CRM and the customer base: retention, segments, win-back
How many customers are active and how many left, how cohorts hold, where value flows between segments, whether winning back lapsed customers pays off and what part of sales and of the base you can see at all.
About the topic
CRM (customer relationship management, the record of customers and their purchases) answers questions that a store without one answers by feel. How many customers do we really have and how many are active? What share returns? How much does a customer bring over time, and does acquiring one cost more than it brings? Whom should we win back, and will the campaign pay off? The topic teaches you to read the base so that a decision stands behind each figure.
The first thing to understand is that CRM does not see everything. Suppose store sales are RUB 100 million a year and receipts for 60 million are linked to customers with a phone or card. Coverage is 60%. If 70% of them consent to mailings, you can reach only 42% of sales. All conclusions about the base apply to that visible part. So you start with coverage: if it is 20%, conclusions about retention are conclusions about cardholders, not about all buyers.
Then cohorts: customers who made a first purchase in the same month. Example: 1,000 new customers arrived in January; three months later 250 bought at least once more, a year later 150. A retention curve of 25% and 15% shows how fast customers leave. If the curve drops sharply to zero, the customer is one-off and the first purchase matters more. If it levels out, you have a core of regulars, and growing it matters. The sum a cohort brings over its life (LTV, lifetime value) is compared with acquisition cost (CAC, customer acquisition cost): if LTV is RUB 7,000 of margin and acquisition is RUB 2,500, the customer pays back almost threefold.
Segments (for example RFM: recency, frequency, monetary value) show where you lose value: when your best customers drift into the dormant group, it is a signal before sales fall. A win-back campaign must pay off after counting those who would have returned alone. Example: 2,000 lapsed customers, a contact costs RUB 100, total RUB 200,000. 4% respond (80 people), but 1.5% (30) would have bought without the message, so the true uplift is 50 buyers. At an order margin of RUB 2,500 that is RUB 125,000 against RUB 200,000 of cost: the campaign loses money. To break even you need 80 incremental buyers.
Start with coverage and definitions (who counts as active and repeat), and fix them once. Traps: mixing new and returning customers, measuring response without a control group, counting revenue from those who would have bought anyway. Consents for mailings and data storage are set by law, so check with a lawyer.
Key ideas
CRM coverage limits every conclusion
Coverage changes on its own, and that breaks trends. If in March coverage rose from 40% to 55% because of a card-at-the-till promotion, “active customers” grew by 37.5% (55 / 40) even though demand may not have changed. Compare periods at the same coverage, or count shares within the database.
Do not mix cohorts
Cohorts must not be mixed: the average hides the weak one. In January 1,200 regular newcomers arrived and 300 returned (25%); in February a promotion brought 800 newcomers and 96 returned (12%). The combined 19.8% looks tolerable, yet the promotion cohort retains half as well. Look at each cohort separately and by acquisition channel.
LTV is compared with acquisition cost
The LTV to CAC ratio does not say when the money comes back. A customer with a CAC of RUB 2,500 brings RUB 900 of margin from the first order and then RUB 150 a month. The cost is recovered when 900 + 150 × n reaches 2,500, which is after 11 months. The acquisition money stays tied up for almost a year, and the budget must allow for it.
Segments catch churn early
RFM splits customers by recency, frequency and spend. When top customers have not bought for six months they are already at risk even though sales overall hold. Migration between segments shows where value flows.
Win-back is counted by uplift
The uplift threshold can be calculated before launch. A mailing to 5,000 lapsed customers at RUB 80 each costs RUB 400,000; with an order margin of RUB 2,400 you need a net uplift of 400,000 / 2,400, about 167 buyers, which is 3.3%. If past campaigns gave 2%, do not run it on everyone: narrow the segment to those with a higher uplift.
Interview, survey, A/B: the method is chosen by the question
The question “why do customers leave” is solved by interviews about the last purchase (JTBD): 8–15 conversations until saturation. The question “how many” is solved by a survey: for ±5 pp 444 answers, 3,700 invitations and RUB 53,280 are needed at 12% responding. The question “what causes” is solved by A/B on randomly split traffic. The result of an interview is hypotheses and not shares; the decision rule is written down before collecting.
What to do: Start with CRM data and interviews, then a survey of the base, then A/B; do not survey until you know the answer options.
Message frequency has an optimum
The return saturates while each message costs sending and unsubscribes: in the example the optimum is 4.76 messages a month, profit RUB 27.49 per subscriber against RUB 24.47 at eight.
Formula: f* = k × ln(ceiling ÷ (k × (cost + unsubscribe share × future margin)))
What to do: Check the frequency with a test on groups and count channels separately.
The reward for a friend is paid to legitimate ones, the benefit comes from incremental ones
In the example of 1,000 who came 880 are legitimate, 660 incremental, profit RUB 308,000, break-even reward RUB 1,350.
Formula: Break-even reward = incremental × lifetime margin ÷ legitimate
What to do: Pay after the first paid purchase.
How to work through it
- Determine CRM coverage: the share of sales linked to customers and the share of customers consenting to contact.
- Fix definitions: active, repeat, new and returning customer.
- Count the base for the period: how many active, repeat, new and returning.
- Split customers into cohorts and build a retention curve for 3 to 6 cohorts.
- Calculate LTV on margin and compare with acquisition cost.
- Segment the base (RFM) and look at migration between segments over two periods.
- Before a win-back campaign set aside a control group and calculate net uplift and the response threshold.
Where to start
First “CRM coverage”: what part of sales and of the base you see. Then “Customer base” for a period, “Cohort retention” over 3–6 cohorts, “Segment migration” and “Win-back” before a campaign. Calculations of the topic in order: “Customer base: active, repeat, new and returning”, “Cohort retention: curve, churn and cohort LTV”, “Customer segment migration: where the value flows”, “Win-back of lapsed customers: campaign payback and break-even response”, “CRM coverage: how many sales and customers you can see and reach”, “Bonus points liability and breakage”, “Survey sample size and margin of error: how many answers are needed and what can be read by segment”, “CRM message frequency: return against unsubscribes”, “Referral program payback”.
Calculations in this topic
Customer base: active, repeat, new and returning
Shows the health of the base over a period: how many customers came back, how many left, the share of new and repeat buyers and how the ticket and revenue per customer differ for new and returning ones. You set what “active” and “repeat” mean, so the figures are comparable only inside your company.
Is the customer base healthy: how many customers came back, how many left, whom revenue rests on.
Cohort retention: curve, churn and cohort LTV
For cohorts of customers (the month of the first purchase) it builds the retention curve: what share came back in the 1st, 2nd, 3rd and 4th period after the first purchase. It compares cohorts by LTV over a common horizon and shows which cohort brought more valuable customers.
How fast cohorts lose customers and which cohort brought more valuable customers.
Customer segment migration: where the value flows
For your segments (for example from RFM), ordered from the best to the worst, it counts how many customers moved up, stayed or moved down a step over a period and what it cost in money. Shows where the base loses value and what one step down costs.
Where the base loses value: from which segments customers move down and what it costs.
Win-back of lapsed customers: campaign payback and break-even response
Calculates whether a campaign to win back lapsed customers pays off: how many respond, how much margin is left after the discount, what part would have bought anyway and at what response the campaign breaks even. The discount and the response are your numbers.
Will a win-back campaign pay off and at what response does it break even.
CRM coverage: how many sales and customers you can see and reach
Shows what share of receipts and revenue is tied to a customer (card, phone), what part of the base consented to advertising messages and how much of it can actually be reached. Without this, conclusions about customers rest on a smaller part of sales.
What part of sales and of the base you see and can reach: what customer analytics stands on at all.
Bonus points liability and breakage
From points issued, redeemed and expired calculates the points balance, the obligation to members at face value, the historical breakage of a closed cohort, the reserve allowing for expected breakage, the cost of the program in revenue and the addition to the reserve already created.
How much the brand owes members in points, what part will expire and how much reserve to hold.
Survey sample size and margin of error: how many answers are needed and what can be read by segment
From the confidence level, the tolerable margin, the expected share, the size of the base, the effect of quotas and the response rate counts how many answers are needed, how many invitations to send and what it costs. For an already collected sample shows the margin of the overall estimate and of each segment: a segment several times smaller gives a wide interval.
How many answers are needed for the chosen margin, how many invitations to send, what it costs and what margin each segment of an already collected sample has.
CRM message frequency: return against unsubscribes
Finds the number of messages per month at which profit per subscriber is highest: the return grows ever more slowly with frequency and runs into a ceiling, while each message costs money to send and gives a share of unsubscribes, and an unsubscribed customer loses their future margin. Shows the optimal frequency, the profit per subscriber and for the whole base at the optimum and at your current frequency, and a table by frequency.
How many messages per month give the highest profit per subscriber taking unsubscribes into account.
Referral program payback
Counts whether the reward for an invited friend pays back: from all who came by the link it subtracts the fraudulent ones (self-referrals, fake), then those who would have come without a reward, and compares the margin of the rest (over the customer lifetime) with the rewards paid to all legitimate ones. Shows the profit, the return, the cost of an acquired customer and the break-even reward.
Whether the reward for an invited friend pays back and what it can be for the program not to become loss-making.
Self-check questions
- What share of sales and customers is visible in CRM, and how reachable is it?
- How do we define an active and a repeat customer?
- Are we mixing new and returning customers?
- How fast and to what level does cohort retention fall?
- What is LTV on margin and how many times does it exceed CAC?
- Where do top customers migrate and how many go dormant?
- What share of responders would have bought without the campaign?
- What response threshold is needed for the campaign to pay back?
- Do we have consent for mailings and a data storage procedure?
Knowledge check
Multiple-choice questions. The correct answer and the explanation follow each question.
A mailing to 10,000 lapsed customers, a contact costs RUB 5. 2% will respond, 1.5 orders per responder. Basket RUB 2,000 excl. VAT, margin before discount 50%, promotion discount 20%. 25% of responders would have bought without the letter anyway. What is the campaign result?
- A. RUB 105,000
- B. RUB 55,000
- C. RUB 130,000
- D. RUB 180,000
Correct answer: B. RUB 55,000. 200 responded, 300 orders. The promotion order margin is 2,000 × (0.5 − 0.2) = RUB 600, in total RUB 180,000. “Own” orders are 300 × 25% = 75 with a margin without discount of RUB 1,000: RUB 75,000, they are not credited to the campaign. The additional margin is RUB 105,000, costs are 10,000 × 5 = RUB 50,000, the result is RUB 55,000. RUB 180,000 is the margin without deductions, RUB 130,000 is without deducting the “own” ones, RUB 105,000 is before the contact costs.
At the start of the period 5,000,000 points, 3,000,000 accrued, 2,000,000 redeemed, 1,000,000 expired. For the closed cohort 2,000,000 were accrued, 1,200,000 redeemed, 800,000 expired. A point costs RUB 1 on redemption. A reserve of RUB 2,600,000 has already been created. How much more should be accrued?
- A. RUB 400,000
- B. RUB 1,000,000
- C. RUB 2,400,000
- D. RUB 3,000,000
Correct answer: A. RUB 400,000. The closing balance is 5,000,000 + 3,000,000 − 2,000,000 − 1,000,000 = 5,000,000 points, RUB 5,000,000 at face value. The cohort breakage is 800,000 / 2,000,000 = 40%, the reserve is 5,000,000 × 0.6 = RUB 3,000,000. The top-up is 3,000,000 − 2,600,000 = RUB 400,000. RUB 2,400,000 is the face value minus the created reserve without the breakage adjustment, RUB 3,000,000 is the whole reserve, RUB 1,000,000 is the points that expired in the period. Check the reserve accounting with an accountant.
It is needed to find out why customers return dresses. Which method is suitable first?
- A. An A/B test of the price
- B. A survey of 3,000 people without preliminary conversations
- C. Interviews about the last return, then a survey for the shares of reasons
- D. Only sales data
Correct answer: C. Interviews about the last return, then a survey for the shares of reasons. The question “why” is solved by interviews: they give reasons that are then measured by a survey. A survey without conversations asks about invented options, A/B checks a reason and does not look for it. Covered in: Interview, survey, A/B: the method is chosen by the question.
For an error of ±5 pp 444 answers are needed, 12% respond. How many invitations should be sent?
- A. 444
- B. 5,328
- C. 3,700
- D. 1,850
Correct answer: C. 3,700. 444 ÷ 0.12 = 3,700. Sending 444 invitations means getting about 53 answers. Covered in: Interview, survey, A/B: the method is chosen by the question.
One answer costs RUB 120, 444 answers are needed. What is the survey budget?
- A. RUB 53,280
- B. RUB 444,000
- C. RUB 1,440
- D. RUB 26,640
Correct answer: A. RUB 53,280. 444 × 120 = RUB 53,280: the price is for an answer, not for an invitation. Covered in: Interview, survey, A/B: the method is chosen by the question.
What is lost with each unsubscribe?
- A. The future margin of the subscriber
- B. Only the sending cost
- C. Nothing
- D. The whole brand
Correct answer: A. The future margin of the subscriber. It is counted in the loss per message. Covered in: Message frequency has an optimum.
What is the break-even reward of a referral program?
- A. The revenue of all who came
- B. The cost of a registration
- C. The average check
- D. The margin of incremental ones divided by the number of legitimate ones
Correct answer: D. The margin of incremental ones divided by the number of legitimate ones. The reward is paid to all legitimate ones. Covered in: The reward for a friend is paid to legitimate ones, the benefit comes from incremental ones.
Common mistakes
- Calculating retention without coverage. Consequence: conclusions about cardholders are taken for all buyers. Avoid: show coverage next to every figure.
- Calculating LTV on revenue. Consequence: acquisition payback is overstated. Avoid: use margin net of service costs.
- Judging a campaign by responses. Consequence: paying for those who would have returned. Avoid: a control group without the mailing.
- Changing the definition of active from report to report. Consequence: false dynamics. Avoid: fix once and write it down.
- Mailing without consent. Consequence: fines and complaints. Avoid: store consents and provide an opt-out.
Counting only opens
What it looks like (illustrative figures): An open is not money.
Paying for a registration
What it looks like (illustrative figures): Self-referrals get the reward.
Russia and the EAEU
In Russia customers' personal data is regulated by Federal Law 152-FZ on Personal Data. For CRM this means: consent for processing, a stated purpose, a storage period, notifying the personal data regulator, and storing data of Russian citizens on servers in Russia (initial recording and updates). Mailings by phone number and e-mail are governed by the advertising law 38-FZ: prior consent of the recipient is required and unsubscribing must be simple. Consents are best stored with a date, a source and the text.
Loyalty programs: accumulated points are a liability of the brand to the customer, and its accounting and tax consequences depend on the scheme (discount or payment by points, expiry period). Check with your accountant and lawyer. Use a till that transmits the buyer’s contact in the receipt if you need to link purchases to customers: a receipt under 54-FZ can carry a phone number or e-mail.
Linking sales to a customer in Russia works through loyalty cards, a phone number at the till, a personal account and apps. On marketplaces the customer belongs to the platform and the brand gets limited data, so CRM sees little of marketplace sales: take this into account when assessing coverage. Do not use retention and LTV figures in Russia as norms: every company has its own base.
CRM calculations are the same; rules on personal data and mailings differ. Guide, please verify: Belarus, Kazakhstan, Armenia and Kyrgyzstan each have their own personal data protection laws, with their own requirements for consent, storage and transfer of data abroad. Some countries require data of citizens to be stored on national territory. A customer base must not be transferred between legal entities of different countries without checking. Payment habits and communication channels differ (messengers, apps, local marketplaces), so calculate CRM coverage per country separately.
Categories and audiences
- Apparel
- If it is apparel: purchase frequency is moderate and mostly seasonal, so it is more convenient to define active as 12 months rather than 3. Measure retention by season. Size returns distort LTV, so count it after returns.
- Footwear
- On footwear: bought less often and at a higher price, repeat after a year or more, so the activity window is longer and the cohort slower. Define the leaving segment by the wear life of the model. Size returns are high, so count LTV after returns.
- Accessories
- In accessories: often a gift, so buyer and user differ and repeat rate is lower. But the check is small and the purchase quick. For CRM occasions (holidays) and a phone number at the till matter, otherwise coverage is very low.
- Kids
- For children’s wear: the customer is the parent, purchases are tied to the child’s age and size, so the next purchase can be predicted from growth and season. A strong segment is families with several children. Store data about children with care, minimally and with consent.
- Men
- Men buy less often but purposefully and respond less to discount mailings. Reminders of the season change and sets work better. Retention is lower than for women, so the first purchase and the convenience of a repeat matter more.
- Women
- Women buy more often and across several categories, and respond more to mailings and private pre-orders. The retention curve is usually higher and segments more numerous. Watch for mailing fatigue: frequency of communication should be measured, not maximal.
Questions this topic answers
- What part of the customers came back and what part left?
- How fast do cohorts lose customers and which cohort is worth more?
- Where does value flow between segments?
- Will a campaign to win back lapsed customers pay off?
- What part of sales and of the base do we see at all?
Key measures and terms
- Customer retention and churn
- Cohort and cohort retention
- Repeat customer share
- RFM
- CAC payback and LTV/CAC
- Self-selection
- Cohort LTV over a common horizon
- Segment migration
- Win-back
- CRM coverage (identification and reachability)
- Customer base report (CRM dashboard)
- Customer lifecycle
- Churn
- CRM communications and trigger campaigns
- Customer data quality and duplicates
Study first
What next
Related: “Advertising and customer acquisition”, “Marketing budget and testing impact”, “Store sales and receipts: traffic, conversion, ticket”.
Documents on this topic
- CRM analytics: RFM, cohorts, retention, LTV/CAC, loyalty
- Circular models: resale, upcycling, repair, rental
- Customer service: SLA, FCR, claims and consumer rights
- Voice of Customer: return and complaint reasons → actions
- Unit economics with buyout: CM1-CM3, LTV/CAC, customer survival, channel payback
- Cohort analysis: retention table, curve, LTV forecast and RFM as mathematics
- Marketing channels in Russia: Telegram, VK, MAX, Dzen, Rutube, Yandex Direct, bloggers and placements, content plan and metrics
- CRM chains and the loyalty program: triggers, frequency, consents, tiers, points lifetime and payback by cohort
- Customers’ personal data: Federal Law 152-FZ for a brand and a retailer. Grounds and consents, storage, localization, loyalty, mailings, liability
- Customer behavior and research methods: from a hypothesis to a decision (interviews, surveys, A/B, JTBD, sample size)
- Measuring marketing: path attribution, advertising carry-over, the budget optimum, the price test, NPS and the referral reward
- Site, CRM and receipts: abandoned carts, the delivery threshold, payments, message frequency, the lapsed threshold, day indices and pairs of items
- Market research: TAM/SAM/SOM, CustDev, UX tests and benchmarking