How to match HVAC contractor data to your CRM

A useful match process separates existing-record matches, possible new companies or locations, and rows too ambiguous to classify. Preserve that uncertainty rather than forcing every row into “match” or “new account.”

1Choose the record unit before match keys

Decide whether the CRM record represents a legal entity, brand, headquarters, branch, or service location. Decide separately what one incoming row represents. A correct company-level match can still be a wrong location-level match.

Document parent companies, trade names, franchises, multi-location operators, and multi-trade businesses. Keep the source record and proposed CRM record distinguishable throughout review.

2Use several comparison fields

Common firmographic keys include business name, website domain, business phone, street address, city, state, and ZIP code. No single field is universally decisive: a phone may serve several locations, a domain may cover a parent, and an address may have changed.

Request only fields needed for the test. The HVAC data-fields guide explains the proposed fields and limitations. Directory fields do not establish a person, consent, intent, deliverability, or permission to contact.

3Preserve raw values when normalizing

Keep each unchanged source value beside any normalized value. Normalization may standardize capitalization, punctuation, company suffixes, phone formatting, URL schemes, state codes, and ZIP formatting. Keep ZIP codes as text so leading zeroes survive.

Normalization makes values comparable; it does not prove either value is current or correct. Record the normalization version so the process can be reproduced.

4Define confidence and conflict rules

Write decision rules before inspecting results. A team might accept only candidates supported by multiple independent fields, review weaker candidates, and leave conflicts unmatched. Actual rules should fit the buyer’s CRM and risk tolerance.

Use explicit outcomes: accepted match, possible match—review, no match, and unresolved. Do not silently break ties or overwrite values when name, domain, phone, and location evidence disagree.

5Review one-to-many and many-to-one cases

One incoming company may correspond to several CRM locations; several incoming rows may correspond to one account. Queue these cases with supporting and conflicting fields.

Keep matching separate from merging. A probable match does not authorize deleting duplicates, replacing buyer-maintained values, reassigning ownership, or creating accounts.

6Build a labeled test set

Select a reproducible sample from the exact segment. Have an authorized reviewer label expected outcomes using agreed evidence and a fixed review date. Report denominators, unresolved rows, and exclusions.

Measures can include confirmed accepted matches, recovered reference matches, unresolved rate, and incorrect company- or location-level links. Do not combine different measures into an unsupported universal “match accuracy” claim.

7Set import and rollback controls

Test in a sandbox or reversible import path where possible. Assign a batch identifier, preserve source lineage, restrict ambiguous-match approval, and define how to reverse created links or records.

Share only the minimum buyer fields needed through an agreed secure method. Record access, retention, and deletion expectations. Never place CRM exports or credentials in an ordinary quote form.