ICP Definition Framework: One Profile per Solution, Step by Step

Ask a revenue team for their ideal customer profile and you usually get one slide. A revenue band, a headcount range, four industries, a tech stack, maybe a trigger like "recently raised". The same slide is used whether the rep is selling the flagship platform, the compliance module, or the services retainer. That is the flaw, and it costs money in a specific, traceable way: the account clears the bar, a rep works it for six weeks, and the product they led with was never the one that account needed.

A single company-wide profile answers the question "is this account worth our attention at all". It cannot answer "which of our seven things should we open with, and why this quarter". Those are different questions, and the second one is where most multi-product revenue is won or lost. If you are still settling the basics of what the acronym covers, start with what ICP stands for in business, then come back here for the part the standard guides skip.

Where one company-wide ICP breaks

The collapse into a single profile fails in three places, and each one is visible in your own CRM if you go looking.

The profile describes your biggest product, not your company

Most ICPs are reverse engineered from closed-won data, and closed-won data is dominated by whatever product has been selling longest. So the "company" profile is really the flagship profile wearing a company badge. Every newer product then gets aimed at accounts selected for a different reason. The newer product underperforms, and the internal conclusion is usually that the product is weak rather than that the targeting was inherited.

Good-fit accounts get sold the wrong thing

This is the failure that shows up in margin rather than in pipeline. An account can be a textbook fit for your company and a poor fit for one particular product, and selling them that product makes the relationship worse. The evidence here is older than most of the current sales tooling and it is still the clearest warning in the literature. Denish Shah, V. Kumar and colleagues published Unprofitable Cross-Buying: Evidence from Consumer and Business Markets in the Journal of Marketing in 2012, and the finding that the American Bankers Association's Banking Journal pulled out of it is blunt: roughly one in five cross-buying customers is unprofitable, and that minority accounts for about 70% of the firm's total customer loss. Cross-selling raised average profit per customer overall. It also concentrated the losses.

Read that as a targeting result rather than a product result. Nobody sets out to sell the wrong module. They sell the module the account qualifies for under a profile that was never built to distinguish between modules.

Territory and routing inherit the flaw

Once one profile is the system of record, it propagates. Scoring models, routing rules, territory carve-ups and quota setting all read from it. By the time a rep sees an account in their queue, the decision about which product it is a candidate for has already been made upstream by something that was never asked that question. Forrester analysts Stephanie Sissler and Katie Fabiszak made a related point in January 2026 in Multiply B2B Growth: Why Your Ideal Customer Profile Alone Isn't Enough: an ICP sharpens your aim, but on its own it does not do the rest of the work you are asking of it.

What a per-solution ICP actually contains

A solution-level profile is not a copy of the company profile with one field changed. It answers a different question, so it holds different fields. The company profile is a gate. The solution profile is a match.

Field	Company-level ICP	Solution-level ICP
Purpose	Should we spend attention here at all	Which of our products fits this account, and how well
Firmographics	Revenue band, headcount, geography, industry	Same, but narrowed to where this solution has ever worked
The problem	Usually implicit	Stated explicitly, in the buyer's words, one problem per solution
Owner of the problem	A department	A named role that is measured on the problem
Disqualifiers	Rare, usually size or region	Specific and numerous: an incumbent, a contract cycle, a process they already run
Evidence of need	Fit attributes only	Observable signals with a date and a source
Refreshed	Annually, by RevOps	Per account, when something changes

The two rows that carry the most weight are the disqualifiers and the evidence. A company-level ICP rarely disqualifies anyone, because its job is to keep the funnel wide. A solution-level ICP should reject aggressively. If your compliance module is useless to a company that already runs the process in-house, then "runs the process in-house" is a disqualifier, and any account with that trait should be routed to a different product rather than worked and lost.

Building the framework, one solution at a time

The work is finite. For a company with five or six sellable solutions this is about two weeks of RevOps and product marketing time, done once, then maintained.

1. List what you actually sell as separate motions. Not the price book. The unit is a thing a rep can lead a conversation with. Three SKUs that always sell together are one solution. One SKU that two different buyers purchase for two different reasons is two solutions.
2. Write the problem each solution solves, in the buyer's language. One sentence, no product nouns in it. If you cannot write the sentence without naming your product, you do not yet know the problem, and the profile built on top of it will be a description of your feature set.
3. Pull the closed-won accounts for that solution alone. Filter by the line item, not the logo. Twelve accounts is enough to see a pattern. If a solution has fewer than five wins, mark its profile as a hypothesis and label it that way so nobody scores against it as though it were evidence.
4. Find the shared trait that is not firmographic. Company size will show up first and it is usually a red herring shared with every other solution. Look for the operational fact: they run a multi-entity finance close, they sell through distributors, they hold a regulated data category, they just absorbed another company's book of business. That trait is the profile.
5. Write the disqualifiers from closed-lost. Lost deals are more useful here than won ones. The reasons that repeat across losses are the disqualifiers, and they are what stop the profile from being a wish list. Your lead qualification process gets sharper the moment these are written down rather than carried in a tenured rep's head.

The output is one page per solution. Problem, owner, fit traits, disqualifiers, and the two or three observable things that suggest the problem is live right now. Nine solutions means nine pages. That is the correct amount of work, and it is less work than a team of reps guessing for a year.

Fit is an attribute, timing is a signal

A profile built only from firmographics tells you an account could buy. It says nothing about whether they are going to do anything this quarter. That second half is what turns a list into a decision, and it comes from evidence rather than from fields in a database.

For each solution, write down the two or three observable events that mean the problem just became someone's priority. A new VP in the function that owns the problem. A posted role whose responsibilities are the problem. A regulatory deadline in their sector. An acquisition that doubles the volume of the thing your solution handles. A statement in an earnings call or a press release. These are ordinary, checkable facts, and they are the difference between a target account and an account worth calling on Tuesday. Our writeup on buying signals in sales covers how to define them so that two people reading the same source reach the same conclusion.

This is the whole reason the per-solution split matters for expansion revenue. Account whitespace analysis gives you a grid of which customers own which products, and every empty cell in that grid looks identical. A per-solution profile plus a live signal tells you which of those empty cells has a reason attached to it this quarter, which is a different and much shorter list. If you are working out where a second product fits against an existing relationship, the distinctions in cross-sell versus upsell matter more once the profiles are separate, because the buying committee for a genuinely new solution is often not the committee that bought the first one.

This is the job PitchSmart does. It reads every lead on your list against what you sell, matches each account to the solution that fits, and returns the reason with its source attached, so the plan for that lead says which product to raise and why now instead of leaving the rep to guess from a grid.

Keeping nine profiles from becoming nine spreadsheets

The honest objection to per-solution ICPs is maintenance. One profile is already stale within a year. Nine profiles, each needing per-account evidence, sounds like a research tax paid by the people with the least time to pay it. Salesforce's State of Sales research, drawing on more than 4,000 sales professionals, puts 60% of a rep's time into non-selling tasks. Manual per-account research is exactly the kind of work that lands in that 60%.

Three rules keep the framework from collapsing under its own weight.

• Separate the stable part from the perishable part. The profile itself (problem, owner, fit traits, disqualifiers) changes once or twice a year. The evidence per account changes constantly. Store them separately and refresh them on different clocks. Most ICP decay is actually evidence decay being blamed on the profile.
• Give every profile one owner and one review date. Product marketing owns the problem statement, RevOps owns the traits and the scoring. Unowned profiles drift into marketing copy within two quarters.
• Automate the evidence, never the judgment. Gathering the signals for every account on a list is mechanical work that should not be done by hand. Deciding what the signals mean for a specific deal is the rep's job. Any system that hides its sources has moved the judgment somewhere you cannot audit it, which is why every claim in a PitchSmart plan carries the source it came from.

What to do this quarter

You do not need the full framework to get the first benefit. Start with the gap that is costing the most.

• Take your two highest-revenue solutions and split the single ICP into two. Just two. Put them side by side and mark every field where they differ.
• Pull the last twenty closed-lost deals for the smaller of the two. Write down every repeated loss reason as a disqualifier.
• Apply those disqualifiers to the current open pipeline for that solution. The deals that fail them are the forecast risk you have been carrying without naming it.
• Pick the three signals that mean the problem is live, and check your top fifty accounts against them. That list is your next quarter, and it will not look like the list your single profile produced.
• Once the two profiles are working, add one per quarter rather than attempting all nine in a single sprint. Profiles written in a rush are firmographics with adjectives.

The test of the framework is simple and you can run it in a pipeline review. Point at an account and ask which solution it is a candidate for, and what makes you say so. If the answer is a revenue band and an industry, you have one profile doing the work of nine. If the answer is a problem, a role that owns it, and a dated fact showing the problem is live, the framework is doing its job.