A "parents in 90210" card campaign is the textbook failure
Picture a card issuer building an in-market audience: parents, ages 30-45, married, living in one high-income postal code, excluding anyone with thin credit files. Every element is plausible marketing; in the US and Canada, most of it is off-limits for personalized advertising under Google's policies. The policy language groups credit-related products and services with banking products and services and certain financial planning and management services, then names the signals that may not be used to select or target personalized ads: gender, age, parental status, marital status, and postal code. The examples of covered credit-related products are explicit — credit cards and loans, bank and checking accounts, debt management products. Income, occupation, and stated interests are absent from that list. The restriction is a closed list of demographic and geographic descriptors attached to a broad product category, not a ban on card advertising.
The category, not the creative, triggers the restriction
Two misreadings waste most revisions. The first is thinking the rule judges your ad copy; it judges the audience-selection inputs instead. A factual card offer targeted by marital status fails; a blunt rate comparison without those signals does not, on this policy alone. The second is treating "credit card" as a narrow label. The policy's example list folds cards together with loans, deposit accounts, and debt management products, so pairing a rewards card with balance-transfer messaging keeps the campaign inside the same category regardless of media-plan splits.
The restriction is also geographically scoped. The policy states the US and Canada limitation for these categories specifically, so an audience built for a compliant Canadian or US buy cannot be copied into another market and assumed compliant. Multi-market card programs need a separate targeting matrix per market.
Financial vulnerability is a separate, data-type ban
The demographic list is not the only barrier. Personalized ads also may not use inferred or actual sensitive information, and the policy's own example is financial: information indicating a low credit rating or high debt burden, plus other detailed personal financial information. That prohibition is structured differently from the credit-category rule. It concerns a data type, not a product type, and sits in a list that also covers health records, ethnicity, religion, criminal history, political affiliation, union membership, and sexual behavior.
For a card campaign, an "improve your credit" audience built from credit-rating signals fails on a second, independent ground even after age and postal code are stripped out. Audit the seed data, not only the interface checkboxes: CRM imports, lookalike sources, and list partners can reintroduce exactly the attributes the policy names.
Documentation the same policy expects alongside targeting
Compliance here is not only subtraction. If you use Google products or ad code for personalized advertising, the policy expects that you hold all necessary rights to use audience data such as cookie lists; that the advertiser add an in-ad notice stating the ad is interest-based, for example through the Ad Options icon; and that all parties follow applicable industry self-regulatory codes, such as the Digital Advertising Alliance's Self-Regulatory Principles for Online Behavioral Advertising or IAB Europe's framework in Europe. Separately, publisher privacy policies must disclose data collection, sharing, and use arising from Google products, including cookies, web beacons, IP addresses, and other identifiers. Under-13 activity and child-directed properties are carved out of interest-based and remarketing targeting entirely.
Treat that paperwork as part of the launch path; it is easier to assemble before creative is final than after a review query.
How Google's compliance examples classify financial promises
Google's AFS compliance materials describe two grades: a normal violation, where five equal one strike, and an egregious violation, where one equals one strike. Egregious cases are marked by significant scale of impact, a notable failure to follow best practices whether intentional or negligent, particular deceptiveness toward users or Google, or advertiser harm. Among the listed egregious examples are promising loans and a specific promise to open a named high-yield savings account "with no credit check." Normal examples are milder: vague promises, or information offered but never delivered.
Read this as a classification signal rather than a schedule that applies identically to every advertiser. For card copy, a promise you cannot control — guaranteed approval, a fixed APR for every applicant — sits higher; an imprecise but deliverable statement sits lower.
A review sequence that catches the usual misses
- Classify the offer against the policy's example list before briefing creative; cards, loans, deposit accounts, and debt products share one category.
- Inventory audience inputs at every layer — platform segments, first-party lists, partner files — and flag age, gender, parental status, marital status, and postal code for US and Canada buys.
- Screen for credit rating and debt-burden signals separately, since that failure can survive a demographic cleanup.
- Confirm data rights for cookie-list audiences, the interest-based ad notice, and the self-regulatory code your buy falls under.
- Review privacy disclosures for cookie, beacon, and IP-address language.
- Stress-test claims for promises outside your control.
- Keep a written record of which signals were removed and why, for the next compliance review.
What this policy does not answer
The policy lists prohibitions, not permissions; a signal not named in it is not therefore acceptable. It does not address consumer-finance disclosure duties, fair lending obligations, or state-level advertising rules that may apply to card offers; those lie outside Google's policy and belong with qualified legal counsel in each market. Scope matters too: the restriction addresses personalized advertising, so the first question for any placement is whether it is personalized at all, and EU markets carry a consent framework US buys do not.
Start by pulling one live card campaign and listing its audience inputs next to the five named signals. That single comparison usually shows whether the real problem is creative, data, or geography.
研究の知恵