How to Segment Customers in Your CRM

Customer segmentation in a CRM works by combining three layers of data: firmographic (industry, headcount, revenue, region), behavioural (contacts in the last 90 days, product lines bought, payment habits) and value (annual revenue, gross margin, days to collect). Put those three in one table, write a single segment label on every record, and the segmentation is done. Keep the number of segments between four and six — beyond that, no sales rep remembers them, and a segment nobody remembers is a segment nobody uses.

Why does segmentation fail in most CRMs?

Because the segment field gets filled in once and never touched again. Eighteen months later, 12 of the 40 accounts tagged "Key Account" have not placed an order in two years. A segment is a photograph of the day the record was created; the customer moves on, the label stays put.

The second reason is subtler: the segment is just the industry name. "Construction", "Textiles", "Food" are filters, not segments. A segment defines a group you intend to treat differently. An account buying $400,000 a year and one buying $4,000 a year cannot sit in the same segment because they happen to share an industry code.

Which data actually builds a segment?

• Firmographic: industry, headcount, annual revenue, city, reseller vs. end user.
• Behavioural: last order date, order count over 12 months, average order size, quote acceptance rate, average days to pay.
• Value: trailing 12-month revenue, gross margin percentage, cost to serve (support tickets, site visits).

If you cannot measure one of these three layers, do not start segmenting — instrument that field first. In a list without margin data, "most valuable customer" collapses into "largest revenue", and that will point you the wrong way: your biggest buyer is often your deepest discount.

How many segments should you have?

For a B2B company with roughly 600 accounts, this structure holds up:

1. Key (A): the top 15-20% that produce 60-70% of revenue — usually 60 to 100 accounts.
2. Growth candidate (B): small spend today, but the same size and sector profile as your A accounts.
3. Steady (C): regular, small, low-effort volume.
4. At risk: late on payment, or no order in the last six months.
5. Dormant: inactive for more than 12 months.

Pareto is not a coincidence here: in most B2B companies 20% of customers generate close to 70% of revenue. The first job of segmentation is to make that 20% visible by name.

What are the steps to build it?

1. Export a 12-month account / revenue / margin table from the CRM.
2. Sort by revenue and add a cumulative percentage column; everything up to 70% is your A candidate list.
3. Add last order date and average days to pay, then split out At risk and Dormant.
4. Among the remainder, promote accounts matching your A profile by size and sector into B.
5. Create exactly one Segment field in the CRM — a picklist, never free text.
6. Write a rule that maintains the label: no order in six months moves to At risk, twelve months to Dormant.

What action belongs to each segment?

A segment that does not turn into a calendar is decoration. A working example:

• A: a planned conversation monthly, an executive-level review quarterly, a contract discussion annually.
• B: contact every six weeks, two new-product sessions a year.
• C: automated reminders and campaigns; no field visits.
• At risk: a call within 48 hours that covers collection and satisfaction together.
• Dormant: two win-back attempts a year, then out of the active list.

Here is the concrete payoff: monthly contact with 80 A accounts means roughly 20 conversations a week across a team of 12. That is a workload you can plan for — not the sentiment "we should pay more attention to important customers".

How do you measure whether segmentation worked?

Three numbers are enough: (1) the revenue share of accounts in A, (2) the number of accounts that moved from B to A — real growth shows up here, (3) average time spent in the At risk segment. If you are not reading these quarterly, the segmentation is a slide, not a system.

What are the most common mistakes?

• Making Segment a free-text field: 600 records will produce 40 spellings and every filter breaks.
• Never refreshing the label.
• Segmenting on revenue while ignoring margin.
• Not telling marketing: the same email lands in both A and Dormant.

How do sales and marketing share one definition?

When the two teams use different segment definitions, budget goes to the wrong place. The fix is a single agreed table: marketing builds campaign targets on the segment code, and sales reports back using the same code. The output of a campaign to segment B stops being "220 clicks" and becomes "14 conversations and 3 quotes from B".

The fastest way to enforce this is to make the CRM Segment field a mandatory filter when marketing pulls a list. A record with an empty segment cannot enter a campaign — that single rule measurably improves data quality within a few months.

How often should segments be refreshed?

• Monthly, automatic: At risk and Dormant labels recalculate from last order date.
• Quarterly, manual: the sales manager reviews the A and B lists, and the team discusses who entered and who left.
• Annually, structural: question the definitions themselves. If company revenue grew, the A threshold has to grow too — a fixed threshold will eventually put 300 accounts in A and make the label meaningless.

One caution: never demote an account silently. If you are going to stop the monthly visit, record it as a deliberate decision. Otherwise, six months later, the meeting asking "why did this customer drift away" will find nothing in the record.

The hard part of segmentation is not the rules — it is producing the data. What the customer said last time, why they did not buy, who actually decided: most CRMs never capture any of it. Closync extracts that from the conversations themselves and writes it into the CRM, so segments stay a current description of reality rather than a frozen label.