# B2B SaaS Ideal Customer Profile

> Build an ICP from closed won data: firmographic, technographic and behavioural criteria, a negative ICP, and a scoring model wired into routing and ad audiences.

Source: https://saas-marketing.net/guides/b2b-saas-ideal-customer-profile/
Topic: B2B SaaS Marketing
Type: guide
Published: 2026-09-11
Last updated: 2026-09-11
Publisher: SaaS Marketing (saas-marketing.net)
License: CC BY 4.0. Quote or republish with attribution and a link to https://saas-marketing.net/guides/b2b-saas-ideal-customer-profile/

## Short answer

A B2B SaaS ideal customer profile is an account level definition of the companies your product wins with, retains and expands, derived from closed won, churn and expansion data rather than a persona workshop. It combines firmographic criteria (size, revenue, geography, industry), technographic signals (the tools already in the stack), behavioural signals (usage or research activity) and trigger events. A working ICP also names who you refuse to sell to, and it feeds scoring, routing and ad audiences so it changes daily behaviour.

## Key takeaways

- Derive the ICP from closed won, churn and expansion cohorts, not from a whiteboard persona session with no data behind it.
- Segment win rate, ACV, cycle length, net revenue retention and support cost together, because one metric alone will mislead you.
- A negative ICP raises win rate by removing deals that consume capacity and close at under ten percent anyway.
- ICP describes the account, persona describes the human, and committee roles sit underneath both in a B2B SaaS deal.
- Score accounts on four to seven weighted criteria, then wire the score into routing, sequencing and paid audiences within one quarter.
- If your ICP excludes nobody, it is a poster, not a filter, and sales will ignore it inside a month.

---

Most ICP documents are written in a two hour workshop, pasted into Notion, and never opened again. The version that survives is built from a spreadsheet of your own closed won and churned accounts, and it is specific enough that a rep can look at an inbound lead and say no in four seconds. This page walks the data pull, the four criteria layers, the negative profile, and the scoring model that turns the whole thing into routing rules. It is the account side of [B2B SaaS marketing](/b2b-saas-marketing/), and it sits upstream of nearly every targeting decision you make.

## Start with the data pull, not the workshop

Pull every closed opportunity from the last 18 to 24 months and every churned account in the same window. You want one row per account with these columns: employee count, revenue band, industry, geography, ACV, days from first meeting to close, win or loss, net revenue retention at month 12, and support tickets per seat in the first 90 days.

That last column is the one nobody pulls and the one that changes the answer. A segment can win at 35 percent and still be a bad segment if it burns four times the support hours per dollar. We have watched teams discover that their "best" vertical by win rate was the worst by gross margin once implementation hours were loaded in.

Read the table by segment, not by metric. Sort by win rate and you find the easy sells. Sort by ACV and you find the big ones. The accounts you want sit where win rate, ACV and 12 month NRR are all above median at the same time, and that intersection is usually smaller and stranger than anyone expects.

Blending all segments into one average hides everything. A 22 percent blended win rate can be a 41 percent segment and an 8 percent segment sitting in the same bucket. Always break the pull by at least two dimensions before you draw a conclusion.

| Segment | Win rate | Median ACV | Cycle (days) | 12mo NRR | Support hrs / 10K ARR |
|---|---|---|---|---|---|
| Fintech, 200-1000 staff | 38% | $54,000 | 71 | 118% | 3.1 |
| Agencies, 20-100 staff | 31% | $9,400 | 24 | 86% | 9.8 |
| Healthcare, 1000+ staff | 12% | $91,000 | 188 | 104% | 14.2 |
| Devtools, 50-300 staff | 34% | $27,500 | 44 | 126% | 2.4 |
| Retail, 200-1000 staff | 17% | $31,000 | 96 | 79% | 7.6 |

Read that table the way a CFO would. Fintech and devtools are the profile. Healthcare has the biggest deals and is still the worst use of a quarter, because 188 days at 12 percent means a rep working five healthcare deals closes fewer than one a year. Retail is the negative ICP, and it is the row that gets defended in every meeting because somebody has a logo there.

## The four criteria layers, and which ones actually predict

A usable profile has four layers stacked on top of each other. Firmographic is the coarse filter. Technographic is the sharpest single predictor in most SaaS categories. Behavioural is what makes the list timely. Triggers are what make outreach land.

**Firmographic.** Size proxy, revenue band, industry, geography, and business model. Pick one size proxy and stick to it. If you charge per seat, employee count beats revenue. If you charge on usage, the usage driver (transactions, endpoints, data volume) beats both.

**Technographic.** What is already in the stack. If your product ingests data from Snowflake, accounts on Snowflake convert at a different rate than accounts on a legacy warehouse, and that difference is usually larger than any industry effect. Tools like Clearbit, BuiltWith, HG Insights and enrichment inside Clay can attach this to a list. Test three or four stack signals against win rate before you commit to any of them.

**Behavioural.** Product usage for PLG motions, research activity for sales led ones. Two people from the same domain reading your pricing page in a week is a stronger buying signal than any firmographic attribute you own.

**Trigger events.** New VP hire in the function you sell to, funding round, a compliance deadline, a migration announcement, a past champion changing jobs. Triggers do not define the ICP, they time it. The cleanest signal in B2B SaaS is still a champion who used your product at their last company landing somewhere new.

**2x to 4x** Typical spread in win rate between a company's best and worst firmographic segment

If you need a starting structure, the [ICP template for SaaS](/templates/ideal-customer-profile/) lays out these four layers as fields you can fill from your own pull, and [how to define your SaaS target market](/guides/saas-target-market/) covers the wider market sizing question that sits around the profile.

## Why the negative ICP raises win rate

The negative ICP is the list of accounts you will decline. Not "lower priority". Decline. It typically has three sources: segments where you lose more than 90 percent of the time, segments that churn inside 12 months, and segments whose support load destroys gross margin.

Writing it down does two things. It gives a rep social permission to disqualify, which is the single hardest behaviour to install in a sales team that is behind on quota. And it gives marketing a suppression list, which matters more than it sounds: excluding a bad segment from LinkedIn and Google audiences usually improves blended lead quality faster than any improvement to your good targeting.

"Retail companies under 1000 employees with no in house data team. Win rate 17 percent, 12 month NRR 79 percent, implementation requires 40+ hours of our solutions engineering. We do not pursue these. If one arrives inbound with budget, route to partner referral."

Note the last sentence. A negative ICP without a disposal route just creates arguments. Give the segment somewhere to go: a partner, a self serve tier, a lower touch plan. That also removes the objection that you are turning down revenue.

The honest cost here is real. In the quarter after you publish a negative ICP, total lead volume drops and somebody senior will notice before pipeline quality improves. Expect roughly one quarter of ugly dashboards before the conversion rates catch up. If you cannot hold that line politically, do not start.

## ICP versus persona versus committee role

These three get collapsed constantly and they do different jobs.

The ICP is the account. The persona is the human you write to. The committee role is what that human does inside the deal, and in B2B SaaS there are usually five: economic buyer, champion, end user, technical or security reviewer, and procurement. Gartner's research puts six or more people in the typical B2B buying group, and a chunk of them never speak to a rep.

The practical failure is producing five practitioner facing assets and nothing for the security reviewer or the CFO. Your champion has to sell internally with whatever you gave them. If the only thing in the pack is a feature comparison, that is what the CFO sees, and that is where deals stall at legal. Build a battlecard for competitive moments (the [SaaS competitive battlecard template](/templates/competitive-battlecard/) has the one screen format), and build a business case asset for the economic buyer separately.

## Scoring 200 accounts: a worked example

Here is the model we would ship for the data set above. Four weighted criteria, 100 points, no more than seven inputs total, because every input past that adds maintenance cost and no accuracy.

| Criterion | Weight | Scoring |
|---|---|---|
| Industry (fintech, devtools, insurtech) | 30 | 30 / 15 / 0 |
| Employee count 50 to 1000 | 25 | 25 in band, 10 adjacent, 0 outside |
| Data warehouse present (Snowflake, BigQuery, Databricks) | 25 | 25 yes, 0 no or unknown |
| Trigger in last 90 days | 20 | 20 funding or relevant VP hire, 10 job change, 0 none |

Run 200 target accounts through it and you get a distribution, not a binary. In a typical run you will see something like 18 accounts above 80, 61 between 50 and 79, and 121 below 50. That shape is the point. Tier 1 (80+) gets named account treatment and human research. Tier 2 gets sequenced and retargeted. Tier 3 gets nurture and nothing else.

**Wiring the score into behaviour**

That last step is the honest test of the whole exercise. Most ICP work is never validated. If tier 1 and tier 3 convert at the same rate, you have built a description of your existing customer base rather than a predictor, and those are different things.

## Where the ICP feeds the rest of the go to market

An ICP that only lives in a marketing doc is decoration. It should show up in at least five places within a quarter.

- Lead routing and qualification rules in the CRM, with a documented threshold
- Paid audience definitions and suppression lists across LinkedIn, Google and any ABM platform
- Outbound list build criteria, so SDRs stop sourcing accounts by whoever replied last
- The definition section of your [sales and marketing SLA](/templates/sales-marketing-sla-template/), which is where the arguments get settled
- Content and campaign planning, because vertical specific pages only pay off inside a defined vertical

Segment strategy flows from here too. If your model says 200 to 1000 employees, the [mid market SaaS marketing playbook](/playbooks/mid-market-saas-marketing/) is the motion that fits, and the [B2B SaaS go to market plan template](/templates/b2b-saas-gtm-plan/) is where the whole thing gets assembled into a quarterly plan. Pricing needs to agree with the profile as well. Selling to 50 person companies with a 45 day procurement process and an enterprise price list is a contradiction, and the [B2B SaaS pricing strategy](/guides/b2b-saas-pricing-strategy/) page covers where the packaging has to bend.

## What we would do differently, and what this costs

Two honest tradeoffs.

First, a narrow ICP shrinks your addressable market on paper, and if you are raising a round in the next six months, your investor deck and your operating ICP will not match. That is fine but you need to be ready to explain it. The deck describes where you could go. The ICP describes where you win this year.

Second, technographic data is expensive and decays. Enrichment vendors will happily sell you 40 attributes. Buy three. The cost of maintaining stack data across 5000 accounts is real, and most of it never enters a scoring model.

The position we hold: an ICP that does not exclude anyone is a marketing poster. If your team cannot name, out loud, the segment they refuse to sell to, you have a description of your total market rather than a profile. Start with the exclusion list. It is faster to build, it produces measurable capacity gains in one quarter, and it is the part nobody writes down.

## What to do this week

Pull the closed won and churn table with the six columns above. Break it by industry and size band only, ignore everything else for now. Find the two rows where win rate, ACV and NRR are all above median, and the one row where all three are below. Write those four sentences down, put the exclusion row in front of your head of sales, and see whether anyone can defend it. Then build the score, and read the [ideal customer profile](/glossary/ideal-customer-profile/) definition if you need a shared vocabulary before the meeting.

## Frequently asked questions

### What is an ideal customer profile in B2B SaaS?

It is a definition of the account type your product serves best, built from firmographic, technographic, behavioural and trigger criteria. Unlike a persona, it describes a company rather than a person. A useful B2B SaaS ICP is derived from closed won and retention data, is narrow enough to exclude real segments, and is expressed as a score that routing and targeting systems can read.

### How is an ICP different from a buyer persona?

An ICP answers which companies to pursue. A persona answers which humans inside those companies to talk to, and what they care about. In a committee sale you need both plus the committee roles: economic buyer, champion, end user, security reviewer and procurement. Marketing targets the ICP with account selection and targets personas with message and format.

### How much data do you need to build a reliable ICP?

Around 40 to 50 closed won deals gives you segment level signal you can defend. Below that, treat the ICP as a hypothesis and refresh it quarterly as deals close. Early stage companies should combine the small closed won sample with churn reasons and with win rate by segment, since churn data often reveals the exclusion criteria faster than wins reveal the inclusion criteria.

### Should the ICP include company size or revenue?

Include whichever correlates with your value metric. If you price per seat, employee count usually predicts account value better than revenue. If you price on usage, the volume driver matters more than headcount. Pick one primary size proxy, test it against ACV and net revenue retention, and drop the one that adds nothing to the model.

### How often should you update your ICP?

Rerun the underlying data pull quarterly and rewrite the profile once or twice a year. Pricing changes, new product lines and a shift from self serve to sales assisted all move the boundaries. Any quarter where win rate in a named segment moves more than eight points is a signal to rerun the analysis early rather than waiting for the calendar.

### What is a negative ICP and why write one down?

A negative ICP names the accounts you will not pursue: segments where you lose, churn, or spend disproportionate support hours. Writing it down gives reps permission to disqualify without looking lazy, and gives marketing a suppression list for paid audiences. Teams that publish one typically recover selling capacity within a quarter because low probability deals stop entering the forecast.

### Can you build an ICP before you have customers?

You can build a hypothesis. Use design partner interviews, the segments where your founders have distribution, and competitor review data on G2 to guess at fit. Label it clearly as provisional, review it every month, and be ready to throw it away. The first 20 closed won deals will tell you more than any pre launch research exercise.