Strategy
Customer Segmentation Fundamentals: A US Marketer's Guide to Getting It Right
How US marketers build customer segments: data sources, variable families, a five-step process, and validation checks before budget goes out the door.
What to take away
- Segmentation starts with a decision you can act on, not with a spreadsheet.
- Four variable families cover most US programs: firmographic, needs-based, behavioral and value-based.
- Data comes from what buyers tell you, what they do, and what you already hold in the CRM.
- B2B work needs account and buying-committee variables; B2C leans on behavior and life stage.
- A segment ships only after it passes size, stability, reachability and difference checks.
Write the decision before you touch data
Segmentation earns its cost when it changes a choice. Pick one: which accounts get a named rep, which customers see the new price, where next quarter's ad budget goes.
Vague goals produce vague clusters. "Understand our customers" is not a decision. "Decide which 200 accounts get onboarding calls" is.
If two groups would receive identical treatment, merge them. A segment that changes nothing is a report, not a strategy.
Where segmentation data comes from
Three source types cover most needs.
Three Segmentation Data Sources
Declared
- Examples
- Surveys, interviews
- Strongest for
- Needs, intent
- Trade-off
- Slow, expensive
Observed
- Examples
- Purchase history, CRM logs
- Strongest for
- Recency, frequency, value
- Trade-off
- Cheap, past-biased
Appended
- Examples
- Census, firmographics
- Strongest for
- Industry, territory fit
- Trade-off
- Quick, generic
| Source type | Examples | Strongest for |
|---|---|---|
| Declared | Surveys, interviews, win-loss calls | Needs, buying criteria, intent |
| Observed | Purchase history, web analytics, CRM activity logs | Recency, frequency, monetary value |
| Appended | Census tracts, firmographic files, NAICS codes | Firmographics, territory fit, industry |
The SBA guide to market research and competitive analysis explains how to study an industry and its buyers before committing budget.
Declared data is slow and expensive. Observed data is cheap but biased toward past behavior. Appended data is quick and generic. Most programs blend all three.
Choosing segmentation criteria that hold
Segmentation criteria are the tests a variable must pass. The standard set: measurable, sizable, reachable, stable, differentiable and actionable.
B2B and B2C segmentation differ mainly in the unit of analysis. A B2B segment is usually an account plus the committee inside it: economic buyer, technical evaluator, end user.
B2C segmentation leans on behavior, life stage and channel preference. Household income bands and ZIP-level data are widely used in the US.
Age by itself rarely separates buyers. Pair it with a behavior variable and it usually earns its place. The practical detail is set out in A working brief on marketing planning.
The customer segmentation process, step by step
- Name the decision and its owner. One person signs off on the final cut.
- Pull one clean extract from the CRM or transaction system. Remove duplicates before you group anything.
- Build candidate variables across all four families. Ten to twenty is usually enough.
- Size each group by revenue, margin and customer count.
- Test the top two or three groups with a small campaign before a full rollout.
A first pass usually takes about six weeks. Longer projects go stale before anyone uses them.
Five-Step Segmentation Process
- Name the decision and its owner
- Pull one clean CRM extract
- Build ten to twenty candidate variables
- Size groups by revenue and margin
- Test top groups with a small campaign
Validating a segment before you spend against it
Run this checklist on every candidate group.
- Sizeenough customers to matter at your average order value.
- Stabilitythe group holds together for at least two quarters.
- Reachabilityyou can contact them with tools you already own.
- Differencethe group responds differently from the rest of the file.
- Cost to servemargin survives whatever extra attention you promise.
Check the claims you aim at a segment
Advertising aimed at a segment still faces the truth-in-advertising and substantiation standard. The FTC advertising and marketing guidance explains what US businesses must prove before a claim runs.
Example: a regional equipment distributor
A distributor with 4,000 accounts sorts by revenue and finds a long tail. Half of those accounts order once and stop.
Adding two observed variables, order frequency and service ticket volume, produces three groups. Frequent, low-support accounts go to self-service. High-support accounts get a named rep. One-time buyers enter an email sequence.
The work took three weeks and one analyst. It changed call routing, not the product line.
Who does the work and what it costs
Most mid-size US firms hand segmentation to a marketing manager plus an analyst. The Bureau of Labor Statistics publishes wage data for advertising and promotions managers by metro area, which helps with budget planning.
Fundamentals carry a program a long way. Customer segmentation in 2027 adds decision-specific layers, need-state research and activation checks.
Common questions
How many segments should we build?
Three to six usually works for a mid-size program. More than that and sales or service teams cannot hold the differences in mind.
What separates market segmentation from customer segmentation?
Market segmentation divides a broad market before you sell into it. Customer segmentation divides the people and accounts you already have. Both use similar variables, but data access differs.
Do we need new software?
No. A CRM export and a spreadsheet handle a first pass. Buy tooling only after the segments survive a quarter.
How often should segments be refreshed?
Review them every six months and rebuild when the decision changes. Buying signals move faster than annual planning cycles.
