Overview
This guide explains how to use the BlueConic AI Agent to identify which profile attributes and behavioral signals are most strongly associated with high-value customers.
By completing this process, you will have:
A ranked list of predictive signals
A clear, measurable definition of “high value”
Actionable inputs for building scalable audience tiers
This framework can be applied to any retail or DTC brand, but the prompt can of course be adapted to your business model as needed.
Before you begin
Ensure the following prerequisites are met:
Your tenant contains meaningful purchase and/or engagement data.
Key profile properties are available (e.g., revenue, transaction count, recency metrics, predictive scores).
You have permission to run AI Agent analyses.
If your definition of value relies on time-based behavior (e.g., last 12 months), confirm that relevant timeline events or rollup properties are available.
Step 1: Define What “High Value” Means
Before running the AI Agent, define high value in measurable, outcome-based terms.
Examples:
High lifetime revenue
Multiple purchases within the last 12 months
Strong engagement or behavioral score
Progression to premium products or subscriptions
Be specific. Avoid vague definitions such as “good customer.” If helpful, establish thresholds (e.g., top 20% by revenue, 3+ purchases in 12 months).
A precise definition will directly improve the quality of AI output.
Step 2: Access the AI Agent
Log into BlueConic.
Navigate to the AI Agent.
Start with an empty prompt
Example Prompt
Analyze BlueConic profile data to identify which profile properties, behavioral events, and calculated attributes are most strongly associated with high-value customers for [Brand Name].
Define “high value” as profiles that show one or more of the following outcomes:
High conversion frequency or revenue contribution
Strong engagement depth or repeat usage
Progression to premium products, subscriptions, or advanced use cases
Determine:
Which profile properties (e.g., demographics, preferences, declared interests, consent choices) are most common among high-value profiles
Which behavioral signals (e.g., content consumption, feature usage, campaign interactions, frequency/recency patterns, engagement score, purchased products) best differentiate high-value vs. low-value profiles
Return the results as:
A ranked list of the top 10 most predictive properties/signals
A short explanation of why each signal matters
Recommendations for how each signal can be used in audience segments, personalization, or activation
Assume the goal is to create scalable, explainable high-value audiences that can be reused across channels. Send the results to [name]@[brand].com.l.
Step 4: Review and Interpret the Output
The AI Agent will typically return an email containing:
A ranked list of predictive signals
Value-related indicators (e.g., lifetime revenue, purchase frequency)
Recency indicators (e.g., months since last purchase)
Engagement signals (e.g., click activity, product affinity)
Risk indicators (e.g., churn score, inactivity)
Focus on patterns across signals rather than a single top metric.
Look for clusters such as:
Spend + frequency
Recency + predicted purchases
Engagement + product affinity
High-value customers are usually defined by combinations of signals rather than one standalone property.
Note: This is just an example, different prompt will results in different outputs depending on your goal.
Step 5: Document the Key Signals
Create a concise working summary that includes:
The 5–10 strongest predictive signals
Any recommended thresholds
Observed signal combinations (e.g., “high-value profiles consistently show X + Y”)
These signals will serve as the foundation for building tiered audiences and value-based segmentation strategies.
Step 6: Optional Troubleshooting
If the results are unclear or overly broad:
Refine the definition of high value
Add timeframe constraints (e.g., “within the last 12 months”)
Remove overly broad criteria
Re-run the analysis with a narrower focus
Tip: AI output quality depends on the precision of your outcome definition and the relevance of the underlying data. Sometimes, the BlueConic Agent cannot process your request when it becomes too complex. Giving it another try by refining or shortening the prompt can help to overcome this situation.
Step 7: Build and Activate Your Audiences
Using the signals documented in Step 5, begin building Segments in BlueConic based on the strongest predictive properties and signal combinations.
Translate the identified signals into clear segmentation logic. For example:
High revenue + high purchase frequency
Recent activity + high predicted value
Strong engagement + premium product progression
Previously, identifying the right properties often required manual review of the data dictionary, validating which fields were populated, interpreting definitions created by other teams, and running exploratory segment tests. This process could take hours before even arriving at a usable audience definition.
With the AI Agent, the inventory, validation, and classification of available data are accelerated. The AI surfaces which properties matter, how they relate to your defined outcome, and which combinations are most predictive.
The human role remains essential.
You validate whether the signals make business sense, apply contextual knowledge about your brand and strategy, adjust thresholds, and refine the segmentation logic. The AI identifies patterns; you apply judgment and translate those patterns into activation strategy.
This human-in-the-loop approach reduces time spent searching for data and increases time spent optimizing audience strategy, testing, and driving measurable impact.