Next Best Action (NBA) – Choosing an Appropriate Follow-up Interaction
Contents
What does next best action mean?
Next best action, or NBA, is a recommendation for an appropriate next step in a customer relationship. It may suggest delivering requested material, arranging a conversation or pausing contact. Recommendations can be generated by business rules, statistical models or AI methods.
The term describes decision support rather than an automatically correct answer. What counts as appropriate depends on the objective, context, available information and actions that may legitimately be taken.
How can NBA support pharmaceutical organizations?
Communication across channels produces different signals and outstanding tasks. An NBA approach can help convert this information into practical priorities, particularly when several teams interact with the same professional.
A recommendation should have an understandable reason. An explicit question can justify a follow-up activity, for example. A technical email-open event, by contrast, may provide little evidence of a particular information need. Teams should distinguish strong evidence of a request from weak behavioral signals.
How should an NBA approach be designed?
Before selecting a model, organizations should define the permitted decisions and the conditions attached to them.
- Define the objective: Explain the need the recommendation should address and what a useful outcome means.
- Limit the options: Include only approved content, permitted channels and services that are actually available.
- Check prerequisites: Consider current permissions, consent, ownership and contact limits before execution.
- Observe results: Assess recommendations, rejections and actual outcomes, then update the logic accordingly.
Choosing not to contact someone should be an available outcome alongside recommendations for another activity. Otherwise, the model may systematically favor unnecessary communication.
What risks and limitations require attention?
Incomplete or outdated information can produce consistently inappropriate suggestions. Optimizing only for short-term clicks may also damage long-term relationship quality. Clear criteria and regular professional review are therefore necessary.
Employees need a way to reject a suggestion for a documented reason. Consent and other legal prerequisites must be current for the actual action rather than treated merely as historical model inputs. Use of sensitive personal information requires its own assessment and must not be assumed lawful because a model can process it.
Compare recommendations with existing practice
Before deployment, define what would make a recommendation better, such as faster responses or fewer unnecessary contacts. Employee rejections can reveal missing information or unsuitable rules. A controlled comparison can help assess the recommendation’s incremental contribution. Counting how many suggestions were executed does not answer that question.
Conclusion: Make recommendations bounded and assessable
Next best action can translate complex information into practical suggestions. Its usefulness depends on suitable data, legitimate options and understandable reasoning. Human accountability and current permissions remain essential even when the recommendation process is automated.
