GUIDE · CUSTOMER SEGMENTATION 16 October 2026 · Completed example and blank manual form Customer segmentation groups customers by shared characteristics for a particular decision. A company can belong to two research segments when it meets two independent rules. Adding segment sizes can double-count that company: count each company once in the combined list. If reporting or work allocation needs one primary group, define a priority rule without deleting the other attributes. Keep missing evidence separate from a confirmed failure to meet a rule. Define the segment around a decision. “Furniture companies” is a starting market. But if you offer two different services, the same opening question may not suit every company. A business accepting made-to-measure projects may need a different conversation from one presenting products to export buyers. Segmentation also applies to existing customers. When a company has not bought from you, call it a prospect or research candidate. Do not assign a purchase-history value segment to a company with no purchase history in your business. Match the grouping method to the evidence you have Approach | Evidence you could use | Limit Company characteristics | Product category, service region and commercial role. | Sharing an industry does not establish a shared need. Needs | A specific difficulty described in a conversation. | A website feature does not establish a problem or budget. Behavior | A quotation, order or conversation in your own relationship. | One contact’s activity does not establish a company-wide buying decision. Value | Actual revenue and service cost within a defined period. | A new prospect’s value remains unknown without the relevant history. [OpenStax’s B2B chapter](https://openstax.org/books/principles-marketing/pages/5-2-segmentation-of-b2b-markets) explains different segmentation approaches. This guide uses inspectable rules for a small company list, rather than building a clustering algorithm. First define the companies you can serve in your [ideal customer profile](https://anivoapp.com/guides/ideal-customer-profile). Write the group name, conditions and unknowns together. Choose the unit first. One record represents one company in this example. Do not add contacts, branches or orders to the company count. Resolve uncertain company identities using the [company-record decision guide](https://anivoapp.com/guides/companies-with-same-domain). A fictional web agency is researching furniture businesses in Istanbul for two possible conversations: reviewing the quotation-form journey and reviewing English product information. Membership describes observed attributes. It does not establish a service need or a purchase approval. Fictional rule card dated 16 October 2026 · version v1 Group | Inclusion rule | Work to investigate A · Projects and quotation form | A statement offering made-to-measure projects AND an available quotation-request form. | Ask whether the form collects the right project information. B · Export catalogue | An English product catalogue AND a company statement offering export supply. | Prepare a question about how international buyers obtain product information. The population is only the four companies in this exercise. Sources and statements are fictional. This is not a measure of Istanbul’s market size or demand. **Yes:** Dated evidence supports every required condition. **No:** Evidence establishes that at least one required condition is not met. This does not make the company unsuitable for every offer. **Unknown:** Missing evidence could change the decision. An inaccessible page is not proof that an attribute is absent. Record the company ID, rule version, source URL or document ID, review date and a short reason. [Shopify’s official reference](https://help.shopify.com/en/manual/customers/customer-segmentation/reference-guide/components) distinguishes AND, which requires all conditions, from OR, which accepts at least one. Make that distinction explicit in your own record. Put separate groups of conditions in parentheses instead of assuming how a tool will combine them. Keep both attributes when the company meets both rules. Mavi Raf, Ada Ofis, Kuzey Mobilya and Delta Raf are entirely fictional. F01–F04 are stable company IDs for this exercise. These are not real businesses, customers or Anivo results. All evidence belongs to the same fictional review snapshot dated 16 October 2026. v1 membership ledger · one row represents one company Company / ID | Recorded evidence and fictional source | A / B | Next check Mavi Raf · F01 | MR-1: made-to-measure projects and a quotation form. MR-2: the company explicitly states that it sells only within Turkey. | Yes / No | Ask which project details the form fails to collect. B’s export condition is not met. Ada Ofis · F02 | AO-1: the company states it sells standard stocked items only and does not accept made-to-measure projects. AO-2: English catalogue and export-supply statement. | No / Yes | Ask how requests for English product information are handled. Kuzey Mobilya · F03 | KM-1: made-to-measure projects and quotation form. KM-2: English catalogue and export statement. | Yes / Yes | Retain both attributes. Establish which task matters first before starting two separate initial approaches to the same company. Delta Raf · F04 | DR-1: the website cannot be accessed during review; no verified evidence for the other conditions. | Unknown / Unknown | Hold the evidence check. Do not mark the company as a confirmed member or a confirmed rejection in either group. Source IDs are not real URLs. They show how to keep the example traceable. In your own record, enter the actual address, relevant statement and access date. [KOBİTEK’s original practical example](https://kobitek.com/musteri_segmentasyonu_nasil_yapilir) also explains that one type of business can appear in two groups. Overlap is therefore not automatically a classification error. First check what decision each grouping supports. Missing evidence is not negative evidence. F01’s “No” follows an explicit statement that contradicts the export condition. F04’s inaccessible website supplies no such statement. Turning both into “No” would erase a useful distinction. Two segments of two companies may contain three companies. Here A = {F01, F03} and B = {F02, F03}. Only F03 appears in both lists. For two groups, the distinct combined count is the number in A plus the number in B minus the number in both: **2 + 2 − 1 = 3**. Different counts from the same fictional snapshot Count | Company IDs | Result Members of A | F01, F03 | 2 companies Members of B | F02, F03 | 2 companies A AND B · intersection | F03 | 1 company A OR B · distinct union | F01, F02, F03 | 3 companies Membership awaiting evidence | F04 | 1 company; excluded from confirmed membership totals There are 4 segment memberships, 3 distinct confirmed members and 4 companies in the reviewed population. Those numbers answer different questions. For three or more groups, combine all member IDs and count distinct company IDs. Do not extend the two-group shortcut by casually adding more segment totals. Multiple intersections can count the same company again. Membership is not a recipient list. Companies, contacts and messages are different counting units. Finding three distinct companies does not give permission to contact three people. Check recipients, prior approaches and stop records in the tool you use. Do not assume automatic deduplication. Use an explicit priority when reporting or tasks need one group. Mutually exclusive primary groups can be the right choice. A distribution totaling 100%, or a work queue with one first task per company, needs one primary assignment. That does not require deleting the company’s other research attributes. Write the decision you need to make. In this example, the decision is which possible web project to discuss first. It is not a permanent classification of the company’s only need. Define how overlap is resolved. Ask F03 which task takes priority. Until then, its primary assignment can remain awaiting a decision. If you set a priority for your own planning, label it as your choice rather than the buyer’s preference. Keep the primary assignment separate from attributes. F03 can still satisfy A and B while its first task addresses one topic. Avoiding duplicate work need not destroy the second piece of information. [Mailchimp’s documentation](https://mailchimp.com/help/getting-started-with-segments/) explicitly allows a contact to belong to multiple segments whose conditions they meet. That does not establish how every tool handles the same recipient across separate campaigns. This ledger is not a sending instruction. Do not report a rule change as a market change. Now evaluate the same fictional evidence for a different purpose. Version v2 adds a requirement to B: the English catalogue must include product dimensions. AO-2 contains them. The complete KM-2 catalogue was reviewed and contains none. A’s rule is unchanged. This is not a later observation of the market. Changing B’s rule on the same 16 October snapshot Company | B · v1 | B · v2 | Reason F01 | No | No | The export condition already fails. F02 | Yes | Yes | English catalogue, export statement and product dimensions are present. F03 | Yes | No | The reviewed catalogue lacks dimensions, so it fails the new rule. F04 | Unknown | Unknown | The source is still inaccessible; the new condition cannot be assessed either. B’s confirmed membership falls from 2 to 1. No change in the companies was observed: the rule changed. F03’s removal does not prove it stopped exporting or is unsuitable for the agency’s service. Keep the old membership instead of overwriting it. Retain the rule text, version, evidence date, decision and reason for change. For comparisons over time, align the population, IDs, evidence coverage and rule, and explicitly report anything that changed. Put groups in the research brief and decisions in company notes. Prepare a researchable brief. “Research furniture companies in Istanbul. For A, I want to review firms offering made-to-measure projects with a quotation form. For B, look for an English catalogue and an export-supply statement. A company may fit both; missing evidence should not become a definite rejection.” This is a fictional research request. Review and approve the plan. Read the research plan Anivo prepares. Check your offer, region and company types, request corrections where needed, then approve the search. Verify membership yourself. Compare the returned company information with your rule and check the sources and dates. A segment name in a research plan does not mean your v1/v2 membership ledger is being applied automatically. Record the next task. When tracking a reviewed company in CRM, keep the membership reason, unknowns and rule version in a note. Use a company-linked task to follow up the missing evidence or first conversation topic. The form is not an Anivo import format. This guide does not promise automatic customer clustering, a versioned segmentation engine or cross-campaign sending exclusions. Anivo is a local Windows workspace for one user; this decision ledger is a manual aid. Use the [lead qualification guide](https://anivoapp.com/guides/lead-qualification) before advancing a company and the [Anivo user guide](https://anivoapp.com/guides/anivo-user-guide) for product steps. Segment membership does not establish need, budget, authority or timing. Overlap is a methodological option, not a result guarantee. [Zheng and Tuzhilin’s WITS 2008 paper](https://repository.hkust.edu.hk/ir/bitstream/1783.1-7885/1/WITS200_1579.pdf) compares predictive models using overlapping segments on consumer-behavior and synthetic data. The full six-page paper was reviewed. It is not a B2B prospecting study, a test of this manual form or an Anivo sales experiment. The company names, source IDs, rules and numbers here are original fictional exercises. They do not measure market size, response improvements or Anivo performance. Test manual rules against your own task. If a proposed segment would not change the work, simplify the existing grouping instead of creating another. The source notes below identify the sections accessed and their settings. Other products’ filters, data connections and refresh behavior are not Anivo features. BLANK RULE CARD Decision and offer: [ ] Population, region and counting unit: [ ] Group name and rule version: [ ] Inclusion conditions; AND/OR and parentheses: [ ] Confirmed exclusion condition: [ ] Missing-evidence hold condition: [ ] Reviewer and review date: [ ] Different task triggered by membership: [ ] BLANK MEMBERSHIP RECORD · REPEAT FOR EACH COMPANY Stable company ID and name: [ ] Rule version and evidence date: [ ] Actual source URL or document ID: [ ] Relevant statement and review date: [ ] A decision: Yes / No / Unknown and reason: [ ] B decision: Yes / No / Unknown and reason: [ ] Primary assignment and reason if required: [ ] Buyer statement or internal planning preference: [ ] Missing evidence and one next task: [ ] Previous membership and reason for change: [ ] BLANK COUNT CHECK Count date, population, unit and rule version: [ ] Confirmed A member IDs: [ ] Confirmed B member IDs: [ ] Intersection IDs and count: [ ] Distinct union IDs and count: [ ] Pending IDs and count: [ ] Separate membership totals from distinct company totals: [ ] For three or more groups, distinct IDs across all members: [ ] Changes in population, evidence or rules since the comparison: [ ] Sources and access limits Checked on 16 October 2026. Methodological references and fictional applications are separate. These sources do not validate Anivo sales performance. OpenStax · Segmentation of B2B Markets https://openstax.org/books/principles-marketing/pages/5-2-segmentation-of-b2b-markets Principles of Marketing, 2023. Relevant B2B-method sections reviewed; a textbook, not an Anivo experiment. Mailchimp · Getting Started with Segments https://mailchimp.com/help/getting-started-with-segments/ Definitions, multiple membership and AND/OR sections reviewed. No precise publication day visible; describes its own product. Shopify · Components of a customer segment https://help.shopify.com/en/manual/customers/customer-segmentation/reference-guide/components Filters, operators, values and connectors reviewed. No precise publication day visible; not instructions for Anivo filters. Zheng and Tuzhilin · WITS 2008 https://repository.hkust.edu.hk/ir/bitstream/1783.1-7885/1/WITS200_1579.pdf Full six-page conference paper reviewed. Predictive models on consumer-behavior and synthetic data, not sales results or B2B validation. KOBİTEK · How to segment customers https://kobitek.com/musteri_segmentasyonu_nasil_yapilir Relevant sections of the original Turkish practical example reviewed. Multiple membership and actionable grouping; precise publication date unverified. Microsoft Learn · Segments overview https://learn.microsoft.com/en-us/dynamics365/customer-insights/data/segments-overview Membership, rules and logical grouping sections reviewed. Dynamics 365 capabilities are not transferred to Anivo.