How to automate lead enrichment without filling the CRM with guesses

To automate lead enrichment without filling the customer relationship management (CRM) system with guesses, fill only empty fields, accept only matches above a confidence threshold, and record the source and date of every value. Enrichment adds what a form did not ask for (company size, industry, role, location), so a salesperson opens a record that means something.

Enrichment data arrives with the same confidence whether it is right or wrong. A vendor that returns "250 to 500 employees" for a company with forty looks identical in the record to a correct answer. Deciding how to handle that gap is most of the design.

What we'll cover

• What enrichment is actually for
• Which fields to trust
• Matching, where the errors come from
• Building it with CodeWords
• Seven automations to build
• Your own data beats a vendor's
• Recording where a value came from
• Enriching without asking the visitor for more
• Making it survive
• Limits of enrichment
• What to build first
• Frequently asked questions

Key takeaways

• Leave a field empty when the match is weak, because a blank field prompts someone to check and a wrong one gets acted on.
• Match on the email domain, the most reliable signal, and never trust a company returned for a free email address.
• Enrich only empty fields, so the automation never overwrites a value a person entered.
• Record the source, date and confidence of each field, so you can refresh, audit or undo the enrichment later.
• Join your own product, support and billing data before buying a second enrichment vendor.

What enrichment is actually for

Enrichment is for four jobs, and the job decides which fields matter.

Routing. Sending a lead to the right salesperson by territory, segment, or size. This needs company size and location and nothing else.

Qualification. Deciding whether to work a lead at all. This needs only the fields your qualification criteria use.

Personalisation. Giving the salesperson something to open a conversation with, which recent behaviour serves better than firmographics (company-level facts such as size and industry).

Reporting. Segmenting pipeline by industry or size, which needs consistency more than accuracy: a slightly wrong industry applied uniformly still segments.

Not "more complete records". Filling every field because a vendor offers it produces a record nobody reads and a bill that scales with volume.

Which fields to trust

Reliability differs by field type, so treat each field type differently.

Domain and company identity are the most reliable fields, because the match uses something unambiguous.

Company size and revenue are estimates, banded and often stale. A company that grew or shrank in the last year is commonly wrong.

Industry classification is approximate and differs between vendors, so your segmentation shifts when you switch vendors.

Job titles are what somebody wrote on a profile, which differs from their real role or authority to buy.

Technology detected on a website is useful, but it detects only what is visible and misses anything server-side.

Personal contact details deserve particular care, because accuracy is poor and because the legal obligations differ. The privacy limits are covered below.

Matching, where the errors come from

Most wrong enrichment comes from a wrong match, not from wrong data.

Match on the email domain where you can. It is the most reliable signal available.

Free email addresses cannot be matched to a company by definition, so a vendor that returns a company for one has guessed.

Company names are ambiguous. Several businesses share a name, and matching on it produces confident nonsense.

Ask for a confidence signal and use it. A match below your threshold should be left empty rather than filled.

Empty beats wrong. A blank field prompts somebody to find out; a wrong one is acted on. This is the most important principle in enrichment.

Report the match rate, because a sudden drop means the vendor changed something and your records are getting worse.

Building it with CodeWords

CodeWords connects to 3,000+ integrations (CodeWords pricing page, 2026), so one automation can combine a vendor, your own product data, and your CRM and a CRM.

1. Open CodeWords and start a new automation.
2. Describe what should happen in plain language to Cody, the automation builder: which records, which fields, and what to do when the match is uncertain.
3. Authorize the connections: the enrichment source, the CRM, and your own systems.
4. Describe the exceptions: a free email domain, a low-confidence match, a field that already has a value somebody set by hand.
5. Run it against a small set of records you know well and check the results before enriching anything else.

You describe the outcome; Cody builds it, connects it, and deploys it. The Free plan covers light use. Pro costs $39 per month ($33 per month billed yearly) and Business costs $100 per month (CodeWords pricing page, 2026); details are on the pricing page.

Seven automations to build

Enrich on creation, immediately, so the record is useful the first time anybody opens it rather than after a nightly batch.

Routing from enriched fields. Territory, segment, or size decides who gets the lead. This is the concrete payoff.

Fill gaps only. Never overwrite a value a person entered, because the person had a reason and the vendor did not.

Re-enrich on a slow cycle for records that matter, because company size and technology change and a two-year-old value is a guess.

Flag the unmatchable, so somebody can decide whether a free-email lead is worth pursuing rather than it sitting with empty fields.

Enrich from your own product data (usage, plan, support history). This data is more accurate and more useful than anything purchased.

Report the match rate and the fill rate weekly. A weekly report shows vendor quality dropping before your segmentation does.

Your own data beats a vendor's

Your own data is accurate, current and free, and most teams overlook it.

Product usage tells you more than firmographics. Whether somebody has actually used the thing is a stronger signal than how many employees their company has.

Support history says whether this account is happy, which changes how a salesperson should open.

Billing status is definitive where a vendor's revenue estimate is a band.

Previous conversations with anybody at the same domain are usually the most useful context available. Teams miss them because nobody joins records on domain.

Your own data costs nothing per record, so build that join before buying a second enrichment vendor.

Recording where a value came from

Recording the source of each value makes enrichment auditable, and few teams do it.

Record the source per field: entered by a person, supplied by a vendor, or derived from your own systems.

Record when. A value from eighteen months ago should be treated differently from one from last week.

Record the confidence where the source provides one.

Source records change what you can do. Knowing which fields are vendor guesses lets you re-enrich only those, overwrite safely, and explain a segmentation to anyone who questions it.

Source records survive a vendor change. You can identify and refresh everything that came from the old vendor instead of keeping a mixture.

Enriching without asking the visitor for more

Every field you add to a form loses submissions, and every field you leave off has to come from somewhere else.

Ask for the minimum. Email and name convert; a ten-field form does not, and fields people fill in reluctantly are often wrong.

Derive what you can from what they gave you. The email domain yields the company, and the company yields most of the firmographics.

Ask progressively. A second interaction is a better moment to ask for role or team size than the first, and by then the person has a reason to answer.

Let behaviour do the qualifying. The pages somebody read and the files they downloaded say more about intent than a self-reported job title.

Do not ask for what you will overwrite. A form field that enrichment will replace anyway is a field that cost you conversions for nothing.

Test it. The trade between form length and lead quality is measurable, and most teams assume the result instead of measuring it.

Making it survive

Never overwrite human-entered values. This requires the per-field source record described above.

Handle the vendor being unavailable by leaving the record unenriched and retrying, rather than writing partial data.

Deduplicate before enriching, because enriching four copies of the same company costs four times as much and produces four records to reconcile.

Cap the spend. Vendors usually price enrichment per lookup, so an automation that enriches every record on every change is expensive.

Report the outcome. Records enriched, matched, unmatched, and skipped because a person had already filled the field.

Limits of enrichment

Coverage varies by region and company size. Small businesses and non-English-speaking markets are consistently worse covered, and a vendor's headline accuracy figure does not show this.

Pricing is usually per lookup, so cost scales with volume and with re-enrichment frequency.

Data ages. Company size, technology, and roles all change, and a record enriched once is a snapshot.

Vendors disagree, on industry and size, sometimes substantially. Check before treating either as fact.

Personal data carries obligations that differ from company data, and several jurisdictions require you to tell people what you hold and where it came from.

What to build first

Build enrichment at the point of creation, filling only empty fields, with a confidence threshold below which the field stays blank. This makes the record useful immediately, cannot damage anything a person entered, and keeps confident guesses out of the CRM.

Two habits matter most. Leave the field empty rather than filling it with a low-confidence match. Record the source and date on every enriched field, so you can refresh, audit or undo the enrichment later.

Frequently asked questions

Which enrichment fields are reliable?

Domain and company identity are the most reliable fields. Size and revenue are banded estimates that go stale, industry classification is approximate and differs between vendors, and job titles are what someone wrote on a profile, not their buying authority.

What should happen to a low-confidence match?

Leave the field empty. An empty field prompts somebody to find out, while a wrong one is acted on and spreads into routing and reporting. Set a confidence threshold in the automation, and let any match below it stay blank. This is the most important rule in enrichment.

Should enrichment overwrite existing values?

No, never overwrite a value a person entered, because the person had a reason and the vendor did not. Overwriting an earlier vendor value is fine, which is why recording the source of each field matters. The source record tells the automation which values a person set and which a vendor supplied.

Why do the numbers differ between vendors?

Vendors derive size and industry estimates differently. Check both before treating either as fact. The difference also explains why switching vendors shifts your segmentation, so compare a sample of records from the old and new vendor first. Treat the sample as a test, not as proof.

What is the most underused source?

Your own data is the most underused source. Product usage, support history, billing status, and previous conversations with anyone at the same domain are more accurate than purchased firmographics and cost nothing per record. Build the join on email domain first, then enrich only the gaps.

How do I control the cost?

Deduplicate before enriching, enrich only empty fields, re-enrich on a slow cycle instead of on every change, and cap the spend. Vendors price enrichment per lookup, so an automation that enriches too often is expensive. Track lookups per week in the same report as the match rate.

Are there rules about enriching personal data?

Yes, in several jurisdictions. Rules can require you to tell people what you hold and where it came from. Company data and personal data carry different obligations, so ask your legal adviser before enriching personal data, and keep the source record for every personal field.

How many fields should the form ask for?

Ask for as few fields as convert well, usually email and name. The email domain yields most firmographics, and a field that enrichment replaces anyway costs you submissions for nothing. Ask for more only at a second interaction, when the person has a reason to answer.

What predicts intent better than firmographics?

Behaviour predicts intent better: the pages someone read, what they downloaded and whether they came back. A self-reported job title is what someone typed, while behaviour is what they did. Use behaviour for qualification and keep firmographics for routing and reporting.

Related reading

• Automated lead management
• HubSpot and CodeWords
• Automating sales pipeline updates