Why Your ICP Is Too Broad to Guide Sales
Your ICP is too broad to guide sales when its terms match the deals you lost as often as the deals you won, so nobody can tell which account to call next. For founders before or just after launch with a stalling pipeline, the fix is to test each term against closed deals, keep the terms that sort them and add a buying trigger visible from outside. The main constraint is how few closed deals you have.
The symptom and what it does not prove
This guide is for founders before or just after launch who have an ideal customer profile written down, have run a few months of outreach against it and see a pipeline that is busy but not converting. If you have not chosen a segment yet, start with how to choose your first ideal customer profile. If interviews go well but nobody pays, the gap may sit between the yes and the purchase, which when customer interviews say yes but nobody buys diagnoses. This guide covers a different failure: the profile itself is too broad to tell anyone which account to call next.
The symptoms tend to show up together:
- Results are flat across the list. Reply, meeting and pilot rates look about the same in every part of the list, so no group stands out to double down on.
- First calls are spent qualifying. Most of the first conversation goes into finding out whether the account has the problem at all.
- Every win has a different story. Ask why each customer bought and you get a different reason, buyer and use case each time.
- Two people build two lists. Give the profile to two teammates separately and their account lists barely overlap.
- Nothing is excluded. No account the team has met falls outside the profile.
Size labels are a common source of the problem. The U.S. Small Business Administration's contracting guide says size standards vary by industry and are generally based on the number of employees or the amount of annual receipts, and that for federal contracting most manufacturing companies with 500 employees or fewer, and most non-manufacturing businesses with average annual receipts under $7.5 million, qualify as small. A profile that says small businesses can therefore cover a three-person shop and a manufacturer with several hundred employees. Daniel's go-to-market mentoring page contrasts a label such as 'mid-market retailers' with what the first step of his process asks for, in his words: 'One segment, one moment that makes them buy, and the exact job title that signs.' That description is self-reported on his site.
A stalled pipeline does not prove the profile is the cause. The same symptoms can come from an offer that does not fit, a price set too high, a channel that does not reach the buyer or simply too few attempts to read any rate. The test that separates these causes is whether some group of accounts does clearly better than the rest. If the offer were wrong, accounts that match the profile in every respect would lose too. If the profile is too broad, you would expect a narrower group inside it to win while the rest of the list dilutes the result.
The ICP breadth matrix: does each term sort your deals?
An ICP is a set of terms: an industry, a size band, a territory, a trigger, a buyer and a budget. A term earns its place only if it sorts accounts, meaning it is much more common among the accounts that bought than among the accounts that did not. The breadth matrix tests each term against your closed deals. Use one row per term, including the terms in your current profile and any candidate terms you suspect matter.
| Column | What to record | How to read it |
|---|---|---|
| Term | One term from the profile or a candidate term, written so someone else could check it | Keep each term on its own row; do not combine terms |
| Won | How many won accounts match, out of all won accounts | Use counts, not percentages, while deals are few |
| Lost | How many lost accounts match, out of all lost accounts | Include accounts that went silent after a proposal |
| Stalled | How many open accounts with no next step in the past month match | Shows how much of the open pipeline the term would keep |
| Evidence | Said, seen, done or assumed, for how the match was checked | A term matched only by assumption cannot narrow anything |
| Verdict | Sorts, does not sort or exclusion | Decide only after counting |
The reading rules are simple. A term that matches about the same share of won and lost accounts does no work, however true it is; it describes how you built the list, not why accounts bought. A term that matches most won accounts and few lost ones is a narrowing term. A term that matches lost accounts and no won ones becomes an exclusion. Tag each match with the same said, seen, done and assumed labels used in the customer interview notes template. With a handful of closed deals, treat only large differences as signal and write every verdict down as a hypothesis for the next round.
The matrix needs accounts on both sides. If nothing has been lost yet, count accounts that declined a second meeting as lost. If nothing has been won, the matrix cannot narrow anything; use commitment steps short of payment, such as pilots and introductions to the approver, as described in how to turn customer evidence into a narrow ICP.
Check observable buying trigger
A profile that is too broad often lacks a trigger, or carries one that nobody can see. A trigger is the event that makes the problem urgent now: a new location, a hiring push, new equipment, a regulation taking effect, a leadership change or a funding round. Daniel's go-to-market page lists hiring, funding, regulatory changes, leadership moves and visible evidence of the manual process being replaced as signals for selecting accounts, again as his own description.
For the diagnosis, go back to each won account and write down the event that came before its first meeting, then look for a public trace of that event dated before the meeting: a job posting, an announcement, a permit or a news item. Do the same for lost accounts. Add the trigger to the matrix as its own row with a time window, such as a posting in the three months before first contact. If the trigger shows up in won accounts and rarely in lost ones, it is the most useful term you have, because it tells sales when to call as well as whom. If the event is real but leaves no public trace, keep it as a first-call qualifier rather than a list filter.
Investigate specific buyer role
A profile can name the right kind of company and still be too broad because it names a department or a seniority level instead of the person who signs. Check the won column first: who brought you in, who compared options and who approved the purchase, using the job titles the accounts use themselves. Then check the lost and stalled accounts for where the deal stopped. If wins were approved by the owner and losses stalled waiting for a partner, an office manager or a head office, the approval path is a sorting term, and it belongs in the profile as a named title plus an exclusion.
Two different approval paths among the wins often mean two segments sharing one profile. Pick the one with more wins and the shorter path, and move the other to a later round. The ICP template has separate fields for the user and the approver, which helps keep them apart once the diagnosis is done.
Fix the gap in budget and delivery fit
Budget and delivery fit mostly produce exclusions, and exclusions are what a broad profile is missing. For budget, record what each account paid for the problem before you arrived: a person's paid hours, a contractor, another tool or nothing. Won accounts that already spent money or paid time on a workaround, set against lost accounts that tolerated the problem for free, give you a term: accounts without a paid workaround are out.
For delivery fit, list what each lost and stalled account needed that the product does not do, such as an integration, a compliance step or custom setup. If the same gap appears in several lost accounts and in no won ones, write it down as an exclusion even when the excluded group is large. A broad profile that includes accounts the product cannot serve keeps producing meetings that cannot close.
Industry and size terms need to point somewhere specific too. The Census Bureau's Statistics of U.S. Businesses program presents annual data on the number of firms, establishments, employment and annual payroll by geographic area, industry detail and enterprise size. A narrowed profile should be specific enough to point at one industry and one size band in tables like these. If it spans whole sectors and every size, it is still a market description.
A hypothetical diagnosis from pool service routes
The following is a hypothetical example with invented details. It illustrates the method and does not describe a real company, customer or measured result.
A founder sells a route and service log app to field service companies. The written ICP reads: US home service businesses with 5 to 50 employees, owner-operated. After four months of outreach the pipeline holds 40 accounts across plumbing, HVAC, landscaping, cleaning and pool service. Six have bought, nine have been lost and 25 are stalled with no next step. At one of the won accounts, the owner kneels at the edge of a customer's pool, fills a small vial with pool water, adds reagent drops, holds it up to the light to read the color and writes the reading in a notebook kept on the truck dashboard. The founder fills the breadth matrix from the pipeline:
| Term | Won (of 6) | Lost (of 9) | Stalled (of 25) | Evidence | Verdict |
|---|---|---|---|---|---|
| Home service business | 6 | 9 | 25 | Seen on websites | Does not sort |
| 5 to 50 employees | 6 | 9 | 25 | Said on first calls | Does not sort |
| Owner-operated | 6 | 7 | 21 | Said | Does not sort |
| Recurring weekly visits to the same addresses | 6 | 2 | 6 | Seen on service plans | Sorts |
| Posted a field technician job in the 90 days before first contact | 5 | 2 | 4 | Seen on job boards | Sorts |
| Owner approves software alone | 6 | 4 | Not yet asked | Said | Sorts weakly |
| Owner or an office worker builds the week's routes in paid hours | 5 | 3 | 9 | Said; seen at two | Sorts |
| Needs integration with an existing dispatch and invoicing system | 0 | 4 | 10 | Said | Exclusion |
Home service, the size band and owner-operated match nearly every account on both sides, so they describe how the list was built. Recurring weekly visits match all six wins and two of nine losses, and a recent technician posting matches five wins and two losses, which makes them the two strongest terms. The posting is also a trigger a stranger can check on a job board. The paid route-building term is the budget signal. The approval term sorts only weakly and rests on what owners said, so it stays a first-call qualifier. Integration needs appear in four losses and no wins, so they become an exclusion.
Within the recurring-route accounts, four of the six wins are pool service companies, and pool service lost only one deal, against two wins and one loss in lawn care. The founder picks pool service as the beachhead and parks lawn care for a later round. The narrowed statement reads: residential pool service companies in the founder's region that run recurring weekly routes, posted a technician job in the past 90 days, spend the owner's or an office worker's paid hours building the week's routes and do not depend on an existing dispatch and invoicing system, with the owner signing. Of the 25 stalled accounts, three match it and stay active; the rest move to a parked list instead of being chased. The founder writes down the assumptions: that a technician posting signals growth rather than staff turnover, that owners sign without a partner, and that the pool service lead over lawn care is not chance in eight closed deals.
Evidence that the targeted fix is working
The narrowed profile is a hypothesis until the next round of outreach tests it. Before that round, write down what would count as working, then track:
- The list test. Two people build account lists from the filters alone, and most accounts appear on both lists.
- Matched against comparison rates. Run outreach to matching accounts alongside a small comparison group that misses one filter. Matching accounts should reach a meeting and a pilot or payment clearly more often.
- Shorter qualifying. Fewer first calls end with the discovery that the account does not have the problem.
- One story across wins. New customers buy for the same reason, with the same approver, after the same kind of event.
- Weekly records. Daniel's go-to-market page says a session works through choosing what to record each week: conversations, agreed next steps, paid pilots and reasons deals stop. Recording why deals stop is what lets you rerun the matrix.
If matching accounts do no better than the comparison group, the profile was not the main problem, and the next places to look are the offer, the price and the channel. Change the term the evidence contradicts rather than adding new ones, and resist widening the profile again because the matching list looks short. Once a narrow beachhead is working, when to widen it is a separate decision based on the same evidence.
Common mistakes and limits
- Treating a size band as a segment. Employee counts and revenue bands describe how big an account is, not why it would buy now.
- Keeping terms because they are true. Every won account may be a US business, but if every lost one is too, the term does no work.
- Diagnosing from wins alone. Without lost and stalled accounts there is nothing to compare, and every shared trait looks meaningful.
- Using an unseen trigger as a list filter. An event only the buyer knows about belongs on the first call.
- Blaming the profile for an offer problem. If accounts that match every term still lose, look at the offer and the price.
The method has limits. A few closed deals can show a pattern by chance, so every verdict is a hypothesis for the next round, not a finding. It suits products sold to identifiable organizations with an approval path; self-serve and consumer products depend more on usage and purchase data. Public business counts show that accounts exist, not that they will buy. The breadth matrix and its reading rules are an editorial framework, not a measured result.
Related guides and next step
- Choosing among candidate segments: How to Choose Your First Ideal Customer Profile
- Narrowing from customer evidence: How to Turn Customer Evidence Into a Narrow ICP
- Writing the narrowed profile on one page: ICP Template: Buyer, Trigger, Budget, and Current Alternative
- When interviews go well but nobody pays: When Customer Interviews Say Yes but Nobody Buys
- Placing the ICP in a wider plan: How to Build a Startup Go-to-Market Strategy by Stage
Bring your breadth matrix and your list of won, lost and stalled accounts to a free 30-minute strategy session to review your ideal customer profile decision with your actual evidence and constraints.
Frequently asked questions
When is an ICP too broad to guide sales?
When its terms describe how the list was built rather than why accounts bought. Industry, size band and country can match won and lost deals equally, so they cannot tell a rep whom to call. A profile without a dated buying trigger, a named approver and exclusions for budget and delivery fit has the same problem.
What evidence should a founder collect to diagnose an ideal customer profile?
A list of won, lost and stalled accounts, and for each one the industry, the size, the event before first contact with its public trace, who approved or blocked the purchase, what the account paid for its workaround and what the product could not do. Tag each entry as said, seen, done or assumed, and count matches per term.
What are common mistakes in ideal customer profiles?
Treating a size band as a segment, keeping terms because they are true rather than because they sort deals, diagnosing from won deals alone, using a trigger only the buyer can see as a list filter, and blaming the profile when accounts that match every term still lose.
How do you know a narrower ICP is working?
Two people build nearly the same account list from the filters, matching accounts reach meetings and pilots clearly more often than a comparison group that misses one filter, fewer first calls are spent qualifying, and new customers buy for the same reason after the same kind of event.