How to Prioritise a Book of Business

Every article about account prioritisation gives you a framework. Value against risk, four quadrants, three tiers, a weighted score. The frameworks are mostly fine. They all share one assumption that nobody states out loud, which is that your book is a normal size and the method will therefore fit it.

We went looking for what a normal size is. Across 107 live Customer Success Manager and Account Manager postings, four say how many accounts you will carry. Four. The rest describe a job whose entire content is managing a book without ever saying how big the book is.

Where companies do publish the number, the spread is the finding. At one software company, on the same day, an enterprise account manager is told to own around 25 to 50 accounts while an SMB account manager at the same company is told to own 150 to 200. Measured at the starting point, that is a six-fold difference in what a single person is expected to hold, inside one business, under almost the same job title.

A method built for 30 accounts does not survive 200. A method built for 200 is wasted effort at 30. Before you rank anything, you need to know which of those two jobs you actually have.

What the postings actually say

Counts below are the exact phrasing used in live listings, pulled on the same day from public job boards.

Three of those four were advertised by the same employer, on the same day, for three variants of the same job. That is the part worth sitting with. It is not four companies disagreeing about what a book should be. It is one company deciding that an enterprise rep can hold 25 accounts and an SMB rep can hold 150, and publishing both without comment.

The postings say why, which is rare enough to be worth quoting. The 25-to-50 book is for customers with over 1,000 employees. The 150-to-200 book is for customers with 10 to 100. Same employer, same week, and the size of the customer is the thing moving the number.

It does not explain everything. A third role at that company carries 60 to 80 accounts and also serves customers over 1,000 employees, so two books differing by more than double sit behind the same customer profile. Segment sets the rough band. Something else, probably territory or seniority, sets where you land inside it.

Which is the practical point. If anyone has ever given you a number for how many accounts a person should carry, it was a number for a customer size, a contract type and a seniority they did not mention.

The absence matters as much as the numbers. 96% of these postings describe owning a portfolio and never size it. If the companies hiring for the job will not commit to a number, the chance that your own book size was ever a deliberate decision is lower than you would like.

So the first useful thing you can do is count. Not estimate, count. The number you get decides which of the three regimes below you are in, and the regimes do not share a method.

The three regimes

Under roughly 50 accounts. You can hold every account in your head. You know the champion's name, what they bought, what went wrong in March. Ranking is not your constraint, because you already know which accounts are in trouble.

Roughly 50 to 120. You cannot hold them all, but a ranked list still works. This is the regime every prioritisation article is written for, and the only one where the standard advice applies cleanly.

Over roughly 120. A ranked list stops being the answer, because the bottom two thirds of it will never be reached. The question changes from "which accounts first" to "which accounts get a human at all".

These boundaries are judgement, not measurement, and the real line moves with contract complexity. Fifteen accounts with bespoke commercial terms will consume more of a week than a hundred on standard contracts. Treat the numbers as a rough shape, and if your renewals routinely involve procurement and legal, slide yourself one regime up.

For sizing the book in the first place, rather than working one you have already been handed, our accounts per CSM coverage model sets out how the number should be decided.

What to rank on, briefly

This part is not original and it does not need to be. Three inputs, and the standard answer is the right one.

Value. Current ARR plus realistic expansion. Not the logo, not the brand name, and not how much you enjoy the calls.

Risk. How likely the account is to slip, read from leading signals rather than from recall. Usage drifting down, a champion going quiet, seats sitting unused, a support pattern changing.

Timing. Whether there is a renewal, a budget cycle or a decision inside the next 30 to 90 days.

An account that is high on all three is where the week starts. The reason this is worth only three paragraphs is that getting the inputs right is easy and the frameworks all agree. What they disagree about, or rather what they never address, is everything that follows.

Under 50 accounts: ranking is not your problem

If you have fewer than about fifty accounts and you are still losing customers you did not see coming, the problem is not prioritisation. You had the capacity to know. Something else failed.

Usually one of three things. You were measuring attention rather than outcome, so a quiet, healthy-looking account was never examined. You inherited accounts without context and never rebuilt it, which our guide to inherited accounts from a CSM who left covers. Or you were pulled into internal work and the book got whatever was left.

At this size the fix is a standing review of the whole book, not a ranking of it. Thirty accounts can be read end to end in an hour. Do that monthly and you will catch more than any scoring model would.

Building a scoring framework at this book size is a way of feeling organised. It is not a way of saving accounts.

50 to 120: ranking is the job

This is the regime the standard advice was written for, and here it works.

Score every account on value, risk and timing. Work the list from the top. Re-rank when the inputs change rather than when the calendar says to.

Two failure modes are worth naming, because both are common and neither appears in the usual framework articles.

The list that is never reached below line twenty. If the same forty accounts sit untouched at the bottom every week, you do not have a ranked book. You have a book of twenty accounts and a long tail you have silently written off. That may be the correct decision, but it should be a decision rather than an accident, and the tail needs a motion of its own.

Ranking by who contacted you. The accounts that email most are usually the engaged ones, which is to say the ones already getting value. Accounts in trouble tend to go quiet rather than loud. A ranking that follows your inbox will reliably invert the one you need. This is the single most expensive habit in the regime and it is also the easiest to miss, because a full inbox feels like evidence of a well-run book.

If your team is stuck reacting rather than choosing, proactive customer success for a reactive team deals with the operating habits underneath that.

Over 120: ranking quietly stops working

At this size, a ranked list is still produced and still ignored below the fold. The honest move is to stop pretending the list is the plan.

Split the book by motion rather than by rank.

Accounts that get a human. A deliberately small set, chosen on value and timing, reviewed weekly. Small enough that each one genuinely receives attention, which usually means fewer than you want it to.

Accounts that get a trigger. Everything else, covered by signals rather than by a person. Nobody proactively works these. They enter the human set only when something fires: usage drops below a threshold, a renewal window opens, a champion leaves.

Accounts that get a programme. One-to-many. Lifecycle email, in-product guidance, office hours. The aim is not attention, it is a floor beneath which service does not fall.

The thing that makes this regime work or fail is whether the trigger layer is real. If the second group is defined as "I will keep an eye on them", it does not exist, and you are running the first group plus several hundred accounts nobody is watching. A trigger you have not written down and automated is not a trigger. It is an intention.

This is also the regime where a health score earns its keep, and worth knowing that scores go wrong in predictable ways. Why customer health score accuracy is so poor sets out the failure modes before you build one.

Still ranking your book by whoever emailed last?

The tier everyone skips

Published prioritisation frameworks are good at telling you who to work and silent on what happens to everyone else. Both of the most prominent guides on this topic propose a model and then stop, with no guidance at all on the bottom tier.

That silence is a problem, because if you are measured on gross revenue retention, the tail churning still costs you the number.

Two things resolve it, and neither needs more hours.

Keep two lists, not one. Ranking by value answers where to spend effort growing an account. It is the wrong instrument for deciding where to stop a loss. Run a second, much shorter list sorted only on risk and proximity to a renewal, with no value weighting in it at all. A small account on that list outranks a large healthy one, because the question it answers is not what the account is worth but whether it is about to stop existing.

Neglect, not dissatisfaction, is what usually takes the tail. Smaller customers rarely leave after a difficult conversation. They leave after none. A renewal reminder that actually fires, an onboarding sequence that completes, and one check that somebody is still logging in will hold a meaningful share of them without anyone picking up a phone. Built once, that layer runs for years.

The test for whether your bottom tier is deprioritised or abandoned is simple. Name the mechanism that would put a 4,000-dollar account back on a person's desk next Tuesday. If you cannot name it, there isn't one.

What to do when two accounts tie

The frameworks produce scores and then assume the ordering is obvious. In practice ties are constant, because most scoring models have coarse inputs and a hundred accounts will not spread cleanly across them.

When two accounts score the same, three tie-breaks in order.

Reversibility. Prefer the account where acting now still changes the outcome. A renewal 60 days out is worth more of your hour than one 10 days out, because the second is largely decided. Most teams do the opposite and chase the nearest deadline.

Single point of failure. An account with one champion and no second relationship is more fragile than a larger account with four stakeholders, whatever the score says. Breadth of relationship is rarely in the model and should be in the tie-break.

Diagnosability. Prefer the account where you know what is wrong. An hour on a problem you understand beats an hour discovering that you need a different hour next week.

If you cannot break a tie with any of these, the accounts genuinely are equivalent and you should take the larger one. That is the only point in this method where size is the deciding factor.

How often to re-rank

The usual answer is monthly or quarterly. That is a calendar talking, not a book.

Re-rank when an input changes materially, and audit on a cadence. In practice that means three triggers:

A risk signal fires. The ranking is wrong the moment it is wrong, not at month end.

A renewal window opens. Timing is one of the three inputs, so an account crossing into the 90-day window should move without anyone deciding to move it.

Ownership changes. A reassigned account has no accurate risk reading until someone rebuilds the context.

The monthly or quarterly review then does a different job: not re-ranking, but checking that the model is still sorting correctly. The question in that review is whether the accounts that actually churned last quarter were near the top of the list before they left. If they were not, the inputs are wrong and no amount of re-ranking fixes that. Health score coverage as a KPI covers how to measure whether the model is reaching the book at all.

Where GainTrace fits

Everything above rests on one condition: the order of the list has to come from what accounts are doing, not from who reached you most recently. Holding that line by hand is possible at thirty accounts and impossible at three hundred, which is the narrow thing software is genuinely good for.

GainTrace reads the same three inputs across every account, including the ones that have never contacted you, and keeps the resulting order on a single record alongside the owner and the renewal date. Accounts crossing into risk are raised roughly 45 days ahead of their renewal, while a save still costs a conversation rather than a discount. The first 25 companies are free forever.

None of that replaces the judgement in this article. It removes the specific failure the judgement keeps losing to, which is that a silent account produces no reminder and a noisy one produces several.

Want your book ranked on what accounts are doing, not who called last? Start GainTrace free.