Why Firmographic Data Decays and When to Use It

Firmographic data is the set of company-level attributes you use to decide whether an account fits your ideal customer profile. Clay reports the best vendors top out in the mid-80s on accuracy, while contact email clears 95%. The core list is short: company name, industry, headcount, revenue, founding year, location, tech stack, funding stage and corporate structure. Forage.ai and Hans Dekker both list these, and most vendors sell some version of the same set.

Where does it come from? Forage.ai names six common sources: public company websites, business registries, news feeds, social platforms, third-party data vendors, and signals from your own CRM. Every one of those is somebody else's description of a company, written for a different reason than helping you sell. A registry filing exists for compliance. A press release exists for press. Your CRM holds what a rep typed on a busy Tuesday.

That's why two vendors can give you different answers about the same company. Hans Dekker points out that the attributes are standard but the measurement isn't. Employee count is the clearest case. One vendor counts full-time staff only, another adds contractors, another counts the global footprint instead of the headquarters. Revenue splits the same way: audited figures for public companies, estimates for most private ones.

So "firmographic data" isn't one clean thing. It's a bundle of fields with very different reliability, and treating the bundle as one number is how teams end up trusting a filter they shouldn't.

When does this not matter? If you sell to a short, named list of accounts, you don't need a vendor's attributes to decide who fits. You already know. The problem starts when you're filtering thousands of companies you've never researched by hand.

Here's something you can do today. Write down every firmographic field your ICP uses, then mark each one "observable" (industry, headcount, location) or "inferred" (revenue, growth stage). You'll need that split in the next two sections.

Clay's 2026 guide puts firmographic data at the bottom of the B2B accuracy ranking. The best vendors reach the mid-80s. Contact email clears 95%. That gap isn't laziness. Companies don't publish most of this data, so vendors estimate it from incomplete signals, and an estimate has a ceiling.

Then there's time. Forage.ai says static firmographic databases decay at 20-30% per year, and that after 12 months a large share of purchased data is wrong. Companies get acquired, rename themselves, hire, lay off, move and change what they sell. Your CSV doesn't know any of it happened.

Put the two together and the uncomfortable part shows up. The accuracy number on a vendor's sales page describes the day the data was built, not the day you use it. A list that starts at the vendor's best figure is already weaker by the time you've loaded it into your CRM, and the decay compounds every quarter you don't refresh it.

Revenue is the worst field. Hans Dekker says revenue estimates for private companies can be off by 30-50%. Public companies are better because their financials are audited. Most of the companies in a B2B target list aren't public.

What's the strongest counter-case? Decay figures are averages for static databases. A vendor that re-verifies fields continuously will do better, and Forage.ai and others now promote continuously refreshed enrichment as the alternative to a one-off file. That's a fair point, but the sources give no accuracy figure for refreshed data, so don't assume the mid-80s ceiling goes away.

It also doesn't apply if you only use the data once, for a single campaign that ships within weeks of the purchase. Decay hurts the data you keep.

Your next step: ask any vendor when each field was last verified, not just how accurate the file is overall. If they can't say, treat the date as the day you bought it.

Clay draws the line that makes this data usable. Observable fields, meaning industry and employee count, are reliable enough to act as hard gates. Inferred fields, meaning revenue and growth stage, should be weighted in a scoring model and never used as filters.

A gate is a yes or no. A company that fails it disappears from your list and you never look at it again. A score is a nudge. A company with a weaker value on one field can still rise if other signals are strong.

Now think about what a 30-50% error on private revenue does to each. Say your gate reads "revenue above a set threshold". A company that truly clears the bar but whose estimate sits below it is dropped silently. You'll never see it, so you'll never learn the filter was wrong. The same estimate inside a score only lowers that company's rank a little. If it also reads your pricing page, it still surfaces.

Notice the asymmetry. A wrongly included account costs one wasted look. A wrongly excluded account costs a deal you never knew existed, and nothing in your reports will show you the loss.

Industry and headcount aren't perfect either. Employee counts differ by how contractors and global teams are counted, so the usual sensible move is to gate on a band, not a single number. Choose a range wide enough that the counting methods don't push a real fit outside it.

When doesn't this advice hold? When a field is a true constraint, not a preference. If your product can't legally or technically serve a location, location is a gate whatever the data quality. Same if a product physically can't work for companies under a certain size. In those cases accept the lost accounts as the price of the rule.

The decision to make: list your gates, and keep it to two or three observable fields. Move everything inferred into a score, and weight it lower than anything you have seen a company do.

Demographic data describes a person: age, role, location, income. Firmographic data does the same job for a company: industry, size, revenue, location, structure. Mailchimp and Clay both frame it that way, and it's a fair shorthand. If you've ever segmented a consumer list by age band, you already understand segmenting a B2B list by headcount band.

The difference that matters in practice is the unit you're targeting. Consumer marketing starts with the individual buyer. B2B starts with the account, because the purchase is made on behalf of a company and the company's size, industry and budget decide whether a deal is even possible. That's why firmographics comes first in B2B and sits above any single contact.

Don't assume the reliability numbers carry over, though. The mid-80s ceiling and the 20-30% decay figure in the sources are about firmographic data. None of the sources give a matching accuracy figure for demographic data, so there's nothing to compare against, and I won't invent one. What the sources do show is a related contrast inside B2B: company attributes are less accurate than contact email, which clears 95%.

So the practical order is simple. Use firmographics to decide which companies deserve attention, then use contact-level data to reach someone there. Getting the first step wrong is expensive because everything after it inherits the error.

This framing doesn't help if you sell to individuals, such as a consumer app or a freelancer tool. There the account doesn't exist, and demographics is the right lens. It also fits poorly for very small companies, where a founder's own profile and the company's profile are close to the same thing.

If you're unsure which lens you need, ask who signs. If a company signs, start with firmographics. If one individual pays with a personal card, start with demographics.

Here's the gap in most guides on this topic. They stop at the vendor's file. But your website produces a second dataset the vendor can't sell you: which companies actually showed up, which pages they read, in what order, and whether they came back. That's behaviour, and behaviour is a check on the firmographic guess.

The reason it works is simple. A firmographic field says a company fits. A visit says it's interested. When the two disagree, the visit is the better evidence, because nobody estimated it.

You can run this on your own site this week:

1. Pull the companies that reached your site over the last quarter, from whatever company-level view your analytics or identification tool gives you.
2. Mark the ones that read pricing, a comparison page or docs. Those are buying-stage pages. A blog post alone isn't.
3. Note the order. A company that read a comparison page and then pricing is further along than one that read pricing first and left.
4. Mark which companies came back on a different day.
5. Check each marked company against your gate fields. How many pass? How many fail only on an inferred field such as revenue?

Step 5 is the test. If companies that fail your revenue gate keep reading pricing and comparison pages, the gate is throwing away real interest, which is the exact risk the sources describe. Move revenue from the gate into the score. If companies that pass every gate never read anything, your ICP may describe companies that fit on paper and don't want what you sell.

A worked example, as an illustration and not a result: a private company is listed below your revenue line, so a revenue gate would never show it to sales. It reads your comparison page, then pricing, then returns two days later. A score that weights those visits above the revenue estimate puts it near the top of your priority list. A gate would have hidden it.

When does this fail? Low traffic. With few visits there aren't enough companies to see a pattern, and every company name matched from a visit is itself an estimate, so treat each as something to check. Privacy rules also differ between the US and India, and requirements vary by jurisdiction and your consent setup, so use your own legal review before you collect company-level visit data.

PageFox is the wrong choice for a site with little traffic: it tells you which companies are already visiting, it does not bring new visitors.

Gate on industry and headcount, score everything else, and let your own site tell you where the gate is wrong. This week, take the companies that read your pricing page and check them against your gate fields.