CAPTURING YOUR INVISIBLE MARKET

Every revenue leader I talk to is being asked to deliver more growth from channels that are getting harder.

Everyone is targeting the same set of accounts with the same set of channels. The result is spiraling CAC and prospects that never hear what you have to say.

The industry’s answer has been to throw tools at the problem. There are more sales platforms, more AI SDRs, more intent signals and orchestration layers than at any point in the history of B2B. And yet none of it is working. Reply rates are at historic lows. Pipeline is harder to build than it has ever been.

None of it is working because the tools are not the problem.

We believe that the fundamental problem in sales today is that revenue leaders don’t have clarity on what accounts to target or how to allocate resources against them.

This is the first volume of GTM Intelligence, a series where we’ll share what we’re learning from working with revenue teams to define, source & grade their market.

This volume is our first and most foundational piece. On average, B2B companies are missing 60% of the market they could sell to, while 70% of the accounts in their CRM don’t match who actually buys their product at all. We’ll show you why that gap exists, what it’s costing you, and what to do about it.

If you’ve ever had a nagging feeling that your team might not actually be targeting the right accounts… or that you’re missing visibility into large swathes of the market… this one is for you.

Capturing every single account you can sell to is the critical first step in building a revenue engine.

Every decision you make — what accounts to prioritise, what go-to-market strategies to employ, how many reps to hire — depends on having a clear, qualified view of every account you can sell to.

But our research shows that, on average, B2B companies are missing 60% of their market. There is a fundamental mismatch between who companies think they sell to, who they have sold to in the past, the accounts they get from their data provider, and who actually buys their product.

ICP definitions have been built off proxy filters that fail to capture the nuance of who actually buys products; the result is an invisible, uncaptured market of buyers that live outside standard filters.

Your invisible market caps your potential revenue before you even get started. Your TAM model, your territory plans, your hiring forecast… are all built on a foundation that’s missing more than half the picture. There are three reasons why your invisible market exists. Here they are, and how to combat them.

For years, mapping your market meant going to a data provider, accessing the same dataset that every other company could see, and filtering by a combination of employee count, industry and geography.

These three filters were the standard because they were the only data points available. They were simply the only data points that providers could reliably track with enough coverage to know them for every company in their database.

The problem is that these filters were never designed to identify your actual buyers. They were the best approximation available given the data that existed. And now, companies have been using these proxies for so long that many have forgotten what their actual ICP was in the first place.

Ask a CEO who they sell to, and they’ll often say something like: “50 to 500 employee technology companies in North America and Europe.” But that is not actually an ICP — it’s a proxy for one. Their actual ICP might be: “B2B SaaS companies with an established inbound sales motion.” The legacy filters couldn’t capture that, so the proxy became the truth.

70% of accounts in the average B2B company’s CRM do not match their actual ICP. These are the non-ICP accounts that your proxy filters capture by casting such a broad net, or that get added to your CRM by self prospecting and random inbound leads, or that once fit your ICP filters but have slipped out. The result is that 70% of the average B2B CRM is junk — accounts that will never become a deal, but will create noise and consume rep time, marketing budget and management attention.

Capturing your real buyer

Most ICPs are built across three dimensions. These dimensions stay the same; what changes is the data within.

Company size. This has traditionally been proxied through employee count. But employee count is a blunt instrument — it is often selfreported and inaccurate, and it doesn’t tell you what you actually need to know. Depending on your product, “size” might be better captured by the size of specific teams (how big is the sales team? the finance team?), revenue, number of customers, or even web traffic volume. A 200-person company with a 3-person sales team and a 200-person company with a 40-person sales team are completely different prospects if you’re selling a per-seat sales tool.

Geography. This has traditionally been proxied through a single “country” field. But geography is more nuanced than HQ location. It includes where a company’s employees actually sit, where their offices are, where their customers are, and where they’re hiring. A company headquartered in London with 80% of employees outside the UK is a very different prospect to one headquartered in London with 80% of employees inside the UK.

Offering-to-business fit. This is where legacy data has failed most completely. Historically, “does this type of business fit our product?” was proxied through industry. But industry is a spectacularly poor proxy for whether a company needs your product. What you actually need to know is something about how the business operates: do they have an inbound sales motion? Do they process personal data? Do they hire across multiple countries? Do they have a particular tech stack?

Some companies are lucky enough to have clean, binary criteria: “we sell to every eCommerce retailer that uses Shopify.” If that’s you, your ICP translates directly into filters. But most companies have a qualitative ICP: “Companies with an established inbound sales motion,” or “Data-centric businesses that process personal data.” If that’s you, the challenge is finding the right proxies to capture that ICP at scale.

How do you define an ICP without guesswork?

One of the most common mistakes in market mapping is trying to define your ICP qualitatively. You can find thought leaders on LinkedIn who proudly describe how they sat down with a company’s best prospectors and sellers who “know in their gut” what makes a qualified account… but there’s a better way.

Pull a list of qualified accounts (such as your customers), and a list of unqualified accounts (ones you know are not buyers). Enrich them with a deep set of company-level attributes, and run an analysis on what attributes they share. The data will tell you what your ICP actually is — often revealing criteria you hadn’t considered and disproving gut feel assumptions.

If you already have a set of proxies you use to define your market, you can test them the same way: pull your customer list, check how many of your existing customers actually match your current criteria, and see what falls through the gaps.

Once you’ve properly defined your market & worked out the filters that define your ICP, you can run into a second problem: sometimes your ICP can’t be defined by filters at all.

Better data solves most of the “proxy problem.” If your ICP is “companies with a large inbound sales motion,” smarter filters get you a long way: you can look for 10+ person sales teams, 50k+ website visits, a CRM like HubSpot or Salesforce. These are far better proxies than “50–500 employees in Technology,” and for many companies, they’re precise enough to map a market with confidence.

But sometimes there’s a far more elegant solution. “Does this company have an inbound sales motion?” is a question a human could answer in thirty seconds by visiting the company’s website and looking for an inbound CTA. You could try to approximate that through filters — but these are still proxies for a question that a rep could answer at a glance. The problem is that this data doesn’t exist in any structured database. And you can’t send a human to check 500,000 websites.

What are classification models and how do they find hidden accounts?

When trying to answer questions that a human could answer, the natural instinct is to turn to an LLM. You could connect an LLM to a web scraping agent, give it a company’s website and ask: “does this company have an inbound sales motion?” It sounds like the obvious solution, and for a handful of accounts, it might work well enough. But when you need to run that question reliably across your entire market, the cracks begin to show. LLMs are linguistic tools; they read text and interpret it. With grey-area examples, they struggle. At best, LLMs top out at around 85% accuracy on most classification tasks. More commonly they sit around 75%. Think about that in practice: you assign 100 leads based on an LLM classification, and 15 to 25 of them are wrong. If you can’t get your lead assignment to at least 85% accuracy, your reps are going to lose confidence in the leads you give them.

Classification models work differently. Rather than reasoning through language, they learn the underlying quantitative patterns across thousands of labelled company examples. They are trained on a single, well-defined question — “does this company have an inbound sales motion?” — and they learn what the statistical signals of a yes or a no actually look like in the data. The result is that classification models consistently achieve human-level accuracy: 90% or higher.

The training process is a simple loop. You start with a set of labelled examples — say, 20 companies that clearly have an inbound motion and 20 that clearly don’t. The model classifies a new batch, human experts review and correct the outputs, and the model retrains on progressively harder examples. This repeats until the model reaches or exceeds the accuracy of a human reviewer.

Classification models are not always the right tool. They are higher cost than filters, so the decision to use one should be deliberate. The logic is straightforward: if the criteria has a high impact on who is in or out of your market, and the information needed to assess it is publicly available at scale, a classification model is worth building. If it only affects a small percentage of accounts, or the information isn’t reliably available online, a filter or manual check is more practical.

But when the fit is right, classification models let you move beyond proxies entirely. Instead of approximating your ICP through filters, you can answer the actual question — at scale, with human-level accuracy, across every account in your market.

What’s the difference between sourcing and enriching accounts?

There is a fundamental distinction that most revenue teams overlook: the difference between enriching accounts you already know about, and sourcing accounts you’ve never seen.

Most tools in the market — traditional data providers, enrichment platforms, scraping workflows — start from a list of companies you already have and add data to them. That’s enrichment. It makes your existing accounts better, but it can’t show you the accounts that aren’t in your CRM at all.

Sourcing is different. It means building a comprehensive dataset of every company that matches your buyer criteria, regardless of whether you’ve heard of them.

This is why the 60% gap exists. It’s not that companies aren’t investing in data. It’s that they’re enriching what they can see, when the real opportunity is in sourcing what they can’t.

A final and critical point: your market is not static. New accounts are constantly qualifying into your market.

A company hires their 10th sales rep, or opens an office in your region, or launches an inbound motion. Existing accounts become unqualified just as often.

Goodfit tracked over 500,000 unique accounts across our companies’ defined markets. On average, 7% of accounts in a company’s addressable market change every single month. Within a year, the majority of your market looks different from when you first mapped it.

The rate varies depending on how your market is defined. Companies with stable firmographic criteria (geography, industry, size) see 2–4% monthly movement. Companies whose market depends on dynamic signals — technology adoption, team composition, hiring patterns — see significantly more. But no market is static.

If you treated market mapping as a one-off exercise six months ago, a meaningful portion of that market has already shifted underneath you.

This report has described three problems. Your ICP is probably built on proxies that don’t match who actually buys your product. The accounts that match your actual ICP are invisible because they can’t be found through traditional data. And whatever view of your market you do have is decaying every month.

Solving this requires three things:

1. Define your ICP using data, not assumptions. Pull a list of your best customers, enrich them with a deep set of company-level attributes, and let the data tell you what your actual ICP is. Test whatever proxies you’re currently using: how many of your existing customers actually match? What criteria are you missing? What criteria do you think matter but don’t? Everything downstream depends on getting this right.
2. Find every account that matches your criteria. Including the ones that don’t exist in traditional databases. If your ICP can be captured through better filters (team composition, technology stack, web traffic), use them. If your criteria requires answering a question that no structured data field can capture, build a classification model that answers the question at scale, with human-level accuracy. Either way, the output should be a comprehensive, qualified dataset of every account in your market.
3. Keep your market current. When 7% of your market changes every month, mapping is not a one-off exercise. Your market definition should be applied continuously: new accounts qualifying in, stale accounts flagged, contacts refreshed. Your CRM should reflect reality rather than a snapshot from six months ago.

If the numbers in this report gave you pause, there’s an easy way to find out what your own market actually looks like.

Goodfit can build you a market report that shows your true market size based on your actual ICP, how many of those accounts are already in your CRM (and how many you’re missing), the estimated revenue value of the accounts that are invisible to your team, and how many accounts you have that don’t match your ICP and are consuming rep time and marketing budget.

It’s free, and takes minimal work from your side.
You can sign up at the link here.