The ICP Advantage

How do I know if a company is the right fit? An ICP fit check you can run on your own deals

← All posts
ICP fitlead qualityideal customer profileB2B sales

A company can tick every box on your ICP slide and still look nothing like the customers who buy from you. Here's how to check fit against your own won and lost deals, with a spreadsheet.

A lead lands on Monday morning. A 140-person software company in Manchester has booked a demo, and the job title is decent. Your ICP slide says "B2B SaaS, 50 to 500 employees, UK". Every box is ticked.

So is it a good fit?

The slide can't tell you. It was written once, probably in a workshop, and it describes the customers you wanted. Whether this company looks like the customers you actually won is a separate question, and the only place to answer it is your own deal history.

What a good fit means

A company is a good fit when it looks like the companies you have won, on the traits that separate your wins from your losses.

Both halves of that sentence do work. "The companies you have won" means your real closed-won deals, not the persona deck. "The traits that separate your wins from your losses" means a trait only counts if it turns up more often among wins than among losses. If 80% of your wins are UK companies and so are 78% of your losses, being in the UK says nothing about fit. It's just where you sell.

Fit also runs across several traits at once. Most ICP checks in a CRM look at one dropdown at a time: industry is X, employee count sits between Y and Z. Customers rarely split that cleanly. You might do well with small software companies and with large professional services firms, and lose most of what sits in between. One size band can't describe that, and neither can one industry filter.

Working the profile out from the deals you won, instead of writing it from opinion, is what we mean by ICP intelligence. You can do a rough version yourself this week.

How to check if a company is a good fit, step by step

You need a deal export from your CRM (a HubSpot deals view works) and a spreadsheet. There's a longer method for the won side in how to analyse your closed-won deals.

  1. Pull your closed deals, won and lost. The last 12 to 24 months is usually enough. Bring the company properties you fill in reliably: industry, employee count, country, lead source. Aim for at least 20 wins. Below about 10 there isn't a pattern to find.
  2. Clean the fields you'll compare. Put employee counts into bands. Merge industry labels that mean the same thing. Count how often each field is empty, because a field missing on half your deals can't carry much weight.
  3. Count wins and losses for each value. For every value (Software, 50 to 200 staff, Inbound demo request), work out its share of your wins and its share of your losses.
  4. Keep the values with a clear gap. Much more common in wins than losses: a fit signal. About the same in both: drop it. More common in losses: a warning sign.
  5. Check the company in front of you. Go through the kept values and note which it matches. Write the result in words, for example "matches on size and source, differs on industry".
  6. Find its nearest wins. Look for the two or three won deals most like this company. If you can't find any, that tells you something too.

Worked example (illustrative, fictional data). A company with 40 won and 60 lost deals:

Trait value Won deals (of 40) Share of wins Lost deals (of 60) Share of losses What it tells you
Industry: Software 24 60% 18 30% Fit signal
Industry: Professional services 6 15% 21 35% Warning sign
Size: 50 to 200 staff 22 55% 15 25% Fit signal
Country: UK 32 80% 47 78% Nothing (it's where you sell)
Source: Inbound demo request 20 50% 12 20% Fit signal
Source: Partner referral 8 20% 6 10% Fit signal, but small numbers

Back to Monday's lead: software, 140 staff, UK, inbound demo. On this data it matches on industry, size and source, the three traits that separate wins from losses. The UK part adds nothing either way. It's a strong fit, and you can say why in one line.

Now change one fact. Make it a 140-person consultancy. It still matches on size and source, but professional services turns up more than twice as often in these losses as in these wins. Same ICP slide, same ticked boxes, and a very different answer.

Pitfalls that make a fit check lie

Counting wins on their own. Count only wins and every trait that's common in your market looks like a fit signal. The UK row above would read as "80% of our customers are British", which sounds like a finding and isn't. Losses are your control group.

Too few deals. With 15 wins, one big customer and a couple of its friends can make a whole industry look like a pattern. A quick test: take out your two favourite deals and run the counts again. If the pattern disappears, it was never there.

Treating blanks as data. A missing industry is not a different industry. Leave empty fields out of the counts and write down how many there were.

Mixing fit with activity. A company that opens every email and comes to your webinar is interested. That's useful, and it's a separate thing from fit. Add the two into one number and a busy bad-fit company will outrank a quiet good one. We covered how that happens inside HubSpot lead scoring.

Losses that had nothing to do with fit. Some deals die on price, timing or a champion leaving. You usually can't tell which from the CRM, so more deals help more than clever coding of loss reasons.

Doing it once. The customers you won this year may not look like the ones from two years ago. Redo the counts every quarter, or sooner if your win rate moves.

What to do with the answer

Strong fit. Work it first, and give the rep the reason in a line: "looks like our software wins, same size, came in the same way". A rep who knows why a lead matters chases it harder than one handed a number.

Partial fit. Worth a conversation, but find out early whether the missing trait matters. Ask the question your losses would have asked.

Weak fit. Don't bin it on sight. Log it and keep an eye on it. If a few companies like it start closing, your wins are moving, and the counts should move with them.

Wins that split into two groups. If your fit signals point at two quite different kinds of company, you probably have more than one customer type. Check them separately, or the averages will blur both.

Then share the table. Marketing can aim at the traits that separate wins, and sales can see the same evidence, so the "your leads are rubbish" argument has something to stand on. The ideal customer profile template has an evidence column for exactly this.

Where Telepath fits

The spreadsheet method works. It also breaks in the same places every time: it compares one field at a time, each refresh eats an afternoon, and it only checks the companies someone remembers to look up.

Telepath does this comparison for HubSpot teams. The free ICP report reads your won deals and shows what they have in common. In the paid product, Telepath learns from your closed-won deals and gives each HubSpot company a fit score from 0 to 100 (we call it the T-Score): how much it looks like your wins. Where there isn't enough data to score a company, it's marked as not scored rather than given a number. The Companies view lists good-fit companies already in your HubSpot that have no open or won deal. There's more on how that works on the account scoring page.

A T-Score measures resemblance to the deals you've won. It's a fit score, never a forecast.

If you'd like to see which traits your own wins share before you build the spreadsheet, run the free ICP report.

See how Telepath Pro helps your role

Get your free ICP report

Upload your closed-won deals and see which customers you actually win with. Free, takes a few minutes, no account needed.

Get your free ICP report