Sales & PartnershipsFree
Account Scoring Model and Tiering
Tiers derived from your own wins, losses and churn, with the sample size beside every rate and the untested segments named.
Every account scoring template on page one starts the same way: pick your attributes, allocate a hundred points across them, and set the tier thresholds at eighty-five and sixty. The weights come from a table in a blog post and the thresholds come from nowhere at all. What makes a scoring model worth running is not the arithmetic. It is whether the numbers behind it came from your own deals, and whether anyone checked how many deals that was.
River fits the tiers to your closed-won, closed-lost and churned records, then prints the sample beside every rate it reports. A win rate is a proportion, and the interval around a proportion measured on a handful of deals is wide enough to swallow your company average. Segments with too little history are marked as such rather than scored, because a confident number on nine deals is the most expensive kind of wrong. Retention counts too: an account that signed and left inside a year is not a win the model should chase.
Built for the revenue operations lead rebuilding target accounts before a planning cycle, and for the sales manager who has to tell six reps which two hundred accounts are theirs. Run it before territory assignments and again when a quarter of results is in. Clean the list first with lead list cleanup, research the top tier with the account research brief, and see the rest of the tool index and the template library.
The segments you never tried
The flaw in scoring against what closed is quieter than it looks. A model trained on your history can only learn about segments your team actually worked. If nobody has prospected regional manufacturers under five hundred staff, the data holds no wins there, and the model reads that absence as evidence against them. It is evidence of nothing except where the reps have been going, which is usually where the last model told them to go.
So every segment gets one of three labels rather than a score. Measured means enough attempts to say something. Thin means the direction is there and the sample is not. Untested means nobody has tried, and that is a plan rather than a rating. Segments are defined in codes with published thresholds, not adjectives: the size standards table sets employee and revenue cut-offs against each industry code, so mid-market means the same thing to two people.
The last problem is arithmetic nobody does. A tier is only a tier if somebody can work it. Six sellers making four real touches a week can carry a few hundred named accounts between them, so a tier one holding eight hundred is a list with a flattering name on it. The output is a coverage plan: accounts per tier, touches per account, sellers required, and the gap stated in headcount rather than in ambition, so the tiering survives contact with the team it is handed to.
How it works
Bring the history
Closed-won, closed-lost and churned accounts, with the firmographics you had at the time.
River fits the segments
Rates per segment with their samples, and the boundary definitions written so nobody argues.
Read the evidence labels
Measured, thin or untested, so a low tier means low fit rather than no data.
Size the coverage
Accounts per tier against sellers and touches, with the shortfall stated in headcount.
What you get
- Every tier carries the win rate behind it and the number of deals measured
- Churned accounts counted against the segment that produced them, not filed under closed-won
- Segments you have never prospected labelled untested rather than scored low on missing evidence
- Segment boundaries written as industry codes and headcount thresholds, so two people read them identically
- Tier sizes checked against your seller count and the touches each tier is promised
- A named experiment for each untested segment, sized so the result will mean something
Common questions
How much history does it need?
Enough closed deals that the segment rates mean something, and it tells you where you fall short rather than guessing. Split a hundred closed opportunities across four segments and two of them will be too thin to separate from your average. Below that the output is still useful, because the honest finding is which segments you cannot yet judge.
What about intent and engagement data?
They belong in the model and they belong second. Fit tells you who to work; engagement tells you who to work first this week. Mixing them into one number hides which is driving the score, so the register keeps them in separate columns and the tier is set on fit. Engagement moves the order inside a tier.
Why does churn change the score?
Because a customer who leaves inside a year cost more than a clean loss. There was implementation effort, a support load, a reference you cannot use, and a renewal conversation that went badly in front of people who talk to each other. Counting that as a win teaches the model to send more sellers at the same segment.
What do I do with an untested segment?
Run it as an experiment, sized before it starts. The plan names how many accounts, which seller, how many touches and over what period, and what result would move the segment from untested to measured. Without the sizing you get an anecdote in eight weeks, and an anecdote is what the current model was built on. The accounts it spends are counted in the outbound sequence pack.
Can it fill in the tiers we already use?
Yes. Hand over your existing tier definitions and the accounts land in them, with the evidence for each boundary printed beside it. That version is often the more interesting output, because it shows which of your current tiers are supported by your own results and which two were drawn around the accounts a founder liked.
Where does the list of accounts come from?
Wherever you keep it: a CRM export, a purchased file, a conference list, or the three spreadsheets three people maintain. If it came from a vendor, run lead list cleanup over it first, because scoring a file that still contains your own customers produces a tier one you have to unpick by hand. A conference list gets cut to capacity by the event target list.
What happens once an account scores into tier one?
Tier one is a coverage decision, not a plan. The named accounts inside it still need the buying centre mapped, the open opportunities tied to specific stakeholders, and the relationship gaps priced in dollars. That is what a full account plan takes on, once an account has earned that level of attention.
Account Scoring Model and Tiering
Fill in the form and your workspace opens with the work already underway.