Syntha Lab RU

RFM segments from your customer table

Splits customers by recency, number of purchases and amount into segments, shows the share of customers and revenue in each and whom to win back first. Everything runs in the browser; the table goes nowhere.

Customers and loyalty For whom: Owners and marketers of brands and stores with a customer base.

Calculate on your data →

At a glance

What question does it answer?
Who holds the revenue, whom to win back first and which segments are not worth spend.
What data is needed?
An export of customers for one period: number, days since the last purchase, number of purchases, amount.
How to read the result?
First the share of customers giving 80% of revenue, then the “fading” segment: they had value and now do not. The table below says what to do with each segment. Score borders are illustrative: replace with your own. Example on the default values: Customers in the table — 20; Total amount — RUB 540,800; Share of customers giving 80% of revenue — 50%; Revenue share of “champions” — 40.7%.

Why and how to reason

Not all customers are equally valuable, and RFM is a way to see it using three simple signs: Recency (how long ago they bought), Frequency (how many times) and Monetary (how much). Each sign gets a score from 1 to 5 by the customer’s rank in the list (the best fifth gets 5), and recency is inverted: the fresher the purchase, the higher the score. Then the customer falls into a segment by recency and frequency: champion (recent and frequent), loyal, new (recent but few purchases), sleeping (long ago but used to buy often), lost. In the example, of 20 customers 50% give 80% of revenue (classic Pareto concentration would be about 20% of customers, so here revenue is spread more evenly), champions bring 40.7% of revenue, sleeping ones RUB 145,500 over the period, lost ones are 20%. The idea: different segments need different actions, and the cheapest win is bringing back those who bought often but went quiet.

When to use it

It is useful when you have a customer table with last purchase date, number of orders and amount, and need to decide whom to write to, whom to stimulate and where not to spend. You need at least 20–30 customers: with five the scores are arbitrary. It does not fit one-off purchases (furniture, a wedding dress) or seasonal customers with rare but regular purchases whom RFM mistakes for sleeping. Customer data is processed in the browser and goes nowhere.

How to read the result

Read it as a map of actions. Protect champions (early access, personal service), offer loyal ones wider categories, help new ones make a second purchase, win back the sleeping first: money was already invested in them and their RUB 145,500 sets the ceiling of recovery. Be careful with lost ones: they are costly to return and some will not come back. The share of customers giving 80% of revenue shows concentration: about 20% is the familiar Pareto concentration, 50% means revenue is spread evenly, 10% means dependence on a few. The sleeping revenue hints at a win-back budget: not more than a part of it. Define one action and one success metric for each segment (a letter, early access, a call) and a month later compare how the share of sleeping customers changed.

Common mistakes

Scoring a small table; mixing recency units (days and weeks in one table); calling a seasonal customer sleeping; treating scores as absolute (they are relative within your table, not market norms); sending everyone the same message. A segment is a hypothesis: confirm with a test. Do not include staff and test orders in the table: they distort the scores.

What data is needed?

  • An export of customers for one period: number, days since the last purchase, number of purchases, amount.
  • No names or phones: a customer number is enough.

Inputs

FieldUnitValue in the exampleMeaning
Target revenue share for the “core” of customers%80Shows what share of customers gives this share of revenue. Optional.

How it is calculated

StepFormula
Scores 1–5Customers are ranked on each measure and split into five equal groups; for recency the best score goes to the most recent
SegmentRecency 4–5 and frequency 4–5: champions; recency and frequency from 3: loyal; recency 4–5, frequency 1–2: new; recency 1–2, frequency from 3: fading; recency and frequency 1–2: lost
CoreCustomers by descending amount until the running amount reaches the target

Worked numbers

Default values from the app; in the interactive version you enter your own.

Customers in the table
20
Total amount
540,800 ₽
Share of customers giving 80% of revenue
50 %The smaller the share, the more revenue depends on a few customers.
Revenue share of “champions”
40.7 %
Revenue of the “fading” in the period (win-back priority)
145,500 ₽ · 26.9 %
Share of “lost” customers
20 %

Segments and what to do

SegmentCustomersCustomer shareAmountAmount shareAction
Champions420 %220,000 ₽40.7 %Keep: early access to new items, ask for reviews, no discounts.
Loyal315 %95,200 ₽17.6 %Grow frequency: bundles, recommendations, outfit ideas.
New and promising315 %30,500 ₽5.6 %Lead to a second purchase: a message after 2–4 weeks, matches to the first item.
Fading with past value420 %145,500 ₽26.9 %Win back first: a personal message and a reason to return; count the cost of winning back.
Watch list210 %17,700 ₽3.3 %A seasonal reminder, no spend above normal.
Lost420 %31,900 ₽5.9 %A cheap channel; judge whether winning back is worth it.

Whom to write to first among the “fading” (by amount): C-13, C-09, C-19, C-10.

Score borders and segment rules are illustrative: replace with your own. Keep customer data under personal data rules (Russian Federal Law 152-FZ) and check them with a lawyer.

Nuances and limits

  • For seasonal goods take a period equal to the season.
  • Do not mix wholesale and retail in one table.
  • Process customers’ personal data under the law (Federal Law 152-FZ) and check the procedure with a lawyer.

Terms used

Calculate on your data →