RFM

RFM stands for Recency, Frequency, and Monetary value. It is a framework for classifying an existing customer database so you can build meaningfully different audience segments from it, rather than uploading everyone as one undifferentiated blob.

The three dimensions

Dimension	What it measures	Example tiers
Recency	How recently a customer interacted or purchased	High / Mid / Low
Frequency	How often they visit or transact	High / Mid / Low
Monetary	How much they spend	High / Mid / Low

You define what “high”, “mid”, and “low” mean for your client’s business. There is no universal threshold. You can measure monetary value per visit (event-level) or per period (e.g., monthly total); that choice depends on what the client’s data contains and what the business model looks like.

Why it matters

Not all customers behave the same way. A database of 80,000 customers might include:

• Impulse buyers who spend a lot but come rarely
• Deal-hunters who visit frequently but spend little each time
• Loyal regulars who come every week and spend a consistent amount

Treating them as one audience produces generic insights. RFM lets you split them and study each group separately.

How Consumr.AI uses RFM

The recommended workflow is:

1. Classify the client’s first-party database along the three RFM axes.
2. Build segments, typically a high-value, mid-value, and low-value audience.
3. Upload those audiences to Meta as custom audiences. The client does the upload; Consumr.AI never receives raw first-party data directly. This is the privacy-safe model.
4. Find lookalikes at the 1% level for each segment.
5. Run behavior reports on those lookalikes and compare them across segments.

The comparison is where the insight lives. When you put a high-monetary-value lookalike beside a low-monetary-value one, you can see how their interests, behaviors, and media habits diverge, and use that difference to build targeted campaigns, survey respondent pools, or AI twins.

RFM as a seeding method, not a taxonomy

RFM produces segments; it does not describe what those segments care about. The behavioral content (interests, channel preferences, and intent signals) comes from Consumr.AI’s reports after you have built the lookalikes. RFM opens a structured way into the database so the platform has a meaningful seed to work from.

Connection to other concepts

• Data Foundations explains how custom audiences and lookalike audiences work on the platform.
• Respondents & Extrapolation covers how a cohort of uploaded users maps to the full population.
• Sampling & Sample Size explains why each RFM tier needs sufficient size before you can build reliable research from it.
• Design Effect (DEFF) addresses what happens when an audience segment is too clustered to represent the broader population.