decisions. That intelligence is formed through a combination of human-curated rules and machine learning models. Both are required because a next-best action is typically the product of many things happening at once. A simple example is the brand should not show an image of a sweater that is not in stock in the customer’s size or color. Human-curated rules can encompass those types of inventory or supply chain constraints, but also handle other situational details like matching a message to a customer’s location: If a customer enters this geo-fence, they are a candidate for that offer.
Machine learning models are the predictive and prescriptive decisions made about a customer. Is she in danger of churning? Is he ripe for an upsell? Machine learning models ensure that the brand is probing every possible signal to arrive at the most detailed, accurate and precise view of the customer. The human-curated rules may then serve to narrow a potentially long list of possible actions to take based on that complete understanding.
Finally, the intersection of what the machine learning models tell a brand, combined with the human-curated rules, is itself not a static process. As dynamic as the journey it is probing, a real-time engine is continuously arbitrating among a set of decisions. Real time, then, is not just about when a next-best action is presented, it is also about how long an action is valid – how long it remains relevant. Any customer action has the potential to swap out one potential action for another that is more relevant and more natural to the customer journey. In other words, a static list and ordering of decisions – if the customer does A, offer B first, C second – may seem arbitrary or chaotic if the list is based on predictions and/or rules that are not in the context of the journey.
One way to think of real time in the context of a next-best action is to envision the flawless execution of a championship ballroom dance team. One action dictates the next in a seamless combination of maneuvers that appear natural, almost magically so. The ballroom orchestra is the real-time engine that carries a continual beat and underpins the entire performance. The routine is perfectly timed to the crescendos and bar changes, so fluid it is hard to discern which plays off the other. The thunderous applause at the end is a recognition of something special; a series of next-best actions that delight the customer because it makes them feel like they’re with a brand that understands who they are, not just what they want.