Store Customer Segmentation: How to Build Groups and Choose Offers

To segment store customers, decide what action you want to take: whom to offer a complementary product, whom to notify about new stock, or for whom to test a repeat purchase offer. Then create groups using available data about needs and purchases, and set a specific action and outcome measure for each group.

Start with one product category, an agreed period and a small trial. Separate sales to known shoppers from anonymous receipts, check customer identifiers and the legal grounds for using the data. Assess the result by repeat purchases and financial contribution after returns, cost of goods and offer expenses.

Decide what action the segment should guide

A customer segment is a group of shoppers that a store brings together under a defined rule for a specific task. For example, the rule “bought replacement cartridges but have not returned for a long time” can help prepare an offer for relevant consumables. This decision requires a history of purchased products and purchase dates.

We recommend writing down a short brief before selecting customers:

• Decision: which offer and for whom you are testing it.
• Scope: product category, one store or the entire chain, sales channel and observation period.
• Rule: which specific purchases or stated needs include a customer in the group.
• Exclusions: a returned main product, no permission for the relevant message, recent contact or unreliable identification.
• Outcome: what change you are measuring, over what period and with what costs.

For the first launch, choose a few groups for which the store can already prepare different, relevant offers. If you plan to use the same product, text and service approach for two groups, consider whether splitting them is useful.

Prepare data so shoppers are identified correctly

We recommend creating a working table with a minimum set of fields: internal customer identifier, receipt number and date, store, purchase channel, category or product, amount actually paid, discount and any related return. You will need cost of goods to assess the offer economics. Keep contact details and permitted messaging channels where only authorized employees can access them.

Align the period and stores

Consistent rules must use the same period. If you use one month of purchases for one store and the entire operating history for another, comparing purchase frequency will be inaccurate. Separately note when new stores opened, any operating interruptions and periods when the required product was unavailable. Account for these circumstances before concluding that a customer has lost interest.

You can test a shorter observation window for products that are replenished regularly. For seasonal or durable goods, choose a period that covers the relevant need cycle. Record the date on which you create the segments so that later purchases do not change group membership during the assessment.

Distinguish people, cards and anonymous receipts

• Known shoppers. Use a consistent internal identifier across all the stores and channels included in the analysis. Before combining records, check that they really belong to the same customer.
• Anonymous sales. Keep these separate for work on assortment and in-store offers. Without a reliable link to a shopper, these receipts cannot establish their repeat purchases.
• Shared family cards. Mark them as shared. The history of such a card reflects the combined purchases of its users, so a personal message about one person’s supposed preferences may be inappropriate.
• Duplicates. Review matches manually. The same phone number or surname alone does not prove that records belong to one person.

Calculate the share of sales that can be reliably linked to customer records. If a significant portion of receipts remains anonymous, assess segments as groups among identified shoppers. For unknown visitors, you can test product displays, bundles and general offers without creating personal profiles.

Check that customer data is used lawfully

For a store processing shoppers’ personal data in Ukraine, the applicable law is the Law of Ukraine dated 01.06.2010 No. 2297-VI “On Personal Data Protection.” Under Article 6, the purpose of processing must be specific and lawful, and data must be accurate and not excessive in relation to that purpose. Data may be kept in a form that allows a person to be identified only for as long as necessary for the lawful purposes of processing.

Article 11 of this Law provides several grounds for processing, including consent, entering into and performing a transaction, compliance with a legal obligation, and protecting legitimate interests under the conditions set out in the article. The legitimate interest ground is limited: the need to protect a person’s fundamental rights and freedoms must not outweigh the relevant interest.

In practice, before segmentation, identify the purpose and legal ground for the specific planned use of the data. Having a phone number for order delivery does not by itself confirm permission to send advertising messages. If a new purpose is incompatible with the original one, Article 6 requires consent to processing for the changed purpose, unless otherwise provided by law.

We recommend organizing the work as follows:

• Explain to the shopper who uses the data, why the store collects it and how to contact the store about its processing.
• If you rely on consent, keep evidence of what it covered and that it was given, including for the selected messaging channel.
• Use purchased categories and needs stated directly by the shopper to select offers. Do not infer health status, beliefs, family circumstances or psychological traits from purchases.
• Set a schedule for reviewing segments and a data retention period consistent with the purpose and applicable obligations.
• Limit employee access and file sharing: an internal identifier and amounts without names or phone numbers are often enough to assess results.

Under Article 8, shoppers have, among other rights, the right to access their data, withdraw consent, object to processing on reasoned grounds, and request changes to or destruction of data processed unlawfully or that is inaccurate. A response confirming whether data is being processed and describing its contents must be provided no later than 30 calendar days from receipt of the request, except in cases provided for by law. Set up a process for receiving such requests and updating communication lists promptly.

Article 24 requires personal data to be protected against unlawful processing or access, accidental loss or destruction. A natural person who is an individual entrepreneur is personally responsible for protecting the data they hold.

Before sending advertising, separately check the specific rules for sending it through the chosen channel. The provisions of the Law “On Personal Data Protection” cited here are not enough for a complete legal assessment of advertising messages. As a practical approach for a pilot, use confirmed permission for the relevant communication and provide a clear way to decline future offers.

Choose criteria that help you prepare an offer

The table shows recommended approaches. These are examples to test in your own store; the expected result must be confirmed through a trial.

Distinguish the place of purchase, the location where the goods are received and the chosen messaging channel. An order through a website does not confirm a wish to receive advertising by email. Buying an expensive product also does not justify inferring a person’s income.

How to use RFM without complex setup

The RFM method, from the English words Recency, Frequency and Monetary, groups customers by how long ago they last purchased, how often they purchased and the monetary value of their purchases. To start, you can calculate these measures in a working table, defining how receipts and returns are counted.

Illustrative example: a store analyzes purchases from the last 90 days. Only customers with at least one purchase during that period that was not fully returned are included. In this example, F is the number of such purchases, M is the amount actually paid after discounts minus related returns, and R is the number of days from the last such purchase to the date the groups are created.

These thresholds are purely illustrative. Before using them, review the actual purchase distribution and the category cycle. Keep customers with no purchases in the selected window, fully returned purchases or incomplete data in separate groups so they are not assigned arbitrary scores.

Use the following rules to select offers. Here, F and R refer to the scores in the table, not the number of purchases or days:

• New purchase: F = 1 and R = 3. Test a useful message about using the product or a compatible add-on.
• Regular purchases: F = 3 and R = 3. Test the convenience of repeat ordering or a message about availability in the needed category.
• Long break after several purchases: F = 2 or 3, R = 1. First check the category’s usual interval, then assess whether an offer to return is appropriate.

Use M to assess acceptable spending on service or an offer. Specify the product category for each group: the same scores do not mean the same needs. If a customer qualifies for several offers, set priorities and contact limits so they do not receive all the messages at once.

Clarify needs through brief conversations with shoppers

We recommend speaking with shoppers from different groups: after a first purchase, after several purchases and after a break. Ask what task they chose the product for, what made the choice difficult, what was missing from the assortment and which way of receiving the product was convenient. Phrase questions so the person can give their own reason.

Use the answers to form hypotheses and refine offers. A few conversations do not show how common a need is among all customers: people who agree to talk may mainly have a particular kind of experience. Likewise, no response or no purchase in the database does not explain someone’s behavior. Mark unverified assumptions and test them against purchases and pilot results.

Test the offer with a small group

1. Set the conditions. Choose one segment, one offer, available products, a budget and an assessment period. Write down what you will count as a repeat purchase.
2. Keep a comparison group. Randomly divide eligible participants into a group receiving the offer and a control group that does not receive this specific offer. Keep a shared family card in one group.
3. Check the list. Remove opt-outs, duplicates and unsuitable contacts. Check product availability and make sure the terms are clear to sales staff.
4. Count all participant purchases. Count shoppers and results in each group, including purchases made without a promo code. Track message delivery, opt-outs and complaints separately.
5. Wait for returns. Choose an observation period based on the category cycle and returns. Record later returns if they change the conclusion.

The main measures for this pilot are the share of participants who purchased and the financial contribution per participant. For a management calculation, subtract returns, the cost of goods that remained sold and additional costs for messages, gifts, packaging or delivery from the revenue actually received. If the discount is already reflected in the payment, do not subtract it again.

Illustrative numerical example: each group has 100 customers. In the offer group, 20 people purchased; in the control group, 16 did. The percentages of participants who purchased are 20% and 16%, respectively, a difference of 4 percentage points.

• Offer group: revenue after discounts of UAH 10,000 − returns of UAH 1,000 − cost of goods kept by shoppers of UAH 5,400 − additional expenses of UAH 600 = UAH 3,000.
• Control group: revenue after discounts of UAH 8,000 − returns of UAH 800 − cost of goods kept by shoppers of UAH 4,320 = UAH 2,880. There are no additional expenses in this example.

Contribution per participant is UAH 30 versus UAH 28.80. The difference is UAH 1.20 per participant, or UAH 120 for a group of 100 people. This is a contribution toward other expenses, not the store’s net profit: fixed costs are not included in the example. Such a small pilot provides a preliminary assessment; random differences between groups can affect the result.

Expand the offer after reviewing the financial result, returns and shopper responses. If there is little data, repeat the trial under comparable conditions. Change one significant element at a time, such as the product or delivery terms, to understand what caused the result to change.

How to create customer groups in Торгсофт (Torgsoft)

In the “Маркетинг → Клієнти” form (“Marketing → Customers”), you can filter customers by region, discount type, whether a card was issued and other available parameters, and sort them by the date of their last purchase. To filter by behavior, open “Аналіз → Аналіз активності покупців” (“Analysis → Buyer Activity Analysis”): set the activity period and accounting center, the minimum or maximum number of purchases, and whether a product was purchased.

If you leave the period and accounting center blank, the analysis covers the entire period of recordkeeping and the whole chain. The “Виручка реалізацій” (“Sales revenue”) metric in this report does not account for returns, so they must be accounted for separately when checking the offer result.

For grouping by interests, the official description of Torgsoft customer accounting software includes an additional paid CRM option, “CRM. Взаємодія з клієнтом” (“CRM. Customer Interaction”); CRM means customer relationship management. The described work with the additional CRM option requires a “Торгсофт Ультра” (“Torgsoft Ultra”) or “Торгсофт Термінальна версія” (“Torgsoft Terminal Version”) license. First configure interests and link product groups to them; then the “Заповнити інтереси за придбаннями” (“Fill in interests based on purchases”) action fills them based on previous purchases.

Use these labels as information about product preferences. For filtering by interests specifically in “Аналіз активності покупців,” the help documentation separately states that the “Профіль клієнта та дзвінки Binotel” (“Customer Profile and Binotel Calls”) option must be activated, so check whether the required filter is available in your configuration.

Adding selected customers to a mass mailing requires a separately activated mass messaging option. The official description also states that an active subscription to the TurboSMS service is required for the described mailings. Before sending, check the permitted channel, current consent and opt-outs; activating the feature itself does not grant permission to message. Save the selection rule and group membership to compare results, and calculate RFM thresholds in a working table using agreed rules.

Update groups after reviewing the result

For each segment, keep a short description: purpose, data sources, period, inclusion rule, exclusions, offer, responsible employee and next review date. Update group membership before a new launch, taking purchases, returns, data corrections and messaging opt-outs into account.

If a group does not help you choose a product, change a service or test a specific offer, review its criteria. Keep rules that employees can apply consistently and that allow the owner to assess benefits for shoppers and the store’s financial result.