Segmentation sounds like a slide, but it is really an operating decision: who gets what motion, who gets which playbook, and which signals you measure against. This guide walks revops, marketing, and CS leaders through a 3-5 day sprint to cluster accounts, name the segments, wire them into routing and playbooks, and instrument the metrics that prove the model works. The result is a shared vocabulary your GTM team can actually run on.
Time: 3-5 days
Segmentation is not one artifact. It is a choice between four related jobs: routing inbound marketing leads, tiering customer success coverage, grouping accounts by lifecycle stage, or isolating expansion plays. Each job wants different dimensions and different owners. Picking the purpose first keeps the project from drifting into a 40-tab spreadsheet nobody uses. Write a one-sentence problem statement, the decision the segments will drive, and the team that will act on them. If more than one purpose matters, run the sprint twice. One segmentation per decision is faster to ship and easier to defend than a universal model that tries to serve everyone.
Pick 6-10 dimensions, not 40. The best B2B models combine three families. Firmographic dimensions (industry, employee count, region, tech stack) describe who the account is. Behavioral dimensions (product usage frequency, feature depth, support tickets, campaign engagement) describe what they do. Value dimensions (ARR, expansion potential, strategic logo, renewal risk) describe what they are worth. Pulling from all three families keeps you from building a model that is only a size cut or only a usage cut. Write each candidate dimension down with the source system, the refresh cadence, and whether it is reliably populated for at least 80 percent of accounts. Drop anything below that bar.
Export your account list with the dimensions you kept and cluster them. For most B2B teams, a rules-based cut (small, mid, enterprise crossed with high, medium, low engagement) is enough for a first pass and takes an afternoon. Larger datasets with ten or more dimensions benefit from k-means or hierarchical clustering in a notebook. Either way, aim for 4-7 segments. Fewer than 4 and you are just doing size tiers. More than 7 and nobody will remember them. Validate the clusters by sanity-checking a handful of real accounts in each one. If the cluster labels do not match how your reps describe those accounts in Slack, the model is wrong and you need to revisit dimensions.
Give each segment a short, memorable name that describes the behavior, not just the size. "Enterprise retail with heavy admin usage" is forgettable. "Power retailers" is not. For each segment, write a one-page profile: the defining signals, representative accounts, approximate count and ARR share, what good looks like, and the known failure modes. Add the owner, the primary motion, and the handoff criteria to adjacent segments. This document is the contract between marketing, sales, and CS. If your sales ops lead cannot read it and route an account correctly without asking a question, rewrite it.
A segmentation that lives only in a slide is dead on arrival. Push the segment label onto the account record in your CRM and treat it as a first-class field. Then connect the segment to the three places it has to drive behavior: inbound lead routing, outbound campaign targeting, and CS playbook assignment. In Strkr, that means a workflow writes the segment onto every account, routing rules read the segment on new leads, and CS playbooks are templated per segment with required check-ins, QBR cadence, and expansion plays. Make sure every segment has at least one named playbook. If a segment has no motion, it is noise.
Pick the two or three metrics that prove each segment is behaving as the model predicts. For marketing routing segments, that is usually MQL-to-SQL conversion, speed-to-first-touch, and pipeline per segment. For CS tiering segments, it is time-to-value, product adoption depth, NRR, and gross retention. For lifecycle segments, it is stage velocity and churn-risk signals. Build a dashboard that shows these numbers side by side across segments so an anomaly in one segment is obvious at a glance. Set guardrail thresholds (for example, NRR below 95 percent in the expansion segment triggers a QBR review) so the segmentation drives action, not just reporting.
Market shifts, product launches, and new ICPs all pull the segmentation out of alignment. Put a 60-minute segment review on the quarterly calendar with marketing, sales, CS, and revops. Walk through the dashboard, flag any segment where the metrics drifted more than 15 percent off plan, and decide whether to re-run the clustering, redefine a segment, or retire one. Resist the urge to add new segments without retiring old ones. The model only works if it stays small enough to hold in memory. Capture the decisions in the segment profile doc so the audit trail is intact. Expect to retire at least one segment per year as the business evolves.
Four to seven is the sweet spot. Fewer than four and you are usually just doing size tiers. More than seven and the model stops fitting in anyone's head, routing becomes brittle, and reps revert to gut feel. If you think you need more, split the segmentation into two separate models (one for marketing routing, one for CS tiering) rather than growing a single model past seven buckets.
An ICP describes the single kind of account you most want to win. A segment is a bucket inside your actual customer base used to drive a specific motion. ICP is a target. Segments are an operating model. Many companies have one ICP but five segments, because real customer bases include non-ICP accounts and because different segments earn different playbooks even when they share the ICP profile.
Review quarterly, re-cluster annually, and re-architect the model only when the business changes materially (new product line, new market, major pricing change). Re-running the clustering too often causes churn in routing rules and playbooks. Reviewing too rarely lets drift accumulate until the segments no longer match reality.
They should share a common vocabulary but can run different cuts. Marketing routing often needs firmographic plus intent dimensions. CS tiering often needs value plus adoption dimensions. If you force one model on both teams, one of them ends up ignoring it. Share the account-level segment field, but let each team keep a secondary classification tied to their motion.
No. Most B2B companies get 80 percent of the value from a rules-based cut built in a spreadsheet in an afternoon. Clustering algorithms become useful when you have more than ten dimensions and more than a few thousand accounts. Start simple, prove the segments drive action, and only invest in a model-based approach once the manual cuts are clearly leaving value on the table.
Watch for three warning signs. First, the segment has fewer accounts each quarter and no clear replenishment path. Second, routing rules for the segment produce the same action as an adjacent segment more than 80 percent of the time. Third, no team has used the segment-specific playbook in a quarter. Any one of those is a candidate for merging or retiring at the next quarterly review.
Strkr unifies CRM, marketing, and customer success so segment labels live on the account record and drive routing, playbooks, and dashboards from day one.
No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.
