What is next best action?
Next best action (NBA) is a strategy for determining the single most valuable thing a business can do for a specific customer, at a specific moment, across any type of interaction, not just a sales opportunity. It goes beyond deciding what to sell someone. It asks a broader question: given everything we know about this customer right now, what's actually the best thing to do next?
That "best thing" could be a product recommendation, but it could just as easily be a support message, a check-in after a delivery issue, a churn risk intervention, or simply no action at all if now isn't the right moment. Next best action treats every customer touchpoint, in-store, online, over email, through a loyalty app, as an opportunity to make a smarter, more relevant decision than a generic, one-size-fits-all response.
At Leat, we see next best action as the natural evolution of how businesses use customer data. Rather than running the same campaign to an entire list, or offering the same reward to every loyalty member, NBA uses customer behavior, purchase history, and machine learning to tailor the next interaction to the individual, cross-channel, in real time.
Why "next" and "best" both matter
The word "next" keeps the focus narrow and immediate: not a long-term strategy for a customer, but the single most useful thing to do right now, given where they are in their relationship with the business. The word "best" acknowledges that there are usually several reasonable options - a discount, a recommendation, a support touchpoint - and NBA is about choosing the one most likely to serve the customer and the business well.
The benefits of next best action
Increased customer loyalty
When customers consistently experience relevant, well-timed interactions rather than generic marketing and customer messaging, they start to feel understood rather than targeted. That sense of being known, reinforced over time, is one of the more reliable drivers of long term customer loyalty.
Increased customer lifetime value
Because NBA improves the relevance of every interaction, not just sales-focused ones, its effects compound. A customer who consistently gets useful, well-timed interactions tends to stay engaged longer and spend more over their relationship with a business, which shows up directly in customer lifetime value.
Improved customer experience
NBA's biggest advantage over a purely sales-focused approach is that it accounts for moments where a sales message would be the wrong move entirely. Recognizing when a customer needs support rather than a cross-sell offer, and acting accordingly, measurably improves the overall customer experience.
Increased time to value for customers
By surfacing the most relevant next step immediately, whether that's a product recommendation, a helpful piece of content, or an answer to a question, NBA helps customers get value out of their relationship with a business faster, rather than making them dig for it themselves.
Higher conversion rates
When a business does present a commercial offer through an NBA strategy, it's far more likely to convert, since it's been selected specifically because it fits that customer's behavior and current context, rather than being a generic offer sent to everyone regardless of relevance.
Perfect timing in customer interactions
Even the most relevant offer or message underperforms if it lands at the wrong moment. NBA's real-time nature means the business isn't just picking the right message, it's picking the right moment to deliver it, which matters just as much as the content of the message itself.
How next best action works in practice
At a high level, next best action combines three things: customer data, a decisioning model, and a defined set of possible actions to choose between.
Gathering the inputs
Every NBA decision starts with data: a customer's purchase history, browsing behavior, loyalty activity, support interactions, and any other relevant customer data available. The more complete and current this picture is, the better the resulting decision tends to be.
Running it through a model
A machine learning model, or in simpler setups, a set of business rules, evaluates this data and determines which action is most likely to be the right one for that specific customer, in that specific moment. This might mean recognizing a churn risk signal and triggering a retention offer, spotting a pattern that suggests a strong cross-sell opportunity, or noticing that a customer is in the middle of a support issue and should be routed toward help rather than a promotion.
Delivering the action
Once a decision is made, it needs to actually reach the customer through the right channel, at the right time, whether that's a message in a loyalty app, a personalized homepage on a returning visit, or a proactive check-in from a support team. This is where NBA strategy work moves from theoretical scoring to something a customer actually experiences.
A simple example
Picture a customer who's browsed a product category several times without purchasing, and separately submitted a question through customer support last week. A next best action system weighing both signals might decide that the right move isn't a discount on the browsed products at all, it's making sure that support question was properly resolved first, since an unresolved service issue is a much stronger churn risk than an unconverted browsing session. Once that's addressed, a relevant product recommendation becomes a much more appropriate next action.
Combining next best offer, next best action, and next best experience
These three concepts build on each other, each one widening the scope of what "best" actually means for a customer.
Next best offer is the narrowest of the three, focused purely on the commercial question: which product or service should we recommend to this customer right now? It's a subset of next best action, useful specifically in moments where a customer genuinely is in a buying mindset.
Next best action sits a level above that. It decides whether a commercial offer is even the right move in the first place, or whether the better move is support, information, or no action at all. Next best offer effectively becomes one possible action within a broader NBA framework, chosen only when the context actually calls for it.
Next best experience widens the lens further still, looking beyond a single interaction to the overall journey a customer has with a business over time. Where NBA optimizes one moment, next best experience considers how a sequence of well-chosen actions, across weeks or months, adds up to a coherent, hyper-personalized relationship rather than a series of disconnected good decisions.
In practice, these three work best as layers rather than competing strategies. Next best offer handles the specific "what to sell," next best action decides whether selling is even appropriate right now, and next best experience makes sure each of those individual decisions is building toward a consistent, positive long-term relationship.
How next best action marketing compares to traditional marketing
Traditional marketing typically works from the outside in: a business decides on a campaign, a segment to target, and a message, then pushes that same message out to everyone who fits the segment. It's broad by design, built around planned campaigns rather than individual customer behavior.
Next best action marketing flips that structure. Instead of starting with a campaign and finding a segment to send it to, it starts with the individual customer and works out, moment by moment, what the most relevant thing to show or say to them is. The "campaign" becomes a pool of possible actions the system can choose from, rather than a single message blasted to everyone at once.
A few concrete differences worth highlighting:
Timing. Traditional marketing runs on a schedule: a weekly newsletter, a seasonal campaign. NBA runs continuously, reacting to customer behavior as it happens rather than waiting for the next scheduled send.
Targeting granularity. Traditional marketing typically targets segments, groups of similar customers. NBA aims for the individual level, using a customer's specific data and current context, not just which broad segment they fall into.
Flexibility of the message itself. A traditional campaign sends the same message to everyone in the segment. NBA can select from many possible actions, an offer, a piece of content, a support touchpoint, and choose differently for each customer based on what's actually relevant to them.
Measurement. Traditional marketing is usually judged campaign by campaign. NBA is judged more on continuous improvement over time, since the models behind it are meant to get sharper as more customer interactions feed back into them.
Neither approach fully replaces the other. Broad campaigns still have a place for brand awareness or big, universal announcements. NBA is what fills the space in between those campaigns, the countless smaller, individual moments where a business has to decide what to actually do next for one specific customer.
The importance of collecting and unifying data for next best action marketing
Why unified, cross-channel data matters
A customer who shops in-store, browses online, and reaches out to support should be recognized as the same person across all of it, with a single unified profile rather than three disconnected records. Businesses that fail to unify this data end up with an NBA system that's only ever working from a partial picture, which shows up as decisions that feel slightly off, a product recommendation for something the customer already bought elsewhere, or a promotional message sent right after an unresolved support issue.
The role of real-time data collection
Beyond unifying data across channels, NBA depends on that data being current. A model working from purchase history that's a month stale will miss recent signals, a customer who's just started browsing a new category, or one who's gone quiet after usually being highly active, both of which matter a great deal to choosing the right next action.
Data quality over data volume
More data isn't automatically better if it's inconsistent, duplicated, or poorly structured. A smaller set of clean, well-organized customer data points tends to produce more reliable NBA decisions than a large volume of messy, disconnected data spread across systems that don't talk to each other.
Building the foundation before the models
It's worth being honest about sequencing here: unifying and cleaning customer data is unglamorous work compared to building sophisticated models on top of it, but it's the part that determines whether everything built afterward actually works. A business investing in advanced machine learning models for next best action, without first solving for unified, high-quality data, is building on a shaky foundation.
Models powering next best action
Predictive models
Predictive models use historical customer behavior to forecast what's likely to happen next, whether a customer is likely to make a purchase soon, likely to churn, or likely to respond to a specific type of message. These models form the backbone of a lot of NBA decisioning, since they turn raw customer data into a forward-looking signal the system can actually act on.
Propensity models
Propensity models are a specific type of predictive model focused on the likelihood of a particular action, a propensity to buy, to churn, to redeem a reward, or to respond to a specific offer. NBA systems often run several propensity models in parallel, then compare their outputs to decide which action has the highest likelihood of success for a given customer.
Segmentation models
Segmentation models group customers based on shared characteristics or behavior patterns. While NBA aims for individual-level personalization, segmentation still plays a useful role as a starting point, giving the system a reasonable baseline to work from before it has enough individual history to make a highly confident, customer-based decision.
Recommendation systems
Recommendation systems specialize in surfacing specific product recommendations or content based on a customer's past behavior and the behavior of similar customers. These are especially relevant for the next best offer layer within a broader NBA framework, powering cross-sell and upsell suggestions specifically.
Rule-based models
Not every NBA decision needs a sophisticated model behind it. Rule-based models apply straightforward business rules: if a customer hasn't purchased in 60 days, trigger a win-back message; if a support ticket is still open, suppress promotional messaging. These are simpler to build and easier to understand than machine learning models, and they remain a practical, transparent option for many NBA decisions, especially early on.
Reinforcement learning frameworks
Reinforcement learning takes a more dynamic approach, where a model learns which actions work best over time by directly observing the outcomes of past decisions, and adjusting future decisions accordingly. Rather than being trained once on historical data, these frameworks continuously refine themselves based on real customer responses, making them well suited to NBA systems aiming for genuine continuous improvement rather than a static, one-time model.
Implementing real-time decisioning in your business
Getting to real-time decisioning is rarely a single step, and it's worth approaching it in stages rather than trying to build a fully mature NBA system from day one.
Start with unified customer data.
As covered above, this is the non-negotiable foundation. Without a single, current, cross-channel view of each customer, no model built on top of it will perform reliably.
Define your possible actions clearly.
Before building or buying any decisioning technology, get specific about what "actions" actually means for your business, offers, content, support routing, loyalty perks, and so on. A system can only choose well between options that have been clearly defined upfront.
Start with simple rule-based logic.
Rather than jumping straight to machine learning, many businesses get real value from well-designed business rules first. This builds internal understanding of what good decisioning looks like, while surfacing gaps in the underlying data that need fixing before more sophisticated models make sense.
Layer in predictive and propensity models progressively.
Once the data foundation and basic rules are solid, predictive models can start informing decisions that rules alone can't handle well, particularly around timing and prioritization when multiple possible actions compete for the same moment.
Measure, and treat it as ongoing.
Real-time decisioning is a capability that needs continuous improvement as customer behavior shifts and more data accumulates. Building in regular review of what's actually working is as important as the initial setup itself.
Which industries does NBA work best for?
Next best action tends to deliver the most value in industries where customers interact frequently, across multiple channels, and where the right message at the right moment genuinely changes outcomes. A few stand out:
Retail and e-commerce, where customers interacting across web, app, and in-store touchpoints generate rich behavioral data, and where timely cross-sell and win-back actions have a clear, measurable commercial impact.
Hospitality, where repeat visit patterns, loyalty activity, and timing (a customer who hasn't visited in a while, a customer celebrating an occasion) create frequent, meaningful opportunities for a well-chosen next action.
Financial services, where the mix of high-stakes decisions, churn risk, and a genuine need to balance commercial offers against real customer needs makes NBA's broader framing (support versus offer) especially valuable.
Telecommunications, an industry with famously high churn rates and frequent customer interactions, both across support and billing, making it a strong fit for NBA's ability to prioritize retention actions over sales pushes when the data calls for it.
Subscription-based businesses generally, where the ongoing nature of the customer relationship gives an NBA system many recurring touchpoints to learn from and act on over time.
That said, NBA isn't exclusive to these industries. Any business with enough customer interaction volume and data to train reliable models can benefit; the industries above are simply where the return on investment tends to show up fastest and most clearly.
Final word
Getting NBA right isn't really about the sophistication of the models behind it, though that matters. It's about building the unified, high-quality data foundation those models depend on, and being honest about when a commercial offer is the right move and when it isn't. Businesses that get this balance right end up with something more valuable than a higher conversion rate on any single interaction: a customer relationship that feels consistently well-timed, relevant, and genuinely attentive, interaction after interaction.
