What is RFM analysis?

RFM analysis ranks customers by recency, frequency and monetary value, which means how recently, how often and how much they buy. Intempt uses it to sort every customer into six named buckets, like Champions and At-Risk.

How do you interpret RFM scores?

A high RFM score means a customer bought recently, buys often and spends a lot, so they are one of your best. A low recency score means they have gone quiet and may need a win-back message.

What is a propensity model?

A propensity model predicts how likely a person is to take an action, like buying or upgrading, and gives them a score. Intempt’s Likelihood score is a propensity model you set up just by picking the outcome.

Do we need a data scientist to use this?

No. Intempt computes RFM, likelihood and next best product from the events you already send. You pick the outcome you care about, and the Data Analyst agent builds, trains and refreshes the model.

How current are the scores?

The scores update in real time as new events arrive. A customer who stops buying moves to a new bucket on their own, and no one has to rerun anything.

What are the best predictive analytics tools for marketing?

Klaviyo, GA4 and Segment all offer predictions, but each has limits, like 180 days of order history, 1,000 recent buyers or a Business plan. Intempt scores every profile live and gives every channel the same ranked list.

Which plan includes these scores?

RFM comes with the Professional plan. Likelihood and next best product come with the Organization plan. Seat prices are on the pricing page, and you can start on the free plan.

Predictive analytics for marketing

See who will buy next and what they’ll buy.

Score every customer with RFM, purchase likelihood and next best product, updated live, with no model or data scientist.

Start for free

Live scores
No code
Every channel

track('Order Completed')Data Analyst agent· scoring live

Maya Chenusr_48213 · 14 ordersChampions

Scores on the profile

RFM: R5 F5 M4Champions
Purchase likelihood: 82%
Next best product: Rain Shell Jacket

Updated on this event2s ago

VIP early access

Journey started

Your dialler

Call list

Added · rank 3

Years profitable: 7; years profitable,and counting
rating on G2: 4.5; rating on G2
events processed: 10B; events processed
user profiles: 500M; user profiles
enterprise customers: 50; enterprise customers

StockInvest
“Acquisition and retention finally feel like one connected motion instead of two separate problems.”
Jim StrombergCEO, StockInvest
FieldsUSA
“We finally have one system that sees the whole customer journey.”
Eric GardnerCOO, FieldsUSA
Hoperfy
“With Intempt, we built a signal-led pipeline driven by real behaviors.”
Tadas KertenisCo-founder, Hoperfy

Your team shouldn’t guess who to contact.

Today a score means exporting orders, building a model in Python, and uploading a list to one tool before it goes stale.

Start for free

The Data Analyst agent is scoring every profile live

Moving customers between buckets…

Data Analyst agent

Scoring RFM, likelihood and next best product on every event.

Profiles scored48,210
Bucket changes today318
Moved to At Risk42

Bucket changeWin-back started

Lina Ortiz64 days quiet

RegularsAt Risk

View journeyStep 1 · email today

Scored profilesLive

All buckets

Maya ChenOrder Completed · R5 F5 M4, 14 ordersChampions82%
Arjun PatelOrder Completed · 2nd order this monthPromisingRegulars71%
Lina OrtizWent quiet · no order in 64 daysRegularsAt Risk18%
Tom BeckerOrder Completed · first order, $86New Customers34%
Sofia RossiEmail Opened · no order in 41 daysNeeds Attention27%
Jonah ReyesProduct Viewed · Trail Runner 2Regulars58%
Priya NairScoring

Three scores

Three scores that tell you who to contact.

RFM shows who your best customers are, likelihood shows who will act next, and next best product shows what to offer.

01 · RFM

Sort every customer into six clear buckets

RFM analysis ranks each customer by how recently, how often and how much they buy, then sorts them into named buckets.

Six named buckets like Champions and At-Risk.
Buckets update every time a customer acts.
It’s built from the orders you already send.
There’s no spreadsheet or SQL to maintain.

intempt.com/attributes/rfm-score

RFM scoreAI attributeScore users in segment · All usersTime window 30 days

Share of customers · 6 buckets48,210 scored

BucketTotal%RecencyFrequencyMonetary

Champions6,12012.7%4-54-54-5
Regulars9,84020.4%3-43-53-5
Promising7,31015.2%3-42-32-3
New Customers5,46011.3%511-2
Needs Attention8,95018.6%2-32-32-3
At Risk10,53021.8%1-23-53-5

Segment migrationRegularsAt Risk412 peoplechurnView

What changes

Six things you stop doing by hand.

Data Analyst agent3 of 3 trained
- RFM score6 buckets
- Purchase likelihood5 tiers
- Next best productranked
0analysts staffedNot needed
No data scientist to hire
Intempt computes all three scores for you, so you never have to staff an analyst just to refresh a model.
0WarehouseScoring job
ProfileScoreDirect
No warehouse round-trip
Scores are computed right on the customer profile, so there is no warehouse trip and no scoring job to schedule.
Lina Ortiz64 days since last orderQuiet
RegularsAt RiskR3 → R1
Updated 2s agoNo rerun
Scores that never go stale
Every event updates the score in real time, so a quiet customer moves to At-Risk without anyone rerunning it.
One segment · three channels
Champions
6,120
ranked by score
- Email
- Dialler
- Analytics
One list for every channel
The same ranked segment feeds email, your dialler and analytics, so you build it once and use it in every place.
Describe your segment
Champions likely to buy this month
WhereRFMisChampions
AndPurchase likelihoodisVery High
2,740 people matchLive
Segments you can describe
Your team builds a scored segment by describing it in plain words, not by writing a query or a long formula.
Journey recipes · built in
- At RiskWin-backStarted
- ChampionsVIP early accessReady
- Very High likelihoodSales callReady
Journeys already built
The journey recipes that act on each score are already built, so a bucket change can start the right journey.

The difference

Here is how customer scoring should work.

Before and After
Before	After
You export orders to a spreadsheet and sort customers by hand every month.	RFM sorts every customer into six named buckets and keeps them up to date.
Your email tool won’t predict a thing until you have months of order data.	Likelihood scores any outcome you define, from the events you already send.
Someone builds a model once, and the scores are out of date a week later.	Every new event updates the score, so no one ever has to rerun the model.
The scores live inside one tool, so your sales team never gets to see them.	Scores sit on the profile, so email, calls and analytics all read the same.
Everyone gets the same bestseller, because no one has time to rank products.	Next best product ranks items for each person from their own past behavior.

The offer

What scoring costs the old way.

These are the bills a small team pays when the CDP, the custom models and the person who keeps them fresh are bought apart.

The old way

$97,600/yr

Year one, before warehouse compute.

Intempt

$49

per seat a month

Three custom models²$18,000
Data scientist upkeep³$2,000/mo
CDP licence¹$55,600/yr

The old way

$49 · per seat

Intempt

Built in

No model to host

The scoring model you were about to build

RFM across six named buckets.
Likelihood scores for any outcome.
Next best product for every person.
Scores that update on every event.

Runs on its own

The data scientist you were about to hire

Data Analyst agent
Runs the scoring skill for you.
AI Attributes
Saves each score on the profile.
Segmentation
Filters customers by any score.
Journeys
Acts the moment a score moves.

1Vendr marketplace, retrieved 2026-08-15. Twilio Segment median annual contract $55,600 across 613 purchases. Segment Predictions requires the Business tier.
2Fast Data Science, Reforge (2026). Custom prediction models typically take 20-40 hrs of data science consulting at $100-250/hr ($2,000-10,000 each); three models at a $6,000 midpoint.
3goLance (2026). Experienced freelance data scientists charge $105-178/hr; about 4 hrs a week at $120/hr is roughly $2,000/mo.

Control

You set the goal and the agent does the math.

You choose the outcomes that matter to your business, and the Data Analyst agent builds, trains and refreshes each score.

outcome: Order Completed
window: 30 days
refresh: every event
export: none

You pick the outcome
Choose the event you want to predict, like a purchase or an upgrade, and the score is built around that one goal.
Scores stay on the profile
Each score is saved as an attribute on the customer’s profile, so it never has to be exported to be useful.
Your team pulls the list
Anyone on the team can build a scored segment by describing it, so the request never has to come back to you.
No handoff to a data team
The agent handles the model work from start to finish, so there is no ticket, no queue and no analyst to staff.

Integrations

Wired into the tools your team already uses.

See all integrations

JavaScript
Apache Kafka
Slack
Stripe
Amazon S3
Twilio
Gmail
iOS
SendGrid
Webhooks
Node.js
Google Calendar

Go deeper

Two more ways to put scores to work.

One page shows how to stream scored events to your own systems, and the other shows how scores catch churn early.

Scored recordMaya Chen · usr_48213
```
{ "event": "Order Completed", "scores": {
  "rfm": "champions", "likelihood": 0.82,
  "next_best": "rain_shell_jacket" } }
```
Streaming · 12 ms
Your Kafkaevents.scored.v1
Your S3meridian-lake
Job 01
Event Sync
Stream every event to your own Kafka and S3 with these scores already attached, so your own apps can use them too.
See how Event Sync works
Lina OrtizRegulars · 64 days quietChurn risk rising
Churn risk74%
Risk passed 60%
Save journeyLina Ortiz · At RiskStarted 2s ago
We miss youWait 2 daysSales call
Job 02
Customer Retention
Score churn risk from live behavior weeks before the cancel click, then start the save journey the moment it rises.
See how retention works

The products underneath

Each one reads the same customer profile

Frequently asked questions

Straight answers about predictive analytics for marketing, RFM analysis, propensity models and how our scores work.

Predictive analytics for marketing uses past customer behavior to guess what each person will do next, like buy, return or leave. It helps your team spend its time on the customers most likely to act.

Ready when you are

Start every day witha ranked list.

Connect your sources, and RFM, likelihood and next best product start working on every profile, with no model to build.

Connect your AI to your customer data

$ claude mcp add intempt -- npx -y @intempt-technologies/mcp

See who will buy next and what they’ll buy.

Your team shouldn’t guess who to contact.

Three scores that tell you who to contact.

Sort every customer into six clear buckets

Six things you stop doing by hand.

No data scientist to hire

No warehouse round-trip

Scores that never go stale

One list for every channel

Segments you can describe

Journeys already built