RFM analysis for a shop: find loyal customers and those drifting away

What RFM measures

• Recency: how many days have passed since the customer's last purchase. Fewer days is better.
• Frequency: how many purchases the customer made in the period, for example the last 12 months.
• Monetary value: how much the customer spent in the same period, after returns.

Recency is usually the strongest signal. A customer who bought last week is far more likely to buy again than one who last came six months ago, even if the second one spent more in total. Frequency shows habit, and monetary value shows how much a customer matters to revenue.

What you need before you start

RFM works only with purchases you can link to a person: a loyalty card, a phone number given at the till or a customer account in the online store. Anonymous receipts cannot be scored, so check first what share of your sales is identified. If it is small, the analysis describes your loyalty members, not all your customers, and it should be read that way. Using customer data for marketing may also require consent under data protection law, so check the rules that apply to you.

How to do RFM analysis, step by step

1. Decide which purchases count: identified receipts from all stores and the online store, without test and staff purchases.
2. Choose the period, usually the last 12 months, and a reference date, usually today.
3. Export for each customer the date of the last purchase, the number of purchases and the total spend in the period.
4. Count several receipts from the same customer on the same day as one visit, so split payments do not inflate frequency.
5. Calculate recency as the number of days between the last purchase and the reference date.
6. Score each measure from 1 to 5 by splitting customers into five equal groups; for recency, the most recent group gets 5.
7. Combine the scores into a small number of segments you can act on.
8. Assign one clear action to each segment and a person responsible for it.
9. Repeat the analysis every month or quarter and watch how customers move between segments.

Segments that are easy to act on

• Champions (recency 5, frequency 4–5): your best customers. Thank them, give early access to new products, do not flood them with discounts.
• Loyal (recency 3–4, frequency 4–5): regulars. Keep them engaged with relevant news and loyalty benefits.
• New (recency 5, frequency 1): made a first purchase recently. The goal is a second visit.
• At risk (recency 1–2, frequency 3–5): used to buy often, have not come for a while. The most valuable segment to win back.
• Lost (recency 1, frequency 1–2): bought once or twice long ago. Low priority; include them only in cheap, general campaigns.

Five segments are enough for most shops. A grid of 125 score combinations looks precise but is impossible to act on.

Worked example (illustrative numbers)

A shop with three stores has 2,000 customers with identified purchases in the last 12 months. Split into five equal groups, each score covers 400 customers. Three of them:

• Customer A: last purchase 6 days ago, 14 visits, 820 EUR. Scores R5 F5 M5, a champion.
• Customer B: last purchase 95 days ago, 9 visits, 610 EUR. Scores R2 F4 M4, at risk. Until spring B came almost every month.
• Customer C: last purchase 12 days ago, 1 visit, 35 EUR. Scores R5 F1 M1, new.

Suppose 140 customers fall into the at risk segment. Together they spent far more in the period than the 300 lost customers, so a personal message or an offer for them is worth more than a general campaign to everyone. After the next monthly run the shop checks how many of the 140 have moved back to loyal, and how many new customers made a second visit.

Adjust recency to your purchase cycle

Thirty days without a purchase means little in a furniture shop and a lot in a bakery. Equal groups adapt to your data automatically, but look at the actual day ranges behind each score. If R1 starts at 40 days in a shop where people normally come weekly, those customers are truly drifting away; in a shop with a yearly purchase cycle, a longer period may be needed before the analysis says anything.

How it works in OrgLines

In OrgLines retail analytics are based on a copy of your loyalty programme's operations: receipts, members, points redeemed and the share of receipts with redemption, per store and per seller. Customer analytics place members in a recency and frequency grid. Click a segment to see only those customers, save it and export it as a file for the mailing tool you already use. There is no built-in mailing.

The grid is honest about its limits. When there are too few customers to rank, it is not drawn at all, because it would be an artefact of the sample rather than a picture of your base, and customers who could not be compared are shown as a separate count instead of being hidden. The grid uses recency and frequency; if you also want monetary value, add it in your own calculation on the exported segment or your sales data.

Common mistakes

• Mixing anonymous and identified sales, so the base looks smaller or larger than it is.
• Counting split payments as separate visits.
• Using a period that is too short, so seasonal customers look lost.
• Building dozens of segments and acting on none of them.
• Sending the same discount to every segment, including champions who would have bought anyway.
• Running the analysis once instead of tracking movement between segments.
• Ignoring consent rules when exporting customers for campaigns.

Checklist

• The share of identified sales is known.
• The period and reference date are written down.
• Same-day receipts are merged into one visit.
• Scores are calculated the same way every time.
• Five or fewer segments, each with one action and an owner.
• Exports for campaigns respect customers' consent.
• The next run is in the calendar, and segment movement is compared.

Try it free in OrgLines