The carve happens in a spreadsheet, usually in November. The columns are account name, state, employee count, last year's revenue, and current owner. Somebody sorts by revenue, divides the rows by the number of reps, then nudges accounts between columns until each one sums to roughly the same number. The file gets a version suffix, goes to the VP, comes back with two swaps, and becomes next year's territories.
Nothing about that process is lazy. It uses every piece of data that was actually available at the time. That is the problem worth looking at, because the list of available columns is short, and every column on it describes an account in isolation rather than against the thing you are trying to sell it.
What territories actually get cut on
The Sales Management Association surveyed sales organizations on territory design and Xactly published the findings. Three numbers describe the state of practice better than any framework does. Seventy-six percent of companies assign territories by geography. Seventy-six percent do territory planning once a year. Eighty-three percent use spreadsheets to do it, moderately or frequently.
A fourth number explains the first three: fewer than 40 percent of companies believe they can properly measure the data that territory planning needs. When you cannot measure the thing you want to balance, you balance the thing you can count.
So the inputs become proxies, and each was a good proxy once. Geography stood in for travel cost when a rep drove to accounts. Employee count stood in for budget. Last year's revenue stood in for next year's. Industry code stood in for fit. Stack those four and you get a territory balanced on company size, drive time, and history, which is a reasonable description of a 1995 sales force and a poor description of a software company selling five products into a base that already owns two of them.
Andris Zoltners, Prabhakant Sinha and Sally Lorimer made the general version of this argument in Harvard Business Review: analytics had arrived across sales force decisions, and territory design was still the corner it had not reached. A decade on, the survey data says the corner is mostly still dark.
What the imbalance costs
The same research puts a number on the gap. Organizations that rate themselves effective at territory design hit sales objectives 14 percent above the average. Organizations that rate themselves ineffective land 15 percent below it. That is a 29 point spread on goal achievement, and the population is not evenly split: 64 percent of respondents put themselves in the ineffective or only somewhat effective camp.
Set that next to what quota attainment looks like now. RepVue's Q2 2025 Cloud Sales Index, covering 246 cloud and software companies and roughly 47,000 quota-carrying reps, put average attainment at 42.69 percent, with 57.31 percent of reps missing target. Territory design is obviously not the only cause of that. It is, however, one of the few causes a revenue operations team fully controls, and it is decided months before anyone can do anything about the rest.
The rep-level version is easier to see than the aggregate. Take two books that are perfectly balanced on the usual columns: 60 accounts each, matched on employee count bands, matched on prior-year revenue, contiguous regions, no overlap. One of those books happens to contain eleven accounts that replaced a system you integrate with in the last two quarters. The other contains one. Both reps carry the same number. Only one of them has a year.
Nothing in the spreadsheet knew about those eleven accounts, because the spreadsheet had no column for what an account needs from what you sell.
The column the spreadsheet does not have
It is worth laying the inputs side by side, because the gap stops being subtle once you ask what each one tells you about next quarter.
| Input | Where it comes from | What it tells you about next quarter |
|---|---|---|
| Employee count | Enrichment vendor | Whether they could afford it. Not whether they need it. |
| Industry code | CRM field | Which playbook to open. Not which product to open it to. |
| Last year's revenue | Billing system | What already sold. A weak predictor in a base with room left. |
| Geography | Address field | Travel cost, which is near zero for an inside team. |
| Open whitespace cell | Product catalogue plus billing | What has not sold yet. A gap, not a reason. |
| Matched buying signal | Public sources read against your catalogue | Which product, and why this quarter, with the source attached. |
The first five rows are facts about your filing cabinet or about the account's shape. Only the last row is a fact about the account's year, and it is the only row that requires knowing what you sell. That distinction is the whole argument. A whitespace cell tells you an account has not bought a product. A buying signal tells you why they might now: a hire, a funding round, a compliance deadline, a system change, a public statement about a problem you solve.
This is the job PitchSmart does. It reads every account on a list against what you actually sell, then returns the buying signals that say who needs which product now, each with the source it came from. The output is per account and per product, which means it can be summed into a book and compared across books.
Cutting a territory on matched opportunity
The practical version is not a rewrite of your planning process. It is one extra column, added before the carve rather than after it.
- Write down the catalogue first. Every solution you sell, and for each one, the problem it solves and the observable conditions that make it urgent. Territory math cannot score need against a product line nobody has written down. If your ICP definition stops at firmographics, this is the step that extends it.
- Research the whole list, not a sample. A sampled read gives you a rate, not a carve. You need a score on every account, including the ones nobody has looked at since import, because those are precisely where the surprises live.
- Score on matched need, not on size. An account with three live signals pointing at two of your products outranks a larger account with none, regardless of headcount.
- Balance books on scored opportunity. Equal account count is not the goal and never was. A rep with 30 accounts carrying live signals has a comparable book to a rep with 55 accounts at baseline, and both reps can see why.
- Keep geography as a constraint, not the axis. For a field team, travel cost is real and belongs in the model as a limit. For an inside team it belongs nowhere near the primary split.
- Re-read quarterly. Signals expire. A funding round is loud for two quarters and irrelevant after four. Three quarters of companies plan territories annually, and nearly 40 percent of those already suspect more frequent assessment would be worth something.
The mid-year problem
Annual planning is not defended on principle. It is defended because re-cutting is expensive and disruptive, and because the evidence for a mid-year change is usually one rep's anecdote against another rep's anecdote. That argument is unwinnable without data, which is why it usually ends in no change.
A scored list changes the shape of the argument. If the scoring runs again in April and one book's total opportunity has dropped by a third while another has doubled, that is a reviewable fact with sources behind it. You may still decide not to move anything, because disruption carries a cost too. But you will be deciding rather than defaulting.
What changes at the territory review
The territory review is usually a fairness meeting. Reps arrive believing their book is thin, they argue in account counts and prior revenue because those are the only units available, and the manager splits the difference. Everyone leaves suspicious.
When every account carries a score and each score carries its sources, the unit of argument changes. A rep saying the book is thin now has to say which accounts have no reason attached, and that is a claim anyone can check in a minute. Sometimes they are right and the book genuinely has nothing in it, which is useful to know in January rather than October. Sometimes they are wrong, and the fastest way to find out is to open the eleven accounts they have not called.
The coverage question also becomes answerable. "Has every account in this territory been looked at" is a question that, in most organizations, nobody can answer honestly, because looking at an account means a rep spent 20 minutes on it and there is no record of whether they did. If the research ran across the full list, coverage is a fact rather than a hope, and account mapping inside the top accounts becomes a follow-on step rather than the only step anyone has capacity for.
It also gives you something to feed back. The accounts that converted can be compared against the signals that were present when the territory was cut, which tells you which signal types actually predict a deal for your product. That is the loop that makes scoring improve instead of ossify.
What this does not fix
Territory design is one input. Being honest about the rest matters, because a better carve gets blamed for things it was never going to solve.
- Over-assigned quota. If the total number handed to the team exceeds what the market plus the current capacity can produce, a perfect carve distributes the miss more evenly. It does not remove it.
- Ramp time. A new enterprise rep takes months to become productive regardless of how good the book is. Capacity models that use headcount instead of ramped capacity will keep being wrong.
- Comp plan conflicts. If the plan pays most on new logos, reps will underwork expansion signals in their book no matter how well those signals are scored.
- Bad data hygiene. Scoring a list with duplicate accounts and stale ownership produces a confidently balanced carve of a fictional base.
What it does fix is narrow and worth having. It replaces a balance measured in company count with a balance measured in matched opportunity, and it does that before the year starts rather than after the first quarter proves the carve wrong. Given that most organizations already rate their own territory design as ineffective, and given the 29 point spread between the ones that do it well and the ones that do not, that is a large amount of room for one extra column.
