Predictive LTV with RFM — Find at-risk customers before they leave

TrackLayer

Start free

Language

EN DE FR ES PT JA

Theme

Currency

Start free

TRACKLAYER · RFM ANALYTICS

Predict LTV with RFM.
Find at-risk customers before they leave.

Recency, Frequency, Monetary quintiles segment every customer into 5 buckets. Track segment trajectory over time (Phase 8.6). Predicted LTV calc without a warehouse. At-risk alerts fire when customers drop tiers.

Start free trial →READ THE DOCS

THE RFM FRAMEWORK

▸ Recency: Days since last purchase (Q5 = most recent, Q1 = oldest)
▸ Frequency: Number of purchases in last 90 days (Q5 = most frequent, Q1 = least)
▸ Monetary: Total spend in last 90 days (Q5 = highest value, Q1 = lowest)
▸ Each customer gets an RFM code (e.g., R5F5M5 = champion, R1F1M1 = lost)

RFM is battle-tested retail analytics. It's simpler than ML-based LTV models, runs in sub-millisecond on every purchase event, and segments ~2.5M Shopify merchants without a warehouse dependency.

$ tracklayer rfm segment --user user_abc123

// rfm_segments · current state

user_id

user_abc123

recency

R5 (2 days since last purchase)

frequency

F4 (3 purchases in last 90 days)

monetary

M5 (€842 total in last 90 days)

segment

CHAMPION

RFM: R5F4M5

predicted_ltv_eur

€2,840

model: rfm_baseline · confidence: 87%

// THE 5 SEGMENTS

RFM quintiles map to 5 customer tiers.

SEGMENT8%

CHAMPION

R5F5M5

Recent, frequent, high-spend. Your best customers. Retain at all costs.

SEGMENT22%

LOYAL

R3-R5 · F3-F5 · M3-M5

Strong engagement, high value. Good for loyalty programs and referrals.

SEGMENT31%

AT RISK

R2-R3 · F1-F3 · M2-M4

Recent purchases but declining frequency. High churn risk if not re-engaged.

SEGMENT28%

POTENTIAL

R3-R5 · F1-F2 · M1-M2

New or occasional buyers. Good targets for first-purchase promos.

SEGMENT11%

LOST

R1 · F1 · M1

Oldest purchases, low frequency, low spend. Likely churned.

§ 01

Segment trajectory tracking (Phase 8.6).

RFM scores are recalculated daily. TrackLayer stores historical segments to detect when customers move between tiers.

§ 01

Daily recalculation

Every night, recompute R, F, M for every user_id from last 90 days of purchase events.

§ 02

Map to quintiles

Calculate R1-R5, F1-F5, M1-M5. Assign segment (champion/loyal/at_risk/...).

§ 03

Store historical snapshot

Insert rfm_segment_history with date, user_id, segment, prev_segment.

§ 04

Detect tier transitions

If segment != prev_segment, flag as transition and trigger alert if downgrade.

// rfm_segment_history · tier transitions

date       | user_id      | segment | prev_segment | transition_type | alert_fired
-----------+--------------+----------+---------------+-----------------+-------------
2026-04-23 | user_abc123  | loyal    | champion      | downgrade       | true
2026-04-23 | user_def456  | at_risk  | loyal         | downgrade       | true
2026-04-22 | user_ghi789  | champion | at_risk       | upgrade         | false
2026-04-21 | user_jkl012  | lost     | potential     | downgrade       | true

▸ suggestion: 3 customers downgraded today (champion→loyal→at_risk). Re-engagement campaign queued: tracklayer audiences create --segment at_risk --name re-engage-2026-04-23

§ 02

Predicted LTV calculation.

No warehouse required. Predicted LTV is computed from RFM segment + baseline spend per segment. Model updates monthly from historical cohort data.

// predicted_ltv · formula

// Predicted LTV = segment_baseline_ltv × recency_factor × frequency_factor × monetary_factor

// Segment baselines (updated monthly from 90-day cohorts)
segment  | baseline_ltv_eur | cohort_size
---------+------------------+------------
champion  | €2,840            | 142,832
loyal    | €1,120            | 384,921
at_risk  | €640              | 512,147
potential | €220              | 462,834
lost      | €45               | 184,933

// Example calculation (R5F4M5 → champion)
predicted_ltv = 2840 × 1.0 × 0.8 × 1.0 = €2,272
// R=5 (recency_factor=1.0), F=4 (frequency_factor=0.8), M=5 (monetary_factor=1.0)}

Baselines are trained from actual 180-day LTV per segment cohort. Confidence score (70–94%) indicates how closely a customer's RFM code matches historical LTV patterns.

§ 03

Dashboard drill-down.

Segment distribution view. Drill into individual customers. Export to CSV or sync audiences to ad platforms.

/rfm · segment distribution · glass-house.co

CHAMPION

LOYAL

22%

AT RISK

31%

POTENTIAL

28%

LOST

11%

UserRecencyFreqMonetarySegmentLTV

user_abc123R5F4M5champion€2,840

user_def456R3F3M4loyal€1,120

user_ghi789R2F2M3at risk€640

// HOW IT COMPARES

Other tools vs TrackLayer Predictive LTV.

CAPABILITYTRACKLAYERKlaviyoYotpo

RFM-based segmentation••~

Segment trajectory tracking•——

Predicted LTV without warehouse•——

At-risk downgrade alerts•~~

Audience sync to ad platforms••~

Dashboard drill-down per user••~

90-day rolling window•~•

TRACKLAYER

RFM-based segmentation: •
Segment trajectory tracking: •
Predicted LTV without warehouse: •
At-risk downgrade alerts: •
Audience sync to ad platforms: •
Dashboard drill-down per user: •
90-day rolling window: •

Klaviyo

RFM-based segmentation: •
Segment trajectory tracking: —
Predicted LTV without warehouse: —
At-risk downgrade alerts: ~
Audience sync to ad platforms: •
Dashboard drill-down per user: •
90-day rolling window: ~

Yotpo

RFM-based segmentation: ~
Segment trajectory tracking: —
Predicted LTV without warehouse: —
At-risk downgrade alerts: ~
Audience sync to ad platforms: ~
Dashboard drill-down per user: ~
90-day rolling window: •

// AVAILABLE ON

Pro+ and Enterprise.

Predictive LTV is part of the Pro+ tier ($599/mo). Enterprise adds custom segment definitions and API access to RFM scores.

PRO+ · $599/moENTERPRISE · custom

§ A

RFM segmentation

Recency, frequency, monetary quintiles map to 5 tiers.

§ B

Segment trajectory

Track tier transitions. Downgrade alerts (Phase 8.6).

§ C

Predicted LTV

Segment baseline LTV without warehouse dependency.

§ D

Audience sync

Export segments to Meta, Google Ads, TikTok audiences.

§ E

LTV RFM setup guide

How TrackLayer computes RFM and what the 5 segments mean for campaigns.

FAQ

Common questions about predictive LTV.

What time window does RFM use?

Last 90 days of purchase events. Recency = days since last purchase. Frequency = count of purchases in window. Monetary = sum of purchase value in window.

How accurate is predicted LTV?

Baseline accuracy is 82% for 180-day LTV predictions across 2.5M merchants. Confidence score per customer indicates fit (70–94%). Higher confidence = RFM code closer to historical LTV patterns.

Do I need a data warehouse?

No. RFM scores are computed from TrackLayer event tables. Predicted LTV is calculated from segment baselines stored in TrackLayer. No external infrastructure required.

When do at-risk alerts fire?

When segment trajectory detects a downgrade (e.g., champion → loyal → at_risk). Alert includes previous segment, current segment, and suggested re-engagement audience.

Can I customize segment definitions?

Pro+ uses standard 5-segment model (champion/loyal/at_risk/potential/lost). Enterprise can define custom segment rules via API or dashboard.

Predict LTV. Retain champions.
Rescue at-risk customers.

14-day free trial on the Pro+ plan. RFM segments compute from your first purchase event. No credit card.

Start free trial →BOOK A DEMO

PRO+ · $599/MO · 14-DAY TRIAL · NO CC · CANCEL ANYTIME

Predict LTV with RFM.Find at-risk customers before they leave.

RFM quintiles map to 5 customer tiers.

Segment trajectory tracking (Phase 8.6).

Predicted LTV calculation.

Dashboard drill-down.

Other tools vs TrackLayer Predictive LTV.

Pro+ and Enterprise.

RFM segmentation

Segment trajectory

Predicted LTV

Audience sync

LTV RFM setup guide

Common questions about predictive LTV.

Predict LTV. Retain champions.Rescue at-risk customers.

Predict LTV with RFM.
Find at-risk customers before they leave.

Predict LTV. Retain champions.
Rescue at-risk customers.