Next Best Action (NBA) for dialogues determines, in real time, which dialogue to show to each individual visitor. Instead of relying on fixed rules, NBA uses behavioral data and a self-learning model to select the option most likely to achieve your optimization goal, such as conversions, click-through rate, or revenue. It prioritizes the option with the highest expected value, not just the highest conversion rate. For example, a subscription might convert less often than an ad click but deliver higher long-term value. The model accounts for this and prioritizes accordingly.
For example, NBA can choose between:
A monthly donation versus a one-time donation
A paywall versus ad exposure
Different discount levels within the same placement
The model continuously learns from incoming data and improves its decisions over time.
Availability: This feature is available to a limited set of customers. Contact BlueConic to request access.
Before you begin
Confirm access to the AI Workbench.
Have at least two dialogues.
Relevant profile properties available for personalization.
Create a next best action notebook
Log into BlueConic.
Via the navigation, select More.
Select AI Workbench from the dropdown
Open the Notebooks tab.
Click Add notebook.
Select Next Best Action notebook. The notebook opens with a guided setup interface.
Enter a name for the notebook.
Click Save.
Configure the optimization goal
The optimization goal defines how the model evaluates success.
In the notebook, locate the Optimization goal section.
Select one of the following:
Click-through rate
Conversion rate
Conversion value or revenue
Save your selection.
Note: Ensure that all included dialogues support the selected goal through their conversion settings.
Select dialogues to include
Add the dialogues or variants that the model should evaluate.
Go to the Dialogues section in the notebook.
Click Add existing dialogues.
Use search or filters (such as labels) to find relevant dialogues.
Select one or more dialogues.
(Optional) click Create new dialogue to open a modal to configure a dialogue. Configuration can be completed later in the Dialogues page using the link that will appear in the table.
To remove a dialogue:
Locate the dialogue in the list.
Click the (X) remove icon.
Select profile properties (hypotheses)
Profile properties help the model understand which factors influence outcomes.
Go to the Profile properties section.
Select the properties you want the model to use, such as:
Interests
Behavioral data
Customer type
Save your selection.
These properties are labeled as hypotheses to encourage deliberate selection of meaningful signals.
Configure the model training period
The model training period determines how far back in time the model uses profile activity data to learn and make predictions.
In the notebook, locate the Look back window setting.
Select a time range (for example, last 7 days or last 30 days).
Shorter look-back windows allow the model to adapt more quickly to recent behavior. Longer windows provide more stable predictions by including more historical data.
Choose a window that balances responsiveness and stability based on your use case.
Configure the retraining schedule
Go to the Schedule tab in the notebook.
Click the settings icon to open the schedule window.
Select the appropriate schedule.
Click OK.
Click Save.
Use more frequent retraining when your data changes rapidly. Use less frequent retraining for more stable environments.
View performance metrics
The notebook provides insights into performance.
You can review:
Views and conversions per dialogue or variant
Progress toward the optimization goal
Relative performance between options
You can also add these Insights to Dashboards.
Create a dashboard from the notebook
You can create a dashboard directly from the notebook to quickly visualize performance insights.
Open your Next Best Action notebook.
Go to the Insights tab.
(Optional) Adjust the time period for the insights.
Click Create dashboard.
Open the generated dashboard using the confirmation message.
The dashboard is preconfigured with relevant insights for your notebook, allowing you to immediately analyze performance and share results.
Add dialogue comparison insights to a dashboard
Go to the Dashboards page.
Open an existing dashboard or create a new one.
Click Add insight.
Add one of the following plugins:
Dialogues KPI comparison
Dialogues comparison
Configure the insight:
Select one or more dialogues as the test group (for example, NBA-driven dialogues).
Select one or more dialogues as the control group.
Choose a KPI, such as conversions per view, clicks per view, or average conversion value per view.
(Optional) Set a time interval or comparison period.
Save the insight.
These insights allow you to compare performance between groups and quantify lift directly in BlueConic dashboards.
Monitor model performance over time
Use the Notebook performance insight to track how your model performs across training runs.
In the Dashboards page, click Add insight.
Add the Notebook performance insight plugin.
Select your AI Workbench notebook.
Choose one or more performance metrics. For the NBA notebook, bandit score and conversion value regressor score are available.
Save the insight.
The widget displays model performance as a time-series chart based on completed notebook runs. Use it to monitor model health, detect performance changes over time, and evaluate the impact of retraining.
Next steps
Monitor model performance after the first training run.
Compare NBA-driven dialogues against a control group in your dashboard to quantify lift and validate the optimization goal.
Refine your profile property hypotheses based on early model performance — remove low-signal properties and add higher-value ones.