Why intent data and contact databases miss manufacturing plants

What intent data actually measures

Third-party B2B intent data typically watches reading across a network of publisher sites, matches each visit to a company by IP address or domain, and flags the company when its reading on a topic climbs above its own baseline. What lands in your CRM is a company name, a topic, a score and a week. That is the whole unit.

First-party intent is visits to your own website, matched to a company. Technographic data lists the software detected on a company's domain. A contact database stores people, titles and emails on a company record. Bundle them and vendors call it account intelligence, or sales intelligence for manufacturing when the industry filter says NAICS 31 to 33. The intent data glossary entry has more on the mechanics.

Every one of those records describes a company's online footprint. None of them tells a plant seller what just changed at a specific site, or when.

What triggers a plant purchase, and why intent data sees it late

A trade magazine's 2025 buyer study surveyed more than 250 manufacturing executives and managers and asked what most often starts a significant purchase. The top answer, at 59%, was aging assets or replacement needs, followed by regulatory compliance at 53%, capacity expansion at 52% and technology obsolescence at 44%. Cost savings came in at 39%. All three top triggers start at the plant.

A spindle that keeps throwing the same alarm. An auditor who set a date. A new customer program that needs a second shift. That is where the decision starts, often before anyone opens a browser. When research does follow, much of it starts outside the sites intent data watches: in the same study, 45% said a search engine is their first stop and 23% go straight to a vendor website they already know.

So web-reading intent data sees, at best, the research that comes after the trigger. The trigger itself happened on the floor. The approval path is covered in how manufacturing plants buy.

A plant's operating software rarely shows on its website

Technographic tools read what runs on a company's domain. For a plant, that is usually the marketing site: a content management system, an analytics tag, a contact form. The MES, the maintenance system, the ERP and the machine controls don't show up there, because they live on the shop floor or behind a login.

Often there is nothing to detect at all. One analyst firm estimates that in 2024, 54% of small and mid-size plants worldwide used pen and paper or spreadsheets as their MES, and that just 8% of plants globally run a commercial MES. A clipboard and a PM binder leave no web trace. For an MES seller or a CMMS seller, that clipboard is the real incumbent, and no scanner will report it.

What a seller can act on is the change: an old system reaching end of life, a migration in progress, a new platform turning up on the floor. Clean counts that as a buying moment, a systems change, and dates it.

Many manufacturers barely show up on professional networks

Contact databases and job-change alerts lean hard on professional profiles. That works for software companies, where everyone keeps a profile current. It works less well at a family-owned stamping plant where the owner has never posted and the maintenance manager spends the shift on the floor.

The research habits say the same thing. In the same 2025 study, the main professional network came last of the 12 research sources buyers were asked about, and fewer than 1 in 100 named it as their first stop. Many smaller manufacturers have little or no presence there, so any tool built on that network sees only part of the market.

When the people at a plant keep thin profiles or none, you get no job-change alerts, thin contact coverage and stale titles, and you end up writing to whoever the record happens to hold. Once you have the right name, cold email to plant managers covers what to say.

Most sales data has no field for the machine, the program or the deadline

Most sales data tools store a company, a topic or attribute, and a date. That fits 'this company is reading about ERP.' It has nowhere to put what actually moves a plant: the five-axis machine that just went in, the new automotive program that needs PPAP by a set date, the AS9100 audit on the calendar, the first quality manager the plant ever hired.

Those details are the pitch. A plant that just won an automotive or aerospace program owes a PPAP package or first article reports before it ships a production part, and a topic score can't tell a quality software seller that the clock has started. Clean's database is where that detail lives. It catalogues more than 4,000 early signs across 14 buying moments, and on Clean's own scoring about 7 in 10 are specific enough to fill in exactly those blanks: the machine, the program or the deadline involved.

To be fair, some newer plant databases do list equipment and activity per plant. Ask any of them two questions: when was each fact true, and what is the evidence behind it? Large capital-project trackers have a different gap. They are built around big plants and big projects, so an expansion worth $2 million at a 180-person shop often goes unlisted. More on catching those in new factories and plant expansions.

Multi-site manufacturers get blurred into one account

Intent and contact data resolve to a company, usually through the corporate domain and the headquarters network. Manufacturing is spread across sites. In 2022, the 4,177 US manufacturing firms with 500 or more employees ran 35,367 manufacturing establishments between them, more than eight each on average, and employed 59% of the sector's workers.

When corporate IT reads about MES, the surge lands on the company and every plant inherits it. When the plant two states away adds a line, nothing fires, because nobody there was reading. The plant often trades under a different name than the parent, sits in a different city from headquarters, and may roll up to a holding company with a third name. Several of the changes Clean's research found happened at plants the company's own website doesn't even mention.

That blur costs you, because the buying decision usually happens at the site itself. The plant manager typically owns the problem and, up to a spending limit, the budget. Clean researches manufacturing accounts plant by plant, so each reason to reach out is tied to the building where the work happens.

Intent data vs contact database vs plant-level buying moments

Each answers a different question. Here is what each sees when you point it at a US plant.

Where intent data and contact databases are fine

Keep the contract if it fits. Intent data works when the buyer researches at a desk on the open web: corporate IT at a large manufacturer, security, HR or finance software sold to headquarters, and SaaS-heavy engineering teams. If you sell cybersecurity to the corporate office of a big defense supplier, a topic surge can help with timing.

In our view, visits to your own website are the most useful intent signal in this market. In the 2025 study, 85% of buyers used vendor and supplier websites when researching a complex purchase, the top answer. A plant that lands on your pricing page is telling you something, even if the match to the right site is shaky.

Contact databases earn their keep one step later. Once you know which plant and why, you still need the plant manager's or maintenance manager's name and a working email. Let the database answer who, and let buying moments answer which plant and when. The comparison of manufacturing email lists vs plant buying moments goes deeper on that split.

What to use instead: plant-level buying moments with dated evidence

A buying moment is a dated change at a plant that tends to come before a purchase. Clean breaks a plant's life into 14 of these. The list covers a new site, capacity expansion, a new customer program, a certification push, new equipment investment, a compliance or safety deadline, an ownership change and a first hire in a new function, among others. A slowdown is on the list too, and for most sellers it means hold off. The full set is on the manufacturing buying signals page.

Where a topic score hands you a number, Clean hands you a dated reason with its evidence attached, plus whatever else could explain the change and what would show the reason is wrong. Facts Clean can't confirm are flagged unknown, never guessed. Before a seller sees anything, Clean removes one-person operations, contractors, and repair and service shops.

One moment alone is weak, and so is one week's topic score. Clean lays each plant's history out as dated events in order, then runs a temporal graph network across it, because the order and spacing of changes reveal more than any single event. On top of that, Clean maps 140+ typical chains, the patterns by which one change at a plant tends to set off the next. The aim is to reach the plant before its decision closes, and often before the company has made anything public. See how Clean works. The product page is Clean for manufacturing.

Test your current data against plants you already won

You don't have to take our word for any of this. Run this check in an afternoon with accounts you already closed.

• Pull 20 closed-won plant accounts from the last two years.
• For each, write down what changed at the plant in the months before the first call: a new line, a new customer, an audit date, a new hire.
• Check whether your intent tool flagged the account in that window, and on which topic.
• Check which site the flag resolved to. If it was headquarters, mark it.
• Check whether your contact database had the person who actually signed, with the right title.
• Count the gaps. That is a rough measure of what your current data misses.
• If the gap is big, book a demo. Clean will pull together a live list of plants that fit your product while you're on the call, each one with a dated reason to reach out.