Customer Match & Audience Uploads

9 min read

Learn

Quick Reading

Estimated 9 mins

Prereq

Intermediate

Basic ML concepts helpful

Interactive

Static Playbook

Static guide & reference tables

What Customer Match actually does

Customer Match (Google Ads) and its equivalents across platforms let you upload your own first-party customer list — hashed emails, phone numbers, or mailing addresses — and have the platform match it against its own signed-in user base, turning your CRM data directly into a targetable (or exclude-able) ad audience. It's the activation layer for first-party data: the list itself is useless for targeting until it's matched against a platform's identity graph.

Hashing requirements are strict and platform-specific

Every platform requires SHA-256 hashing of PII before upload, but the normalization rules preceding the hash differ subtly and matter a lot for match rate: emails must be lowercased and trimmed; Google additionally strips periods from the local part of Gmail addresses before hashing (since Gmail treats john.doe@gmail.com and johndoe@gmail.com as the same inbox); phone numbers must be in E.164 format with the leading + and country code. Getting normalization wrong doesn't throw an error — it just silently produces no match, because the hash of a slightly malformed string never equals the hash of the correctly formatted one.

normalize_and_hash_email.ts

typescript

import { createHash } from "crypto";

function normalizeEmailForMatch(rawEmail: string): string {
  let email = rawEmail.trim().toLowerCase();
  const [local, domain] = email.split("@");
  // Google-specific: strip dots from the local part for gmail.com addresses
  const normalizedLocal = domain === "gmail.com" ? local.replace(/\./g, "") : local;
  return `${normalizedLocal}@${domain}`;
}

function hashForUpload(rawEmail: string): string {
  return createHash("sha256").update(normalizeEmailForMatch(rawEmail)).digest("hex");
}

Match rates are never 100%, and platforms won't tell you which records failed

Expect match rates typically in the 40-70% range depending on list quality and platform — not every uploaded hash corresponds to a signed-in user the platform can identify, and platforms deliberately don't expose which specific records matched, both for user privacy and competitive reasons. Plan audience sizing and campaign expectations around realistic match rates rather than assuming the full uploaded list size becomes a reachable audience.

Audience upload policy constraints by platform

Platform	Minimum list size for activation	Key policy note
Google Ads Customer Match	~1,000 matched users for most campaign types	Requires account history and policy compliance review; restricted for some verticals
Meta Custom Audiences	No hard minimum, but very small audiences underperform	Hashed upload via Ads Manager or Conversions API-adjacent audience endpoints
LinkedIn Matched Audiences	300 matched members minimum to activate	Strict B2B data-source policy; company-list matching also supported

Refresh cadence and audience decay

A Customer Match list uploaded once and never refreshed decays — customers churn, emails change, and the audience's relevance to current targeting drifts. Treat audience uploads as a pipeline, not a one-time export: a scheduled job (daily or weekly, depending on data volatility) that re-exports the current qualifying segment from the CRM/warehouse and re-uploads via the platform's API keeps the audience reflecting reality, rather than a snapshot from whenever someone last remembered to export a CSV.

Exclusion audiences are as valuable as inclusion audiences

Uploading a list of existing customers to *exclude* from new-customer acquisition campaigns prevents wasted spend showing acquisition-focused creative to people who already converted — an underused application of the exact same Customer Match pipeline most teams only think of for targeting.

What's Next?

Customer Match uploads are the seed audience; from there, most platforms let you expand reach significantly further using lookalike or similar-audience modeling built from that seed, trading some precision for scale.

I build these systems professionally.

Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.

Work With Me →