LinkedIn Matched Audiences: How to Upload a Contact List That Actually Matches, With a Script to Clean It
LinkedIn Matched Audiences let you show ads to people already on your customer or prospect list, people who visited your site, and companies you name. The contact list is where most setups fail: wrong headers, hidden characters, too few matches. Here is the setup as LinkedIn documents it today, and a Python script that turns a CRM export into LinkedIn's template.
LinkedIn Matched Audiences are audiences built from your own data: uploaded contact or company lists, retargeting of site visitors and ad engagers, or a connected data partner.
In Campaign Manager, a contact list is uploaded under Plan > Audiences > Create audience > Matched Audience > Company / Contact, using the CSV template LinkedIn provides.
LinkedIn needs at least 300 rows to upload a contact list, accepts up to 20 MB or 300,000 records, and an ad set needs at least 300 matched member accounts to deliver.
LinkedIn suggests uploading at least 10,000 email addresses to reach 300 matches, and says building a matched audience can take up to 48 hours.
LinkedIn discontinued lookalike audiences on Feb. 29, 2024; existing ones no longer refresh and stay static, and ad sets using them keep delivering to that fixed audience.
LinkedIn Matched Audiences let you aim LinkedIn ads at people you already know about: contacts from your CRM, companies on your target account list, visitors to your website and people who engaged with your ads, forms or Company Page. For B2B advertisers it is the most direct way to reach a short list of buyers instead of a broad job-title audience.
The part that trips people up is the contact list upload. LinkedIn is strict about headers and sizes, and a list that uploads fine can still come back too small to run. This guide follows LinkedIn's help pages as they read on Nov. 13, 2026, all linked at the end. We also wrote a short Python script that converts a CRM export into LinkedIn's template, and checked its email hashing against the sample values on LinkedIn's troubleshooting page. It matched all three.
What are LinkedIn Matched Audiences?
LinkedIn's help page lists three ways to build one:
Lists: a contact list (people, matched mainly by email) or a company list (organizations, matched by name, website and Company Page URL).
Retargeting: people who visited a page on your site (this needs the LinkedIn Insight Tag), engaged with your Company Page, responded to your LinkedIn Event, opened or submitted a Lead Gen Form, viewed a video ad, or engaged with single image, document or conversation ads. Conversions API data counts too; see our LinkedIn Conversions API setup guide.
Data partners: a marketing partner connected to Campaign Manager sends the segments for you.
An ad account can hold up to 1,000 audience segments across matched and predictive audiences. If the Create audience button is grayed out, LinkedIn says to delete an old audience.
How to upload a contact list in Campaign Manager
From LinkedIn's help page for list uploads:
Open your ad account in Campaign Manager.
Click Plan in the left menu and select Audiences.
Click Create audience and choose Matched Audience.
Select Company / Contact, then click Next.
Name the audience, choose the list type, and click Select list to pick your CSV file. The template download sits in the same screen, under "Upload a list."
Click Agree & Upload.
The audience then shows the status Building, and later Ready. LinkedIn says building can take up to 48 hours, in rare cases up to 72. Once it is Ready, add it to an ad set's audience. You can narrow it with attributes such as job function or seniority, but LinkedIn says contact lists can't be combined with member interests and traits.
What does the contact list file need?
LinkedIn's contact template has these headers, in this order: email, firstname, lastname, jobtitle, employeecompany, country, googleaid. The rules from LinkedIn's help pages:
Keep the template's headers. LinkedIn says changing or removing them causes an upload error. The email column header is email, in lowercase.
Each row needs at least one identifier: an email, a first and last name (adding the company helps), or a mobile device ID.
Emails can be plain or SHA-256 hashed. Plain emails are hashed in your browser before LinkedIn saves them. Email is the only field that accepts hashed values.
Save as CSV, in UTF-8 if names have accents or other special characters.
One list per audience. Two lists means two audiences.
What is LinkedIn's minimum audience size?
There are three different numbers, and mixing them up is the most common reason a list "doesn't work":
300 rows is the minimum to upload a contact list.
20 MB or 300,000 records is the maximum.
300 matched member accounts is the minimum for an audience to run in an active ad set. Below that, Campaign Manager shows "< 300 members" and grays out the list name.
Because only part of any list matches a LinkedIn member, LinkedIn recommends uploading at least 10,000 personal or professional email addresses to reach 300 matches. Location is a required targeting setting on every ad set, and it can shrink a list further. If one list is too small, LinkedIn says you can select several audience segments in the same ad set.
Advertisers targeting the European Economic Area or Switzerland should expect smaller audiences: LinkedIn says it matches only members there who have opted in through its consent screen or their ad settings.
A script that turns a CRM export into LinkedIn's template
CRM exports rarely use LinkedIn's headers, and they carry problems you can't see: a byte-order mark at the start of the file, spaces around emails, zero-width characters pasted in from email signatures, duplicates. LinkedIn's troubleshooting page lists hidden characters as a cause of low match rates on hashed lists.
This Python script (standard library only) maps common column names to LinkedIn's headers, lowercases emails and strips whitespace and invisible characters, blanks malformed emails, drops duplicates and rows with no usable identifier, optionally hashes emails with SHA-256, and warns when the list is under or over LinkedIn's limits.
"""Turn a CRM export into LinkedIn's contact list template.Usage: python3 linkedin_list.py export.csv linkedin_upload.csv [--hash]"""import csv, hashlib, os, re, sysTEMPLATE = ["email", "firstname", "lastname", "jobtitle", "employeecompany", "country", "googleaid"]# Common CRM column names, mapped to LinkedIn's headers.ALIASES = { "email": ["email", "e-mail", "email address", "work email", "emailaddress"], "firstname": ["firstname", "first name", "first_name", "given name"], "lastname": ["lastname", "last name", "last_name", "surname", "family name"], "jobtitle": ["jobtitle", "job title", "title", "job_title"], "employeecompany": ["employeecompany", "company", "company name", "account name", "organization"], "country": ["country", "country code"], "googleaid": ["googleaid", "google aid"],}EMAIL_RE = re.compile(r"^[^@\s]+@[^@\s]+\.[^@\s]+$")HIDDEN = re.compile(r"[\s ]")def clean_email(value): return HIDDEN.sub("", value or "").lower()def main(src, dst, do_hash): with open(src, newline="", encoding="utf-8-sig") as f: reader = csv.DictReader(f) lookup = {h.strip().lower(): h for h in reader.fieldnames or []} cols = {} for target, names in ALIASES.items(): for name in names: if name in lookup: cols[target] = lookup[name] break rows, seen, bad, dupes, no_id = [], set(), 0, 0, 0 for row in reader: out = {t: (row.get(c) or "").strip() for t, c in cols.items()} email = clean_email(out.get("email")) if email and not EMAIL_RE.match(email): bad += 1 email = "" has_name = out.get("firstname") and out.get("lastname") if not email and not has_name and not out.get("googleaid"): no_id += 1 continue key = email or (out.get("firstname", "") + "|" + out.get("lastname", "") + "|" + out.get("employeecompany", "")).lower() if key in seen: dupes += 1 continue seen.add(key) if email and do_hash: email = hashlib.sha256(email.encode("utf-8")).hexdigest() out["email"] = email rows.append(out) with open(dst, "w", newline="", encoding="utf-8") as f: w = csv.DictWriter(f, fieldnames=TEMPLATE, extrasaction="ignore") w.writeheader() for r in rows: w.writerow({k: r.get(k, "") for k in TEMPLATE}) size_mb = os.path.getsize(dst) / 1_000_000 print(f"Columns found: {', '.join(f'{c} -> {t}' for t, c in cols.items())}") print(f"Rows written: {len(rows)} (dropped {dupes} duplicates, {no_id} with no usable ID; {bad} invalid emails blanked)") print(f"File size: {size_mb:.2f} MB") if len(rows) < 300: print("WARNING: under 300 rows. LinkedIn needs at least 300 rows to upload.") if len(rows) > 300_000 or size_mb > 20: print("WARNING: over 300,000 records or 20 MB. Split the list.") if len(rows) < 10_000: print("Note: LinkedIn suggests 10,000+ emails to reach 300 matched members.")if __name__ == "__main__": if len(sys.argv) < 3: sys.exit(__doc__) main(sys.argv[1], sys.argv[2], "--hash" in sys.argv)
How we tested it
We built a six-row test export with the kinds of mess CRMs produce: a byte-order mark, the header "Email Address", " [email protected] " with spaces and capitals, an email ending in an invisible zero-width space, a duplicate of the first contact, a malformed email ("not-an-email") on a row that still had a full name, a row with only a company, and "[email protected]" in capitals. The script:
mapped six columns to LinkedIn's headers and ignored the phone column, which the template doesn't use
wrote 4 rows: it dropped 1 duplicate and 1 row with no identifier, and blanked the malformed email but kept that row, since a first and last name with a company is a valid identifier
with --hash, produced exactly the SHA-256 values LinkedIn's troubleshooting page lists for [email protected], [email protected] and [email protected], even though our inputs had spaces, capitals and a hidden character
warned that the file was under LinkedIn's 300-row minimum, which is correct for a test file
Our view: upload plain emails unless your privacy rules say otherwise, since LinkedIn hashes them in the browser anyway. Hash first only when your policy requires that no plain email leaves your systems, and then clean before hashing, because one stray space produces a completely different hash and a lost match.
Can you still create a LinkedIn lookalike audience?
No. LinkedIn's help page says lookalike audiences were discontinued on Feb. 29, 2024. Existing ones stopped refreshing and became static, and ad sets that use them keep delivering to that fixed audience. LinkedIn's help page now lists predictive audiences as a separate audience type, sharing the 1,000-segment limit with matched audiences; their setup is beyond this guide.
What we don't know
LinkedIn doesn't publish an expected match rate for contact lists, and its help pages don't say how long an unused list stays active, only that expired audiences can be reactivated or re-uploaded. Match rates depend heavily on whether a list holds work or personal emails, so test with your own list before planning a budget around it.
Kiran runs Kihan Marketing, a web design, SEO and ads agency, and manages dozens of websites across real estate, home services and healthcare. He writes about SEO, AI search and paid ads news, and what it means for the people who run websites.