RFM scoring on autopilot:
recency, frequency, value, refreshed daily
Most teams run an RFM export once a quarter in a spreadsheet, segment customers by hand, and the list is stale before the next campaign goes out. We build a model that scores every customer on recency, frequency and value continuously, so a segment used in a campaign today reflects today's behaviour.
The process today
RFM, recency, frequency, value, is one of the oldest and most reliable ways to segment customers, and most teams who know about it still run it the same way they always have: pull a data export, open a spreadsheet, calculate three scores by hand, bucket customers into segments, and use that list for the next campaign. By the time the campaign sends, the data behind it can be weeks old, a customer who bought again last week still shows as “lapsed,” and a champion who has gone quiet for two months still shows as “champion” because nobody has re-run the export.
The manual version also tends to happen only when someone remembers to do it, often tied to a specific campaign push rather than running continuously, so the business operates most of the time without a current view of which customers are slipping from loyal to at-risk, a transition that is far easier to reverse early than after it is complete.
The cost compounds because the segments that matter most for retention spend, champions worth protecting, at-risk customers worth a win-back offer, are exactly the ones that change fastest and therefore go stale fastest in a quarterly process.
What the agent does
The model scores every customer on recency, frequency and value continuously, refreshed daily or weekly depending on your order volume, using segment definitions tuned to your actual customer base rather than a generic textbook cutoff. Segments like champions, loyal, at-risk, lapsed and new push directly into your CRM or email platform as live lists, so a campaign built against “at-risk customers” pulls whoever is currently in that bucket, not a list frozen at export time.
When a high-value customer’s recency score drops in a way that signals they are drifting from champion toward at-risk, a movement alert flags them specifically, giving a retention team a chance to reach out while a win-back offer still has a good chance of working, rather than after the customer has fully gone quiet. A dashboard tracks segment sizes and movement over time, so a marketing lead can see whether the champion segment is growing or shrinking as a trend, not just as a snapshot.
Before go-live, the model’s segment definitions are backtested against your last campaign’s actual results, so you can see whether the segments it would have produced line up with what performed well, rather than trusting a new definition blind.
What stays with humans
What offer goes to which segment, the creative, the discount depth, the channel, stays entirely with your marketing team. The model defines and refreshes who is in each segment; it does not decide what to say to them or when to send a campaign.
Guards
Every segment assignment is logged with the recency, frequency and value numbers behind it, so a marketer can see exactly why a customer landed in a given bucket. Segment definitions are reviewed against historical campaign performance before go-live, and a kill switch reverts to your last manually exported segment list in one message if something looks off.
Price and timeline
| Option | Price | What it covers | Timeline |
|---|---|---|---|
| Single automation | from $700 | Scoring, segment definitions, CRM or email push | 5 to 10 days |
| Department package | from $2,500 | RFM scoring plus automated win-back and loyalty campaigns | 2 to 4 weeks |
Running cost is usually $15 to $50 a month depending on customer base size.
Related
Pair this with win-back of inactive customers so the at-risk segment this model produces triggers a campaign automatically, and with product recommendations so champions see suggestions tuned to their specific buying pattern. For a broader clustering approach, see customer segmentation. The full package breakdown is on the AI agents service page and the automation-everything overview; for real retention results, see the Balkans supplements store case study and the Thailand D2C rebuild case study.
Ready to see your current champions and at-risk customers, not last quarter’s? Get in touch and we will look at your customer data in the first call.
Tired of doing this by hand? We can take the whole routine off your team, not just this step: Routine takeover, from $400 →
FAQ
How much does RFM scoring automation cost?
From $700 for scoring and segment definitions on your existing customer data, live in 5 to 10 days. A department package adding automated win-back and loyalty campaigns usually starts at $2,500.
Is this the same as customer segmentation?
RFM is a specific, well-understood segmentation method based on three behavioural measures. It is simpler and more transparent than a general clustering model, which is why many retention teams prefer it: every segment has an obvious, explainable definition.
How often does the score actually update?
Daily or weekly depending on your order volume. A high-frequency retailer benefits from daily refresh; a lower-volume B2B business is usually fine with weekly.
What happens when a top customer's behaviour changes?
A movement alert fires when a high-value customer's recency score drops meaningfully, flagging them as slipping toward at-risk while there is still time for a win-back touch, rather than after they have fully lapsed.
Where do the segments get used?
Pushed directly into your CRM or email platform as live segments or tags, so a campaign targeting 'champions' or 'at-risk' pulls the current list automatically instead of a spreadsheet someone exported weeks ago.