The Most Expensive Assumption in Sales

‍

Every company has an Ideal Customer Profile. Most of them are wrong.

‍

Not wrong in the way that's obvious and fixable wrong in the quiet, insidious way that looks like a strategy problem or a messaging problem or a competitive problem, when actually it's a targeting problem. The pipeline isn't converting because you're selling to the wrong companies. The win rate is low because you're winning with the wrong buyers. The churn is high because the customers you're closing were never a genuine fit in the first place. Each of these symptoms gets diagnosed and treated as its own separate issue, when in reality they all trace back to a single upstream cause that nobody has examined.

‍

The cruel part is that a wrong ICP can persist for a long time before anyone names it. Sales blames the product for missing features. Marketing blames sales for poor execution on good leads. Leadership responds to declining efficiency by ordering more outreach volume, which simply accelerates the rate at which you contact the wrong people. The actual problem that the company has been targeting a fictional customer instead of the real one goes unexamined because examining it requires a kind of honesty that's uncomfortable to confront. It means admitting that a foundational assumption, one that everything else was built on top of, might have been wrong all along.

‍

This post is about how to confront it. How to take 30 minutes, run a structured audit of your closed-won data, and walk away with an ICP definition that describes who actually buys from you rather than who you wish would buy from you. We've run this exercise with dozens of SaaS companies across a wide range of categories and stages. Every single one found at least two meaningful surprises. Not minor refinements genuine, strategy-altering discoveries about who their real customer is. Here's how to do it.

‍

Why Most ICPs Are Fiction

‍

Ideal Customer Profiles are almost universally built in the wrong direction. They're constructed top-down: a founder, a head of marketing, or an outside consultant builds a profile of who the product should appeal to based on market research, competitive analysis, and intuition about where the largest opportunity exists. The profile describes the aspirational buyer rather than the actual buyer the customer the company wishes it had rather than the customer it's actually winning.

‍

This isn't inherently wrong as a starting point. Early in a company's life, before there's a customer base to analyze, you have to make educated guesses about who you're building for. You have no closed-won data to learn from, so you reason from first principles: who has this problem, who has budget, who's likely to move quickly. That's a legitimate way to begin. The mistake is treating that initial hypothesis as a permanent truth instead of a starting point to be validated and refined by real customer data as soon as that data exists.

‍

Most companies never do the refinement. The ICP built in year one still drives targeting in year three, even though the company now has 50 or 100 closed-won accounts whose actual characteristics could tell a much more accurate and much more useful story. The data exists, sitting in the CRM, waiting to be analyzed. Nobody has bothered to look at it systematically. So the fictional ICP persists, gets baked into the sales playbook, the marketing campaigns, the lead scoring model, and the territory planning, and the entire GTM motion runs against a target that may bear little resemblance to reality.

‍

There's also an organizational reason the fiction persists: the original ICP usually came from someone senior. Revisiting it can feel like questioning that person's judgment. So the ICP acquires a kind of institutional protection that has nothing to do with whether it's accurate. The way past this is to make the exercise about data rather than opinion. You're not arguing that the original ICP was a bad guess you're simply letting the closed-won data update the hypothesis, which is exactly what good companies are supposed to do as evidence accumulates.

‍

The fix is simpler than most people expect. It doesn't require a data science team, a consultant, or a six-week project. It requires 30 minutes, your CRM, and a willingness to be surprised.

‍

The 30-Minute Audit

‍

You need three things: access to your CRM, 30 minutes of focused time, and the willingness to be surprised by what you find. That last requirement is the one most people underestimate. The entire value of this exercise comes from letting the data overturn your assumptions, and that only works if you go in genuinely open to discovering that you've been wrong about something important.

‍

Step 1: Pull Your Last 20 Closed-Won Deals (10 Minutes)

‍

Open your CRM and find your 20 most recent closed-won accounts. Not the biggest logos. Not the most interesting case studies. Not the ones you're proudest of or the ones that make the best stories in board meetings. The most recent 20, in chronological order. Recency matters because your product and positioning have evolved older wins may reflect a customer profile that no longer applies to what you sell today or how you sell it.

‍

For each account, record seven data points. Company size in employees. Company revenue, if you can access it. Industry and sub-industry and be specific here, because "technology" is nearly useless as a category while "vertical SaaS for healthcare logistics" actually tells you something. Geography. The technology stack they were running, specifically in the categories adjacent to your product. The title and seniority of the person who signed the contract the economic buyer, not the champion who advocated internally. And the specific pain point they articulated as their primary reason for buying, in their own words wherever possible.

‍

That last data point is the most important and the most frequently skipped. It requires going back to actual call recordings or email threads, not trusting whatever a rep typed into the notes field months ago in a hurry between meetings. The verbatim pain language is critical because it becomes the raw material for your messaging, your outbound sequences, and your sales enablement materials. "We can't get reliable contact data for our target accounts" is more useful than "data quality issues," which is more useful than nothing. The closer you can get to the customer's exact words, the more valuable this becomes downstream, because those exact words are what will resonate with the next buyer who has the same problem.

‍

Don't editorialize as you collect this data. This is the discipline that separates a useful audit from a useless one. Don't explain why each account doesn't really represent your target customer, or why that one was a special case, or why this one shouldn't count because the deal came through a warm introduction, or why that one was unusual because of the discount you gave. The temptation to explain away the data points that don't fit your existing assumptions is enormous, and it's exactly what you have to resist. Collect what's true. You can form opinions about what it means later, once you can see the whole picture.

‍

Step 2: Look for the Patterns (10 Minutes)

‍

With 20 accounts documented across seven dimensions, lay the data out where you can see all of it at once a spreadsheet, a whiteboard, whatever lets you scan across every account simultaneously. Now look for clustering. Where are the concentrations? Which values show up far more often than you'd expect if your customers were randomly distributed across the market?

‍

In our experience running this exercise with SaaS companies across a wide range of categories, the patterns that emerge almost always include at least two significant surprises. Not edge cases or anomalies genuine structural patterns that directly contradict the company's stated ICP and that nobody in the building had consciously noticed.

‍

Common surprises we've seen repeatedly: a company that believed it was winning in mid-market discovers it's actually winning most cleanly at 50 to 100 employees, well below where its sales team has been focused. A company targeting VP of Sales discovers that 70% of their closed-won economic buyers were actually CROs a different person, with different priorities, who responds to different messaging. A company that has been positioning broadly across all technology sectors discovers that 80% of their wins are concentrated in two specific verticals they've never explicitly targeted or built messaging for. A company that thought it was winning evenly across geographies discovers that its conversion rate is dramatically higher in specific cities or regions where the founders happen to have stronger networks and denser word-of-mouth.

‍

These discoveries aren't failures. They're gifts. They're telling you where your product actually creates enough value to win a deal which is the single most important thing you can know about your business. Every dollar of GTM spend you direct toward the patterns the data reveals is more efficient than a dollar spent against your assumptions. The data knows more about your real customers than your assumptions do, and this is the moment where you stop arguing with it and start listening to it.

‍

Pay particular attention to the pain point language as you scan across the 20 accounts. Is there a phrase or a framing that recurs across many of your wins, regardless of their size or industry? That recurring language is gold. It's the thing your real customers consistently say is broken before they find you, which means it's the single most powerful opening line you can put in front of the next prospect who fits the profile. The pattern in the pain language is often more actionable than the pattern in the firmographics.

‍

Step 3: Stress-Test Against Your Losses (5 Minutes)

‍

Pull your last 10 closed-lost deals. Run the same analysis across the same seven dimensions. Now compare: where does the profile of your wins differ most dramatically from the profile of your losses?

‍

Look for systematic gaps rather than individual explanations. Is there a company size range where you almost never win, no matter how good the conversation seemed? An industry where your close rate drops significantly below your average? A buyer title that consistently shows up in your losses and rarely drives a deal to close? A technology stack combination that correlates with losses far more than with wins?

‍

These systematic gaps define the edges of your real ICP. They tell you where your product doesn't deliver enough value to win which is just as important as knowing where it does, and often more actionable, because it tells you where to stop spending effort. A company that's losing 80% of its deals against organizations with more than 500 employees has a clear data point. That point should either redirect their targeting away from enterprise entirely, or trigger a serious, honest conversation about what would specifically need to be true in the product, the pricing, the security posture, the implementation model for enterprise to become genuinely winnable rather than a place where good deals go to stall.

‍

The losses also serve as a check against over-reading your wins. If a particular industry shows up in both your wins and your losses at similar rates, that industry isn't actually a differentiator it's just where you happen to be spending time. The patterns that matter are the ones that show up in wins but not losses, or in losses but not wins. Those are the real signal.

‍

Step 4: Rewrite the ICP (5 Minutes)

‍

With the patterns from your wins and the contrast from your losses, write a new ICP definition. One that describes your actual best customers rather than your aspirational ones, grounded entirely in what the data just showed you.

‍

The standard you're aiming for is operational specificity. Your ICP should be specific enough that an SDR could use it to build a targeted list within 30 minutes using nothing but LinkedIn Sales Navigator and your data tools. If it's vague enough to describe half the companies in your category, it's not doing its job, and it won't actually change anyone's behavior because it's not specific enough to act on.

‍

Compare these two examples. Version one: "B2B SaaS companies in the growth stage that are looking to improve their sales efficiency." Version two: "Director or VP of Revenue Operations at B2B SaaS companies with 50 to 200 employees, post-Series A, using Salesforce as their primary CRM, experiencing data quality and contact coverage gaps that are slowing SDR ramp time and making accurate forecasting difficult."

‍

Version one could describe tens of thousands of companies. It's the kind of ICP that feels safe because it excludes almost nobody, but that safety is precisely the problem it gives your team no real guidance about where to focus. Version two describes a specific person, at a specific stage, with a specific problem. Your SDRs can search for version two on LinkedIn with filters and find 200 qualified people to contact by end of day. They cannot do that with version one. That's the difference between a decorative ICP that lives in a slide deck and a functional one that actually directs the daily work of your go-to-market team.

‍

Write the disqualifiers, too. Based on your loss analysis, document who is explicitly not your ICP the company sizes, industries, or buyer titles where you consistently lose. An ICP that only describes who to pursue is half an ICP. The other half is the permission it gives your team to walk away from deals that look attractive on the surface but match the profile of your losses. That permission is worth as much as the targeting itself, because the time your reps don't spend on bad-fit deals is time they can spend on good-fit ones.

‍

Why This Changes Everything Downstream

‍

The ICP isn't just a targeting document. It's the foundation of your entire go-to-market motion, and getting it right has downstream effects that compound over time in ways that are easy to underestimate.

‍

When your ICP is correct, your outbound sequences become more precise because you're writing them for a specific person with a specific situation rather than a vague profile that forces generic copy. Your SDRs waste less time on accounts that will never convert, which means the same headcount generates more pipeline. Your sales cycle shortens because you're entering deals where there's genuine fit rather than trying to manufacture fit where it doesn't naturally exist. Your close rate improves because you're qualifying harder on the front end and only advancing deals that match the profile of your actual wins. Your churn drops because the customers you're closing actually have the problem your product solves, which means they actually experience the value you promised.

‍

Perhaps most importantly, your customer success outcomes improve, because the customers who fit your real ICP experience the value your product delivers more reliably and more quickly than customers who don't. That drives expansion revenue, drives referrals, and drives the kind of word-of-mouth in tight professional communities that no amount of outbound spend can manufacture. A well-fit customer base is a compounding asset. A poorly-fit one is a treadmill where you have to replace churned revenue just to stay in place.

‍

The compounding nature of this is what makes the 30-minute investment so disproportionately valuable. A small improvement in targeting accuracy doesn't just improve one metric. It improves every metric in the funnel simultaneously, and those improvements multiply against each other. A 10% better close rate on top of a 10% shorter sales cycle on top of a 10% lower churn rate produces a business that performs dramatically better than the sum of those individual improvements would suggest.

‍

Validating with the Team

‍

Before you finalize the new ICP and roll it out, bring it to two groups of people who will either confirm it or push back on it in useful ways. The quantitative analysis got you most of the way. The qualitative validation is what makes it bulletproof.

‍

First, your sales team specifically the reps who've been doing this long enough to have developed real pattern recognition about who they're having their best conversations with. Ask them directly: does this describe the accounts where you feel the most momentum? Does this feel like the person who gets genuinely excited when you explain what we do, who leans in rather than politely waiting for the call to end? Where does this feel off based on what you're seeing in the field every day? Your best reps have an intuitive model of the ideal customer that they may never have articulated. This exercise gives them the language to make it explicit, and their input will sharpen the definition in ways the raw data can't.

‍

Second, your customer success team the people who see what happens to customers after they close, long after sales has moved on to the next deal. Ask them: which customers experience the most value? Which ones expand without being asked, simply because the product is working for them? Which ones refer others unprompted? The customers with the best post-sale outcomes are almost always the ones who best fit the real ICP, and your CS team's pattern recognition about this is a valuable input that pure CRM data analysis might miss entirely. Sometimes the best-fit customer isn't the one who was easiest to close it's the one who's still thriving 18 months later, and only CS can tell you who that is.

‍

The ICP that emerges from this process quantitative analysis of closed-won data, contrast against closed-lost data, plus qualitative validation from sales and CS will be more accurate, more operationally useful, and more durably correct than anything assembled in a conference room from hypotheses and market research. And it will still be wrong in some small ways, because ICPs are never perfectly static and markets never stop moving. That's fine. Build in a cadence to revisit and refine it quarterly, using the same closed-won analysis you just ran. Let the data keep leading, let your ICP evolve as your product and market mature, and you'll maintain a targeting advantage over every competitor who built an ICP once and never looked at it again.