How to Improve Ad Match Rates: A Diagnostic Guide

"Our match rate is low — is something wrong with the sync?" That question comes up constantly, and the answer almost always reveals a misunderstanding about what match rate is and what controls it. Before you can fix your match rate, you need a clear model of what is actually happening when you upload a customer list.

This post is a diagnostic. It starts with the most commonly misunderstood fact about match rates, establishes realistic benchmarks, and then works through the fixes in priority order — from highest impact to lowest.

The Most Important Thing to Understand About Match Rate

Match rate is not a measure of how well your sync vendor performed. It is not a measure of upload quality in the technical sense. It is the percentage of records you uploaded that the ad platform was able to find in its own identity graph.

The mechanics: you send hashed customer identifiers (email, phone, name, etc.) to a platform's API. The platform compares those hashes against hashed versions of the identifiers in its own user database — the accounts created when people signed up for Facebook, Google, TikTok. Where there is a match, the user is added to your audience. Where there is no match, the record is discarded.

The platform controls the identity graph. You do not. Your sync vendor does not.

The practical implication is this: if you gave two different vendors identical customer data, normalized and hashed identically, and had them upload that data to the same platform, they would get the same match rate. Down to the percentage point. The vendor is a pipe. Match rate is a function of what your data looks like against the platform's user base — not the quality of the pipe.

This matters because it redirects your diagnostic effort to the right places: the data you are sending, not the tool doing the sending.

Realistic Benchmarks

Before you can judge whether your match rate is low, you need a reference point. These are realistic ranges, not guarantees — your numbers will vary based on your audience composition, data quality, and platform.

A few observations on these numbers:

B2B match rates run lower than B2C. The reason is structural: ad platforms are built on consumer identity graphs. When a B2B marketer uploads work email addresses (the CRM's canonical identifier), those addresses often do not match the personal Gmail or Facebook account the professional uses in their personal life. This is not a fixable problem — it is a fundamental mismatch between the identifier types.

Match rate is not response rate. A 50% match rate does not mean half your audience is unreachable; it means half were found in this platform's identity graph. Those users may be reachable through other channels, other platforms, or other identifier types.

Very small segments amplify variance. A 1,000-record list with 50% match rate gives you 500 users. That is a viable audience. A 200-record list with the same match rate gives you 100 users — which may fall below minimum audience size thresholds on some platforms and produce unstable performance. The size of the underlying segment matters.

The Fix List, In Priority Order

1. Fix Your Normalization (This Is the Silent Killer)

The single most impactful intervention for most teams is not sending more data — it is normalizing the data they are already sending correctly.

SHA-256 is deterministic: the same input always produces the same hash, and a trivially different input produces a completely different hash. john@example.com and John@example.com hash to completely different values. If you send John@example.com and the platform has john@example.com, there is no match — even though they are the same email address.

The required normalization steps:

• Email: lowercase, trim leading/trailing whitespace
• Phone: E.164 format — +[country code][number], digits only, no spaces or punctuation. (415) 555-0123 must become +14155550123
• Name: lowercase, trim

These rules are simple but fragile at scale. A CRM with inconsistent data entry, or a sync built by someone who did not read the normalization spec, will silently fail on a percentage of records that is impossible to detect without auditing the pre-hash data directly.

If your match rate is lower than expected and you have not audited your normalization logic, start there. A detailed walkthrough of why normalization matters more than most teams realize is in the customer match audience hashing explainer.

2. Send More Keys

Email-only uploads are the minimum. Every additional identifier you send gives the platform another opportunity to match a user whose email you have but whose account uses a different email address.

The priority order for additional keys:

Phone number (highest impact after email). Many users have updated their phone number in platform accounts when they have not updated their email. Phone in E.164 format, hashed, is often the second-highest match signal after email.

First name + last name. On their own, names are too ambiguous to match. Combined with email or phone, they raise the confidence score of a potential match and can salvage records where only partial data was correct.

Postal/ZIP code + country. Lower individual signal, but useful in combination. Country is required for phone matching on several platforms.

Mobile advertising IDs (IDFA, AAID). For mobile-heavy audiences, device identifiers are strong signals. These are not hashed the same way as PII — they are sent as-is on most platforms. Availability is declining with ATT opt-outs, but for opted-in users they remain a high-quality match key.

If your current integration only sends email, adding phone and name to the payload is usually the fastest way to move the match rate number.

3. Freshen Your Data

Stale data has lower match rates for a structural reason: people change their email addresses, deactivate accounts, and update profile information over time. An email address that was a valid, active account two years ago may no longer exist in the platform's identity graph.

The freshness problem compounds when you use periodic batch uploads — monthly or weekly exports — instead of continuous sync. By the time a monthly export is uploaded, some records are already outdated. Records that have been in the database for years and never refreshed are the most likely to fail matching.

The fix has two parts: use continuous or high-frequency sync instead of batch exports, and prioritize segments of recently active customers over old, dormant ones when match rate matters for campaign performance.

4. Clean Segment Hygiene

Not all records in your CRM are matchable. Some will never match regardless of how well you normalize or how many keys you send:

Role accounts and aliases. Addresses like info@company.com, noreply@company.com, admin@domain.com, or team@agency.com are not personal email accounts. They are not in any platform's identity graph. They consume upload quota and drag down match rate without contributing anything. Remove them before syncing.

Hard bounced addresses. An email address that has hard bounced is almost certainly invalid. It will not match anything. Maintaining a suppression list of bounced addresses and excluding them from audience uploads is basic hygiene.

Duplicate records. If your CRM has the same email address stored under multiple customer records (a common data quality issue), you may be sending duplicates. Platform APIs typically deduplicate on their end, but it is cleaner to deduplicate before upload.

Very old records. Accounts from customers who have not interacted in five or more years are unlikely to be in active ad platform accounts. Defining a recency cutoff for audience eligibility — say, customers with at least one activity in the last three years — often improves match rates by trimming the least-matchable records.

5. Choose the Right Platform for the Segment

Match rate is partly a function of which platform you are targeting. Your customer base may have substantially higher match rates on some platforms than others, depending on demographics and platform usage patterns.

If you have a customer base that skews older, Meta and Google are likely your highest-match platforms. A customer base with higher representation of Gen Z users may match better on TikTok. B2B audiences that use LinkedIn professionally may match well there on email even though LinkedIn match rates are variable.

Testing the same segment across multiple platforms before committing budget gives you a baseline. The SFMC audiences to TikTok Ads guide covers platform-specific considerations for TikTok in detail.

When Low Match Rate Is Acceptable

Not every audience needs a high match rate to be effective. The relationship between match rate and campaign value depends on the use case.

High-LTV suppression lists. If you are suppressing your active, high-LTV customers from prospecting campaigns, a 40% match rate still protects you from showing acquisition offers to nearly half your best customers. Even an imperfect suppression list saves real budget. The ROI calculus is favorable.

Winback campaigns for churned customers. You are spending to re-acquire someone who already knows your brand. Even if only a third of your churned customer list matches, those matched users are high-quality targets. The effort is proportional to the value.

Small retargeting segments. This is where low match rate genuinely hurts. If you have a retargeting segment of 500 high-intent users and only 200 match, you are left with an audience that is too small for meaningful optimization. Minimum audience thresholds on most platforms (typically 1,000 users) mean you may not be able to run the campaign at all. Here, match rate is a direct constraint on campaign feasibility.

The diagnostic question: what is the minimum viable audience size for this campaign, and does your expected match rate produce a segment above that threshold? If yes, your match rate is fine. If no, the prioritized fixes above are your path to viability.

Measuring Improvement Correctly

A common mistake is comparing match rates across different time periods without holding everything else constant. Match rate is affected by:

• The composition of the segment (different mix of customers)
• Platform changes to the identity graph
• Normalization changes
• The number of keys sent

To measure whether a specific change improved your match rate, you need to test the same segment, sent the same way, with only one variable changed. Re-uploading the same segment with a corrected normalization for phone numbers, for example, gives you a clean before/after comparison.

Document the baseline first: for a given segment, which keys were sent, how were they normalized, and what was the match rate? Then make one change at a time.

Match rate improvement is also not purely a match rate story. The end goal is campaign performance — reach, conversions, ROAS. A higher match rate that creates a larger audience is only valuable if that larger audience converts at a comparable or better rate. Match rate is an input, not an output.

Where Cezium Fits

For SFMC teams, Cezium Ads automatically sends the richer key set from your Data Extensions to each platform's audience API — not just email, but phone (normalized to E.164), name, and additional identifier fields where available. Audiences stay fresh through continuous sync, so you are not working from a list that was accurate when it was exported three weeks ago. The normalization logic is applied consistently within your own MC instance before anything is hashed or sent. If you have been working from periodic exports, the move to continuous sync alone typically improves effective match rates because the data reflects your current customer state rather than a historical snapshot.

Match rate is a lagging indicator of data quality decisions you made upstream: which identifiers you collected, how consistently you stored them, how recently you refreshed them, and how carefully you normalized them before hashing. The platforms are not going to change their identity graphs to match your data model. The work is on your side of the equation, and it starts with normalization.

Mounir Nejjai is the founder of Cezium.