RFM segmentation
RFM groups users by three measures: how recently they last transacted (recency), how often they transact (frequency) and how much they spend (monetary). Pushizo gives each user a 1–5 score on these measures, derives a ready-made segment from the scores…
RFM groups users by three measures: how recently they last transacted (recency), how often they transact (frequency) and how much they spend (monetary).
Pushizo gives each user a 1–5 score on these measures, derives a ready-made segment from the scores and shows how users move
between segments. Decision record: docs/adr/0062-rfm-segmentasyonu.md.
Setup
Open the RFM Analysis page in the dashboard and enter the following with Configure model:
| Field | Description | Default |
|---|---|---|
| Events | Events that count as purchases, up to 10 | Required |
| Window | Number of recent days to count, 30–730 | 180 |
| Use monetary value | When off, the segment is derived from recency and frequency only | On |
| Value property | Event property the value is read from | The event's value field |
The first calculation starts when you save. After that, scores are refreshed every day at 04:30 UTC.
Keep in mind:
- Only events from identified users are counted.
- Currencies are not converted; events should be sent in a single currency.
- If your event retention period is shorter than the window, only the retained events are used.
- Scores come from how users rank relative to each other. With fewer than 100 users the results may be misleading; the dashboard points this out.
Segments
| R \ FM | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| 5 | New customers | Potential loyalists | Potential loyalists | Champions | Champions |
| 4 | Promising | Potential loyalists | Potential loyalists | Champions | Champions |
| 3 | About to sleep | About to sleep | Need attention | Loyal customers | Loyal customers |
| 2 | Hibernating | Hibernating | At risk | At risk | At risk |
| 1 | Lost | Lost | At risk | Can't lose them | Can't lose them |
FM is the average of the frequency and monetary scores, rounded up. If monetary value is off, or everyone has the same value, FM equals the frequency score.
Names in the API: champions, loyal, potential, new, promising, need_attention, about_to_sleep, at_risk, cant_lose,
hibernating, lost.
Using it in audiences
You can create an audience segment with one click from a segment row on the page, or pick the RFM fields in the segment builder:
{ "all": [{ "field": "rfm.segment", "op": "in", "value": ["at_risk", "cant_lose"] }] }| Field | Operators | Value |
|---|---|---|
rfm.segment | eq, neq, in, not_in | Segment name |
rfm.recency, rfm.frequency, rfm.monetary | eq, gt, gte, lt, lte, between | 1–5 |
A user without a score only matches neq and not_in conditions.
Transitions
The transition table compares each user's segment at the start and at the end of the selected period. Out of scope means the user had no score on that date: they had no events within the window, or the first calculation had not run yet. History is kept for 400 days.
API
curl -X PUT "https://api.pushizo.com/v1/apps/$APP_ID/rfm/config" \
-H "Authorization: Bearer $PUSHIZO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"enabled":true,"event_names":["purchase"],"window_days":180}'| Endpoint | Scope | Description |
|---|---|---|
GET /v1/apps/{appId}/rfm/config | analytics:read | Model and calculation status |
PUT /v1/apps/{appId}/rfm/config | admin | Configures the model; starts a calculation if enabled |
POST /v1/apps/{appId}/rfm/compute | admin | Queues a calculation (202) |
GET /v1/apps/{appId}/rfm/overview | analytics:read | Segment distribution |
GET /v1/apps/{appId}/rfm/transitions | analytics:read | Transitions between the from and to dates |
GET /v1/apps/{appId}/rfm/users | users:read | Users, with segment, limit, cursor |
Privacy
When a user is deleted, their score and history are deleted too. When app data is purged, both tables are emptied; the model settings remain.
Analytics and reports
Delivery, subscription, event, conversion, funnel, provider health, engagement heatmap and retention reports; date ranges, comparison, CSV; the public analytics API and ClickHouse scaling.
Best channel
Calculates, from past deliveries, the channel each user responds to most. You use the result in two places: in audiences and in the Channel selection step of journeys.