Who should receive an action?
Prioritise entities by expected incremental value rather than a single propensity score.
Next Best Action · Decision Optimization
Choose the action, not just the score.
Combine predictive signals, causal evidence, expected value and real constraints to decide which action should happen next for each customer, account, store or operational unit.
Questions this decision model answers
Start with the commercial or operational choice.
Which action fits best?
Choose among offers, messages, interventions or operating actions based on expected outcome and eligibility.
Is the action worth taking?
Include expected value, treatment cost, margin and downstream consequences.
What should happen first?
Respect contact limits, budgets, inventory, staffing or other scarce resources.
From data to decision
Turn model outputs into a ranked action.
Next-best action is a decision layer. The right inputs depend on whether the business is optimizing retention, cross-sell, marketing, inventory or operations.
Predict likely outcomes
Use propensity, risk, demand, customer value or other predictive signals relevant to the decision.
Estimate intervention impact
Use causal evidence where the goal is to know whether an action changes the outcome rather than merely correlates with it.
Apply eligibility and constraints
Encode which actions are allowed, available and operationally feasible.
Choose the next action
Optimise expected business value and return the recommended action with supporting evidence.
Decision boundary
The highest score is not always the best action.
A customer can have high purchase propensity but low incremental response to an offer. An intervention can create uplift but destroy margin. Next-best action combines these layers so the recommendation reflects expected business consequence, not one model score.
What is likely?
Estimate risk, propensity, demand or value.
What can the action change?
Estimate incremental effect when causal evidence is available.
Which action wins?
Choose among eligible actions under economic and operational constraints.
Related decision problems
Follow the decision chain.
Direct answers
Questions buyers ask about this decision.
Is next-best action just a recommendation engine?
Not necessarily. A recommendation engine often predicts relevance. Next-best action can also include causal impact, economics, eligibility, capacity and business constraints.
Does every use case need causal modeling?
No. The method should match the decision. Causal modeling is most useful when the business needs to know whether an intervention changes an outcome.
Can humans remain in the approval loop?
Yes. The decision object can carry the recommendation, expected impact and evidence while the responsible team keeps final approval.
Can next-best action be delivered through an API?
Yes. Graphite Note can expose recommendations through a decision layer or API so they can enter CRM, ERP, BI, workflow or other operational systems.
Decision IntelligenceFrom signal and prediction to a governed business action.AutoMLAutomate repetitive model-building work without losing business context.Causal Machine LearningSeparate drivers from correlations and estimate intervention effects.Predictive AnalyticsForecast demand, score customers and anticipate business outcomes.Decision OptimisationCompare feasible actions and rank them by expected business impact.Forecasting vs OptimisationSee where prediction ends and the business decision begins.Decision APIDeliver recommendations, expected impact and evidence into workflows.Embedded Decision IntelligenceAdd a decision layer to ERP, CRM, BI, workflow and vertical software.Build vs BuyA practical framework for deciding whether to build, buy or embed.
Bring the decision, constraints and data you already have.
We will structure the model around the action your team needs to take and the evidence required to support it.