Territory Models Ranked: From Postcode Carve-Ups to Account-Tier Assignment
Six ways companies carve up territory, ranked from "recipe for a Slack war" to "actually defensible by Q3."
Ask five VPs of Sales how they carve territory and four will say some version of "geographically, mostly." Ask why, and the honest answer is rarely "because it produces the fairest split of pipeline." It's "because it's the easiest one to put in a board deck." A map with coloured regions looks like a plan. It photographs well. It is, in nearly every B2B org we've diagnosed, the least defensible model on this list — and also the most common, which tells you something about how territory decisions actually get made.
Here are six models we see in practice, ranked from the one most likely to start a Slack war by month two to the one still standing, largely unchallenged, by Q3.
The six, worst to best
| Rank | Model | Why it lands here |
|---|---|---|
| 1 (worst) | Geographic / postcode carve-up | Pipeline density has no real relationship to postal boundaries; parent-child accounts get split across reps; the turf war starts the day a big logo's HQ and its biggest subsidiary land in different patches |
| 2 | Round-robin lead rotation | Removes the geography problem and replaces it with pure luck — no account expertise builds, no rep develops a repeatable motion for a given buyer type |
| 3 | Revenue-tier assignment, no capacity check | Correctly targets account potential, then hands your best rep twelve "whale" accounts nobody could realistically service and calls it a reward |
| 4 | Industry / vertical specialisation | Builds genuine expertise and message-market fit; the risk is vertical-size imbalance if nobody rebalances as markets shift |
| 5 | Account-tier assignment by ICP fit | Ties territory to propensity-to-buy data rather than geography or size alone; only as good as the segmentation model behind it |
| 6 (best) | Account-tier + whitespace-balanced capacity, rebalanced quarterly | Combines propensity data with an honest read on how much a rep can actually work, and re-tests both against reality every quarter |
Why geography keeps winning anyway
Geographic carving survives for reasons that have nothing to do with whether it works. It's legible to a board in nine seconds — here's the map, here's the colour per rep, here's how we'll expand into the white bits next year. It requires no data model, just a postcode list and a decision about where the lines go. And it feels fair in the abstract, because everyone gets roughly the same land area, even when land area has zero correlation with buyer concentration, budget authority or deal readiness.
The cracks show up fast in any business selling into accounts with more than one location. Postcode boundaries don't know that a prospect's headquarters sits in Manchester and its biggest, most promising subsidiary sits in Leeds — so two reps end up with a legitimate claim to the same buying committee, and the first anyone hears about it is a Slack thread with three people cc'd who shouldn't need to be. Multiply that by every multi-site account in the book and you've built a structural source of internal conflict that has nothing to do with the market and everything to do with how the lines were drawn.
Round-robin rotation (rank two) fixes the fairness optics — everyone gets the same volume of leads in sequence — but optics is all it fixes. It actively prevents the thing that makes reps good at selling into a specific buyer type: repetition against a recognisable pattern. A rep who's worked twenty manufacturing accounts back to back gets faster and sharper at manufacturing. A rep getting one manufacturing lead, one healthcare lead and one retail lead in strict rotation never gets the chance.
Revenue-tier assignment (rank three) is where teams start doing something defensible: sizing territory to account potential rather than geography. It falls down on execution almost every time, because "potential" gets measured by contract value alone, with no read on how much service capacity that potential actually demands. The rep who inherits the twelve largest logos in the book isn't being rewarded — they're being handed twelve renewal-risk relationships to manage on top of a full new-business number, with the same number of hours in the week as everyone else.
What "defensible by Q3" actually means
Industry specialisation (rank four) and account-tier-by-ICP-fit (rank five) both start from a better question — not "how do we split the map" but "which accounts is this specific rep best positioned to win" — and that's the question that survives scrutiny six months in. The difference between four and five is data discipline: vertical specialisation can be assigned by instinct, but propensity-to-buy tiering needs an actual model behind it, something closer to an Account Segmentation & Tiering Rubric than a shared spreadsheet someone built during a planning offsite two years ago.
The model at rank six earns its place by doing something the other five don't: checking itself. Account-tier assignment gets you the right accounts against the right reps on day one. Whitespace-balanced capacity, rebalanced quarterly, is what stops that fairness from decaying over the following three quarters as pipeline shifts, accounts churn and one rep's patch quietly becomes twice the work of another's. Run the split through a Territory Design Fairness Scorecard at each rebalance, and have every rep able to answer "is my number actually winnable this quarter" honestly — a straightforward "Is My Quota Actually Winnable?" Self-Assessment does more for territory trust than any all-hands announcement ever will.
Defensible doesn't mean every rep loves their patch. It means that when a rep looks at their own territory next to a colleague's, they can't immediately point to an obvious double standard — and that the person who built the split can explain, in one sentence, why each account sits where it sits. Geographic carving fails that test by month two. The hybrid model at rank six is usually still passing it by the time the board asks for next year's plan.