Balance territories on value, not account counts

Every sales org redraws territories the same way: count the accounts, divide by the reps, nudge until the piles look even. Maybe you weight by pipeline or named-account tier. The spreadsheet says twenty reps, ~75 accounts each — fair. Then a rep opens their book and finds 75 names they'd never call.

An even count is not a fair book

Two reps with 75 accounts each can hold completely different books. One has eight of the segment's twenty best accounts; the other has none of them and 75 long-tail names. Count-based balance calls that fair. Everyone on the team knows it isn't — they just can't point at the number that proves it, because the only number they were handed was the count.

Concentration is invisible to a headcount split precisely because it's about which accounts, not how many. So stop measuring the size of the pile and start measuring its value.

Two questions, kept apart: Capture and Activation

"Is this book fair?" is really two questions, and jamming them into one number is why territory dashboards get ignored. The skill keeps them apart, as two columns a RevOps lead can read without a decoder ring.

Capture — how much of the segment's best business sits in this rep's book. Take the segment's best accounts by ICP score and add up what they're worth: an account in the segment's top 10% counts double, the top 25% counts once, everything below counts zero. A rep at 20% Capture holds a fifth of the segment's best business. Reps in a segment should land close together — roughly within 1.5× of each other. One rep at 3–4× another is hoarding the good stuff; a rep near 0% has been under-supplied, or is ramping.

Activation — of the best business a rep already holds, how much are they working. Same weighting, different question: what share of it has had a meeting, call, or outbound email in the activity window. It's deliberately independent of book size — a rep with 30 accounts and a rep with 300 can both be at 100%.

Splitting them turns a vague "seems off" into a diagnosis. The quadrant that costs you money is high Capture, low Activation: the segment's best accounts are parked with someone who isn't touching them. No account count will ever surface that. It's the single most expensive thing a book can be doing, and it's the first thing this app shows you.

Rank within a segment, not globally

A rep's book mixes segment sizes — a marquee account that falls just below your enterprise line still lives in Commercial — and a Commercial rank of 2 is not comparable to an Enterprise rank of 2. Average ranks across both universes and you get numbers no real book could produce.

So the app ranks each rep's accounts only against their own segment, and shows one honest depth column: the average in-segment rank of the rep's top 25 accounts, where 1 would mean they own the segment's very best 25. One number a reviewer can hold in their head, not a ladder of five nested cutoffs that reads as noise. Coverage adds a whole-book column so you can see how far activity falls off past the top of the book; strength doesn't, because an average rank over a whole book is dragged around by book size and invites exactly the cross-rep comparison it can't support.

The value has to be real

Capture and Activation are only as good as the "value" underneath them. If value means headcount, you've rebuilt the size ranking you were trying to escape; if it means gut feel, reps won't trust the reassignment.

This skill takes account strength from an account-scoring run — a score calibrated to your closed-won deals, where every component links through to the people, teams, and projects behind it. (No scoring run yet? It falls back to Sumble's own account score via the API.) Either way, "the segment's best accounts" means best for you, not biggest, and a rep facing a proposed move can click through to why the account is worth taking.

The flags a balance number hides

Balance is the average; the flags are where the money leaks. The app surfaces five, each a filter on the account list:

• Strong but idle — a top account, owned, and nobody's working it. The most expensive kind of neglect.
• Strong but unallocated — a top account with no active owner at all.
• Wrong segment — the account's size puts it in one segment; its owner sells another.
• Double-allocated — two reps own CRM records that resolve to the same company. Fix these first: they distort every other number.
• Not being worked — owned, but untouched in the window.

A top-N "strong account" cutoff recomputes the strong-account flags live, so you can tighten the definition of "strong" and watch the neglect list shrink or grow.

Suggest, review, export — the CRM is never touched

An automated reassignment that's wrong is worse than the imbalance it fixed. So the balancer only proposes. It measures how uneven each segment's books are (a coefficient of variation across reps), then suggests the smallest set of owner moves that pulls them into balance — respecting per-rep capacity, handing wrong-segment accounts back, and giving unallocated strong accounts an owner.

The proposals land in a review queue as the app's primary call to action. One dropdown per account does everything: accept the suggestion, pick a different rep, or dismiss it. Because the whole app groups by effective owner (current owner plus your accepted moves), the heatmaps and flags update the moment you decide — you see a move help before you commit to it. When you're done, one button writes actions.csv: a plain list of approved owner changes keyed by CRM account id, ready for your admin or a follow-up agent run. Nothing is written back automatically.

What it looks like on a book

The skill's public repo ships a runnable demo built on a fictitious vendor, Northwind, with 1,935 accounts and 20 reps (10 Enterprise, 10 Commercial). The data is synthetic, but it's built to show the shapes you'll recognize:

• Chen Wei holds the biggest Enterprise book and a third of the segment's best accounts (33% Capture) — and works almost none of them (6% Activation). The sitting-on-value quadrant, invisible to an account count.
• Dev Ramesh is the foil: a smaller book worked to 100% Activation.
• Grace Okafor is a ramping rep at 0% — a small book by design, so a low number reads as "new," not "failing."
• Alex Rivera is the commercial star: the biggest book, and works it hardest.

Under the current settings the balancer makes 480 proposals — owners for 445 unallocated accounts, 18 wrong-segment handbacks, 17 rebalance moves — which together would take both segments from uneven (CV ≈ 0.3) to balanced (≈ 0.1) if every one were accepted. Which, of course, they shouldn't be blindly. That's what the review queue is for.

Run it on yours

You need a coding agent (Claude Code, OpenAI Codex, or Cursor), your CRM ownership (Salesforce, HubSpot, a warehouse, or a CSV with account, owner, and a size signal), and ideally an existing account-scoring run for the value column. Connect whatever activity sources you have — Google Calendar, Gong, Fireflies, Granola, Salesforce email — and Activation lights up; skip them and you still get balance and segment fit.

npx skills add SumbleData/sumble-skills-public --skill sumble-territory-planning

Start a new agent session and run it (/sumble-territory-planning in Claude Code; "use the sumble-territory-planning skill" in Codex or Cursor). The interview sets your segments (Enterprise + Commercial by default), whether the segment line is a hard rule or should be calibrated from your data, and your target book size per segment. Then:

cd territory_planning/<your-company>
python3 app.py        # http://localhost:8002

Calibrate the line and the targets, work the suggested-moves queue, export actions.csv.

The part that compounds

A territory plan is usually a once-a-year fire drill that's stale by Q2: someone rebuilds it in a spreadsheet, everyone argues, and the books drift the moment a rep leaves or an account heats up. Keying territories to a live account score changes that. Re-run it, and the accounts that got stronger, the reps who fell behind on coverage, and the new hire who needs a book all show up as fresh moves in the same queue. The plan stops being a project and becomes a habit.

It also sits on the same foundation as the rest of the GTM engine. CRM cleaning resolves every account to a real organization — and catches the double-allocations before they distort your balance. Account scoring tells you which accounts are worth fighting over. This skill puts the right reps on them.