Understanding GTM Data Enrichment with HubSpot AI

Go to your CRM and check this right now: what's your fill rate across the properties your GTM team actually uses? If it’s decent, kudos to you!

But if you’re seeing substandard fill-rates for properties you use now or will use in the future, you’re essentially feeding incomplete information to core systems.

• Lead scoring hands out a zero where it should have a signal
• Routing sends the lead to whoever is the fallback owner
• A segment built for VPs at mid-market SaaS leaves out everyone whose title field is empty, and nobody notices because the list still looks full
• Reports sliced by industry grow a fat "no value" bucket
• AI agents consistently produce unexpected, unusable, and untrustable output

All of this boils down to one major consequence: you, your teams, and your systems are working on an incomplete picture. There’s genuinely useful data sitting in your CRM, but very little can actually be actioned.

Filling data gaps is exactly what HubSpot’s data enrichment exists for: looking up a contact by name and work email, or a company by domain, and filling the matching properties—all of which can be done in bulk or at a record level.

Why run data enrichment at all?

Because more context runs better GTM processes.

Consider how records got filled before enrichment ran on its own: before enrichment lived inside the CRM, a blank record had a few ways out:

• A rep researched it. Company site, a LinkedIn profile, a news search, a guess at headcount. A few tabs per record, done at the moment the rep was supposed to be selling
• Someone bought or pulled a list. Export a CSV, map the columns, import it. The data was as fresh as the day of the export and aged from there
• The form asked. Job title, company size, industry, phone. Visitors typed whatever they felt like typing, or left
• Ops cleaned up in batches. A quarterly project, a spreadsheet, a lookup formula, nobody’s favorite week

Enrichment was an event. Somebody scheduled it. Between events, the CRM drifted.

Now, enrichment runs on its own, with a purpose.

• It happens at creation. Automatic enrichment fills new records as they’re created, including from forms and integrations
• It stays current. Continuous enrichment refreshes previously enriched records monthly
• It shrinks the form. Form shortening hides the fields HubSpot can fill, so visitors answer less
• It researches per record. Smart properties hand the tab-opening to an AI agent, working from a prompt you write

Here, it's worth noting that as helpful as data enrichment is, ‘faster’ isn’t the same as ‘right.’

Dataset enrichment still fills only what it can match, still leaves wrong values alone under fill-blanks rules, and smart properties spend credits. The rep with the browser tab hasn’t disappeared. The tab just moved to the records the dataset couldn’t fill.

How enrichment software works

To understand how HubSpot data enrichment works, it’s helpful to know how data enrichment as an operation generally works.

Where the data comes from

Vendors fill their datasets through a few channels.

• Public web crawling. Company sites, directories, press pages, professional profiles. A crawler can only find what someone published. Thin web presence means thin data
• Contributor networks. Users connect their inboxes, calendars or CRMs and, by doing so, add their contacts to a shared pool. Some vendors describe a network of 2M+ contributors, alongside email engagement tools, third-party providers, and public sources
• Licensed feeds. Firmographic and intent data bought from partners. Quality depends on the partner and on how often the feed refreshes
• Verification. Automated checks, sometimes with people involved, i.e., a team of human researchers reviews records and defines accuracy on first-party data
• Predictive models. Where no source states a value, some vendors infer one. An algorithm looks at patterns across similar companies and predicts a likely revenue band, headcount range or email format

Why it’s not perfect

No database is perfect (not even close), and it’s usually because the process goes through many filters and has many variables involved.

• Matching comes first. Tools match on a key, usually a work email or a company domain. If there’s no key, there’s no match. Free-mail addresses and missing domains drop out before quality even enters the picture.
• Then there's coverage. Data gets gathered from public footprints, so small, new and offline companies are the likeliest to come back thin, and coverage will differ by region too. That's an inference from how the data is collected, not a measured result. No vendor can tell you your number without running a test on your own records.
• Then there's decay. Fields age at different speeds. A job title goes stale far sooner than a founding year. Published decay figures swing wildly, and plenty trace back to nothing. Lusha measured its own database and found that 12.6% of US sales leaders changed roles in a year. But this counts job changes only, so the real rate of records going wrong is likely higher

The short of it is: An enrichment tool doesn't know your contacts. It never did. Rather, it matches what you send against a dataset someone built, so every gap and every error traces back to how that dataset got assembled and how long ago.

How data enrichment works in HubSpot

HubSpot matches the records you select against its commercial dataset, third-party providers and public information, and updates properties when it finds verified business information across multiple sources.

How it works and what it covers

Contacts are matched on first name, last name, and/or work email. Companies are matched on domain. That's the whole key. Without it, nothing gets enriched.

How a record gets enriched decides what gets overwritten, and the methods don't behave alike.

• Settings: Enrichment is off until a Super Admin turns it on. After that, only records created from that point get enriched automatically, plus existing records you select. Everything already sitting in your CRM won't pick it up on its own
• Mapping rules: You can set the overwrite behavior per property. HubSpot's own examples fill first and last names only when blank, and let job title and seniority overwrite so they stay current. You can map enriched values to custom properties too
• Coverage for contacts: On contacts, the documented property list covers name, job title, role, seniority, LinkedIn URL and location
• Coverage for companies: Covers firmographics including industry, headcount, revenue, funding and web technologies. There are no contact email or phone fields, which is why the work email has to arrive first. A record with nothing new to add is skipped

How data enrichment is used in HubSpot

• Form shortening. HubSpot checks its dataset and shows only the fields it can't fill
• Lead scoring. A complete profile gives scoring something to score
• Segmentation. Job title, industry and company size drive your audience lists
• Personalization tokens. Enriched values fill emails and pages more reliably
• Workflow triggers. Enriched properties can enroll records and branch workflows

For instance, we built outbound automation for an innovation management SaaS, which ran email verification first, then enrichment of name, title, LinkedIn URL and location, then a contact summary from the Data Agent, then personalized outreach from the Prospecting Agent. Their existing inbound flow stayed intact, and the AI layer went in downstream of it.

That’s what enrichment typically is in GTM functions: a middle step. Whatever runs before it decides whether it has anything good to work on, and whatever runs after it inherits the result.

Smart properties are enrichment, too

Dataset enrichment looks a value up. A smart property goes and works it out.

A smart property is a record property that HubSpot fills with the answer to a prompt you write.

You pick the property, prompt the AI, and choose where HubSpot should look. HubSpot’s Data Agent does the research and writes the result into the field. It works on contacts, companies, deals, tickets and custom objects, which is more than dataset enrichment covers.

The data source decides what the answer is built from.

• Web research searches the internet. For companies it starts from the domain. For contacts it uses name, email and the associated company’s domain
• Company website reads the company’s primary domain and needs a company record.
• Property data uses one other property on the same record as the input
• Activities and transcripts use logged calls, emails, meeting transcripts and notes, filtered to the types you choose

That makes smart properties a fit for questions a dataset has no column for. Pre-call research. Whether a company matches your ideal profile. Lead qualification.

💡Pro tip: Unlike dataset enrichment, Smart Properties do consume HubSpot Credits. To save credits, run dataset enrichment first, since it carries no credit charge on paid plans. Then, point smart properties at the gaps it leaves. Spot-check the output before running any campaigns or automation. It’s just faster and cheaper that way.

HubSpot data enrichment best practices

Before you turn it on

• Run the scan for enrichment gaps and write down the match rate and per-property coverage for contacts and companies
• Run the coverage check on a segment you sell into. Segments differ, and a portal-wide number hides that
• Pick the properties that matter. Start with the ones your routing, scoring and reports read, and record their current fill rate. That’s your baseline
• Fix match keys at capture. Ask for a business email on forms where you can, and capture the company domain. No key, no enrichment
• Clean up what’s already wrong. Fill-blanks rules leave wrong values alone, so enrichment won’t fix them
• Check which property your reports read. If they read a custom industry property, map to it before the first run
• Decide who can enrich. Give Data enrichment access to a short list, and know who the Super Admins are
• Set automatic and continuous enrichment per object. Contacts and companies each need their own configuration
• Test on a small segment first. Look at the results before you let anything run on a large segment or your entire database

After it runs

• Open the activity log and confirm what ran, who ran it and on which records. HubSpot’s settings page also covers reverting changes from the log
• Spot-check values. Compare a sample of enriched titles and industries against a source you trust before they feed scoring or routing
• Measure the fill rate again on the same properties you baselined, and compare
• Search your workflows for the Enrich record action, since it can overwrite regardless of your mapping rules
• Decide who owns each field. Once a person edits an enriched property, continuous enrichment stops refreshing it. That suits a title a rep confirmed on a call. It hurts a field nobody meant to take over
• Check the forms. With form shortening on, confirm the fields that disappear are the ones you can live without
• Watch credit use on smart properties in the Manage tab, and tighten prompts that return empty
• Treat opt-out notices as work. Super Admins are notified when a contact asks to be removed from HubSpot’s dataset, and HubSpot offers a setting to action those requests automatically
• Keep a trail if you add a tool. Store a source and a last-updated date beside each enriched property, so a conflict can be traced
• Re-run the scan and a value sample whenever you change a mapping rule, add a tool or see a terms update. The first fill rate you measured only means something if you measure again