To create an ideal customer profile that holds up, score every segment of your own customers on three things: how many convert, how long they stay, and one usage signal that predicts both. Then rank the segments on the sales motion you actually run, not on market size.
That is how we rebuilt Scalelist’s ICP in August 2026, when we moved to a phone-led sales motion. Below: the question we asked, the data, the method, what we found, what we changed and the scoring sheet you can copy.
The question: who picks up the phone, and then pays?
The question behind an ICP decides the answer. Our July 2026 ICP asked which customers stay longest. It was scored on retention alone, and it pointed at segments whose buyers rarely answer a cold call.
In August we moved to a phone-led motion, calling prospects directly. So we asked a narrower question: which companies will take a cold call from us, buy, and still be paying months later? An ICP is only ideal for a given motion; the same customers produce a different ICP for inbound than for cold calling.
The data we used
We used our own production data, not a survey or an analyst report. Four sources did the work:
- Accounts. Every account, with the industry and company size it declared at signup. That is the buyer’s own industry, not the industry they prospect into.
- Subscriptions. Start date, cancel date and plan for every paying account.
- Credit transactions. What each account actually did: leads found, emails found, mobile numbers found.
- Live people counts. Scalelist’s own count of how many buyers match each segment today.
We kept only accounts that used credits, and left out the countries we had already stopped targeting. Signups that never touched the product tell you about your signup form, not about your customers.
How to create an ideal customer profile from your data: the method
The method has six steps: define the cohort, measure conversion, measure retention, find the usage signal, rank on the motion, then size each segment.
1. Define the cohort
Keep the accounts that used the product, in the markets you serve. Everyone else adds noise to every rate you calculate.
2. Measure conversion by segment
Conversion is the share of activated accounts that ever paid. Split it by industry and company size band. A segment needs enough accounts to compare; treat a handful of payers as a hypothesis, not a finding.
3. Measure retention by segment
Use the median number of months between subscription start and cancel, and count live subscriptions up to today. Use the median, not the average: a few multi-year accounts would drag an average up.
4. Find the usage signal
Look for one behavior that separates accounts that pay from accounts that do not. Test a few candidates: exports, integrations, a specific feature. Ours was whether an account ever spent credits on a mobile number (20 credits per mobile found).
5. Rank on the motion, not the market
We ranked segments on three criteria, in this order: meetings per hour of dialing, revenue durability (conversion times retention), then market size. The first criterion is judgment, because a database cannot say who answers the phone, and we labeled it that way.
6. Size each segment with a live count
Write each segment’s buyer as one sentence and count the matching people. Sizing a search in Scalelist is free, so this step costs 0 credits. To turn the counts into a market size, use the TAM, SAM and SOM method.
What we found
Five findings reshaped the ICP. The direction matters more than the decimals, so we share the direction here.
Mobile-number usage was the strongest buying signal
Accounts that pulled mobile numbers converted to paid far more often, and stayed much longer, than email-only accounts. That made mobile usage a qualification test, not a feature preference. Our cold-call ICP is not “companies that need leads”; it is companies whose reps dial.
A top-converting segment was the worst one to call
B2B software converted among the best of any industry, but its subscriptions were among the shortest. In our judgment, software sales leaders also rarely take an unknown call. We kept software on inbound and content, where it already converts, and took it off the call list.
Manufacturing had the best mix for phone
US industrial manufacturers with 51 to 500 employees combined strong conversion among mobile users with some of the longest retention in the data. In our judgment, their sales leaders still answer unknown numbers. The weakness: few manufacturing accounts had used mobile numbers yet, so the pitch has to create the phone habit.
Phone-native did not mean a good customer
Finance teams bought mobile numbers readily, then churned faster than any segment with real volume. Freight brokers looked ideal on paper, but too few of them stayed to justify a quarter of dialing.
Retention winners can be slow to reach
Health care vendors stayed the longest of the segments we ranked, but their buyers are the hardest to get on the phone. Great revenue, a poor ratio of dials to meetings, so they ranked third.
The ICP we chose: five segments and one beachhead
The new ICP is five ranked segments, each with a size band and a buyer, and one beachhead we call exclusively for 90 days.
| Rank | Segment | Size | Buyer | US people count |
|---|---|---|---|---|
| 1 | Industrial manufacturers | 51 to 500 | VP or Director of Sales; owner under 200 | 21,000+ sales leaders |
| 2 | Staffing and recruiting agencies | 11 to 200 | Owner, Head of Business Development | 15,000+ owners and executives |
| 3 | Health care vendors | 51 to 500 | VP of Sales, VP of Business Development | 10,000+ sales and BD leaders |
| 4 | Professional services and consulting | 11 to 200 | Founder, Managing Partner | 36,000+ founders, executives and partners |
| 5 | IT services and MSPs, not SaaS | 11 to 200 | Owner, President | 65,000+ owners and executives |
People counts are Scalelist people counts for US companies on 27 September 2026: matching profiles, an upper bound, not verified contacts. Health care vendors were counted as sales and business development leaders at health care and medical device companies.
The beachhead is segment 1: US manufacturers with 51 to 500 employees, sold to the VP or Director of Sales, or the owner below 200 people. It won because the buyer picks up, one good list proves the product, and the account is still paying months later.
For worked profiles of these verticals, see our ICP examples for manufacturing, staffing, healthcare and MSPs.
What we left off the call list
- B2B software: inbound and content only.
- Finance: rejected, because of how fast it churned.
- Freight brokerage: a 100-dial test, not a quarter.
- Construction: only vendors that sell to contractors, not the contractors themselves.
- Agriculture, legal, hardware, fashion and cosmetics: almost none of the accounts that bought mobiles ever paid.
What we changed
The ICP changed who we call, what we say, how we qualify and what we measure.
- Who we call. One segment for 90 days, one buyer title and a sentence we can search. Our cold calling list guide shows how that list gets built.
- What we say. We sell the direct line, not the database. The opening line names the one person their reps cannot reach, such as a plant manager, and offers a small sample list. We do not lead with credits or AI.
- How we qualify. Staffing prospects get three questions: how many recruiters, what data they pay for today, and whether they work on contingency.
- Where software goes. Off the call list and onto inbound and content.
- What we track. Every account is tagged by segment in HubSpot, with four numbers per segment: connect rate per 100 dials, meetings per 100 connects, trial-to-paid rate, and the share of new accounts that pull a mobile number in week one.
That last number is the leading indicator. If trial accounts in a segment are not touching mobile numbers in week one, the pitch is landing as “another email tool” and needs to change before the segment gets written off. The other three belong with your outbound sales metrics.
We also labeled our planning inputs (dials per week, connect rate, meetings per conversation) as estimates, to be replaced with real numbers after week one.
Copy our ICP scoring sheet
Use one sheet per segment, then compare the sheets side by side. Every line has one source: your billing data, your product data, a people count or your own judgment.
2. Accounts that activated (used the product): ____
3. Conversion: accounts that ever paid / line 2 = ____ %
4. Median months retained (live accounts counted to today): ____
5. Usage signal (the behavior that predicts paying): ____
6. Share of accounts showing the signal: ____ %
7. Conversion among accounts with the signal: ____ %
8. Revenue durability: line 3 x line 4 = ____
9. Reach on your motion (phone / email / events): high, medium, low [judgment]
10. Buyer sentence and people count (with date): ____
11. Verdict: beachhead / call next / test with 100 dials / inbound only / never
- Fill lines 2 to 8 from data before anyone argues about a segment.
- Keep line 9 separate. It is the only line where opinion is allowed, so mark it.
- Write disqualifiers from your “never” rows. They become line 10 of the ideal customer profile template.
- Then write the people. Each segment needs one persona per role: see ICP vs buyer persona.
Turn the winning segment into a list
Once a segment wins, write its buyer as one sentence and count it. For our beachhead, “Sales leaders at US manufacturers with 51 to 500 employees” matched 21,882 people in Scalelist on 27 September 2026, at 0 credits.
Scalelist’s Lead Finder shows what it understood from the sentence and the count before you spend a credit. Revealing a verified email costs 1 credit and a mobile number 20, and only when one is found. Scalelist finds and verifies contact details on demand rather than serving a pre-compiled list.
Write your winning segment as one sentence and see how many buyers match before you spend a credit.
Describe your buyer in one sentenceFAQ
How do you create an ideal customer profile from customer data?
Score each segment of your customers (industry, size band, country) on conversion, retention and one usage signal that predicts paying. Rank the segments on the sales motion you run, not only on market size. Write the top segment as your ICP with disqualifiers, size it with a people count, and test it on a pilot list before you commit a quarter.
What data do you need to build an ICP?
Four sources: accounts with their industry and company size, subscriptions with start and cancel dates, product usage (what each account actually did), and a count of the buyers that match each segment today. Keep only accounts that used the product, and exclude markets you have already decided not to serve.
What is a usage signal in an ideal customer profile analysis?
A usage signal is one behavior that separates accounts that pay from accounts that do not. It is found in your product data, not in firmographics. For Scalelist it was pulling mobile numbers; for other products it could be connecting an integration or inviting a teammate. Weight it above market size when you rank segments.
Should your ICP be different for inbound and outbound?
Yes, often. Inbound rewards segments that search and self-serve; cold calling rewards segments whose buyers pick up the phone. In our data, B2B software converted among the best but was the worst segment to call, so we kept it on inbound and built the cold-call ICP around manufacturers, staffing agencies and service firms.
How many customers do you need to build an ICP?
There is no fixed minimum, but each segment needs enough paying accounts to compare rates. With only a handful of payers in a segment, treat the result as a hypothesis and test it with a pilot list. Early-stage companies can start from sales conversations and pilots, then rescore once real customers accumulate.
How often should you rebuild your ICP?
Rescore it every quarter and rebuild it whenever your sales motion changes. We rebuilt ours one month after the previous version because we moved from a retention focus to a phone-led motion. Track a leading indicator per segment from week one so the next rebuild is measured, not argued.
Sources
- Scalelist production data: accounts, subscriptions and credit transactions, analyzed 25 August 2026 (internal, not published).
- Scalelist people count (POST /api/v1/people/count), queried 27 September 2026, 0 credits.
