Remarketing usually reaches the wrong people because the source audience, exclusion or campaign scope does not match the intended customer state. Existing customers may remain eligible when purchases are not recorded, customer lists are stale or identity matching is incomplete. Irrelevant visitors may enter through broad tags, accidental page views, bots, staff traffic or long membership windows.
Fix the system through Google Ads management, analytics and conversion optimisation. Do not assume one exclusion checkbox can identify every customer across devices and browsers.
Translate the campaign intention into explicit sets
Write the target as set logic before opening the advertising platform. For example: people who viewed service A in the last 30 days, excluding people who completed its enquiry, became a customer, work for the company or visited only support and careers pages.
Define each term. What event proves a service view? Which event proves a valid enquiry? How quickly does a new customer enter the CRM suppression list? Does customer mean any historic purchaser or only a person who currently owns that product? Which locations can the business serve?
Create a remarketing audience specification with inclusion, exclusion, duration, data source, refresh interval and owner.
This written logic exposes gaps that an interface can hide. If the business cannot observe the customer state required by the rule, the campaign cannot exclude it reliably.
Audit where audience membership originates
List every source connected to the ad account: website tags, analytics audiences, app events, video engagement, platform engagement, customer lists and imported conversions. Record the account, property, stream, tag identifier and consent configuration for each source.
Inspect tag firing on representative pages and actions. A site-wide base tag may be appropriate, but the audience should still use deliberate conditions. Check staging domains, payment pages, support centres, staff portals, careers pages, supplier pages and redirects. An overly broad URL condition can include people who never showed commercial interest.
Validate event meaning. A conversion event might fire when a form loads rather than after successful submission, or fire twice after a refresh. A purchase event may fail when a payment provider blocks the return path. A single-page application may not report virtual navigation correctly. Test with browser debugging tools and compare event counts with source systems.
Look for internal and automated traffic. Employees, agencies, uptime monitors, security scanners and test submissions can populate lists. Apply appropriate traffic controls in analytics and avoid visiting live journeys with unfiltered test identities where feasible.
Verify conversion and customer exclusions
Build exclusions from the strongest available evidence. This may combine completed conversion events, CRM lifecycle status, product ownership and recent customer uploads. Keep the logic appropriate to the offer: an existing customer may be irrelevant for an acquisition ad but highly relevant for an upgrade or complementary service.
Google Ads documents that advertisers can exclude first-party data segments at account, campaign or ad-group level in its audience exclusion guidance. Choose the level deliberately. An account-wide exclusion can unintentionally block a legitimate retention campaign; an ad-group exclusion may leave another group exposed.
Audit whether the exclusion is actually attached to every campaign and whether campaign duplication preserved it. Check automated, performance-oriented and multi-channel campaign types for expansion settings and controls that differ from standard display configurations.
Customer-list suppression needs an operating process. Define the export query, hashing or platform workflow, upload frequency, failure alert, retention, access and removal process. Compare source record count, accepted rows and matched audience size without exposing personal information.
Exclusions are not perfect. Google's documentation notes limitations that can affect first-party exclusion behaviour, including some iOS contexts. Identity matching also varies with authentication, browsers, devices, cookies and consent. Promise reduction and monitoring, not absolute certainty.
Review membership duration and recency
A person who explored a service yesterday is not equivalent to somebody who visited a general article eleven months ago. Set membership duration according to the buying window, repeat-purchase pattern, offer and privacy expectations.
Create recency bands where volume allows, such as 1 to 7, 8 to 30 and 31 to 90 days. Tailor message and bid to the likely state. Exclude newer converters from all relevant bands and ensure users age out as expected.
Avoid choosing the platform maximum merely to enlarge the audience. Long windows increase the chance that needs, employment, location or customer status have changed. They also make it harder to distinguish recent intent from old curiosity.
Check timezone and data latency. A customer may see an ad between purchasing and the exclusion update. Measure that delay and choose the fastest reliable path for high-volume or sensitive journeys. Where immediate suppression matters, trigger a first-party update from the transaction system and monitor failures.
Find why irrelevant visitors qualified
Segment the list and delivered traffic by landing page, source, geography, device, time, content group and behaviour. Identify which inclusion condition contributed the largest irrelevant population.
Common causes include:
- all-site visitor audiences used for a specialised offer;
- blog traffic with informational rather than buying intent;
- accidental visits from broad social or display campaigns;
- job seekers, students, suppliers or support users;
- service areas the company cannot cover;
- short, bounced or bot-like sessions;
- audience expansion beyond the selected list;
- shared devices or changing identity signals.
Do not solve every problem with a time-on-site threshold. Time is noisy and can exclude a fast, qualified visitor while retaining an idle tab. Prefer events and page combinations that represent meaningful behaviour, such as viewing a relevant service plus pricing, using a calculator or beginning an enquiry.
Check campaign settings outside the audience list
An accurately defined list can still coexist with broad delivery settings. Review targeting versus observation modes, optimised targeting, audience expansion, location options, network, placements, content exclusions, language and demographic settings where applicable.
Confirm which campaign, ad group and asset group owns the setting. Defaults and controls can change by campaign type, so use the platform's current documentation and inspect the live configuration rather than relying on an old checklist.
Review the actual placement and search-term data where available. Exclude unsuitable inventory and unrelated queries based on evidence. Do not infer audience relevance solely from an aggregate conversion rate, because accidental or low-quality conversions can contaminate optimisation.
Inspect budget allocation. A broad ad group may spend most of the budget before a precise remarketing segment can deliver. Separate campaigns when the audience, message, objective, controls or reporting need independent management.
Protect privacy and consent
Remarketing relies on personal data and identifiers, so align collection and use with applicable law, platform policy and the consent choices presented to the person. Document purpose, vendor roles, retention, user controls and data flows. Obtain qualified legal advice for the organisation's circumstances.
Google's personalised advertising policy places restrictions on sensitive-interest targeting and advertiser data use. Its data collection policy prohibits sending personally identifiable information through tag URLs or similar data collection. Inspect page paths, query strings, titles and custom parameters for names, email addresses, phone numbers or other identifiers.
Minimise data sent to platforms. Restrict access to customer audiences, use secure approved upload methods, remove obsolete lists and maintain evidence of consent or another applicable basis. Ensure opt-outs and account deletion requests propagate to relevant systems.
Do not create audience names that reveal sensitive traits to unauthorised users. Use neutral operational naming and access controls. Privacy review is part of campaign design, not an afterthought after the list has accumulated.
Test the audience end to end
Create controlled journeys for at least these cases:
- a relevant visitor who does not convert;
- a visitor to an excluded non-commercial section;
- a new valid lead;
- a new customer;
- an existing customer from the CRM list;
- an opted-out or non-consenting visitor where applicable;
- a record that changes customer status;
- a failed list refresh or tag event.
Verify events, audience entry, exclusion, refresh timing and audit logs. Audience interfaces may report only estimated or delayed sizes, so combine platform checks with tag diagnostics, CRM counts and controlled observation.
After launch, monitor delivered impressions and clicks by relevant segments. Provide customer-facing staff with a way to record reports of inappropriate ads. Investigate patterns rather than dismissing individual reports as inevitable.
Measure the cost of leakage
Track spend to excluded or low-value contexts where reporting permits. More importantly, measure qualified action, lead acceptance, duplicate lead rate, existing-customer complaints, placement quality and incremental conversion.
Compare customer suppression source counts with platform acceptance and match estimates. A sudden decline can indicate formatting, permissions or upload failure. Monitor audience-size changes against website traffic and conversion volume.
Set alerts for zero or abnormal list updates, tag loss, conversion drops and unexpected audience growth. Give one named owner responsibility for responding, and record configuration changes alongside performance shifts.
A practical repair sequence
Use this order to avoid random changes:
- define the intended audience and exclusions in plain language;
- verify source events and URL conditions;
- reconcile conversion and CRM customer evidence;
- attach exclusions at the correct levels;
- shorten or segment membership duration;
- inspect campaign expansion and location settings;
- test consent, privacy and PII controls;
- run representative journeys;
- monitor delivery quality and suppression latency;
- document ownership and recurring audits.
Change one major layer at a time where possible. Otherwise a temporary improvement may be impossible to attribute and can disappear during the next campaign duplication.
My take: treat remarketing as a changing state model
My take is that remarketing works best when it represents a current customer state, not a permanent list of everyone who once touched the website. Use explicit intent signals, fresh customer evidence, deliberate exclusions and short feedback loops. Accept technical limits, measure leakage and keep the message respectful enough that an occasional mismatch does not damage trust.
Frequently asked questions
Why did a customer still see an ad after purchasing?
The conversion may not have fired, the exclusion may update slowly, the person may use another device or identity, or the campaign may not contain the correct exclusion. Test each layer rather than assuming one cause.
Should all existing customers be excluded?
No. Exclude them from offers that are irrelevant, but include appropriate customers in carefully designed retention, upgrade or complementary campaigns with suitable consent and messaging.
Can remarketing exclude every irrelevant visitor?
No. Signals and identity are imperfect. Improve precision with meaningful events, page combinations, recency, location and first-party data, then monitor actual outcomes.
How often should customer lists refresh?
Match the refresh interval to transaction volume, campaign risk and how quickly customer status changes. Monitor every scheduled update and define a safe response when it fails.
Repair audience quality before increasing spend
If your business needs controlled remarketing and trustworthy measurement, get in touch for Google Ads management and conversion optimisation.
Related guides
For further background, read our google ads ppc management and conversion tracking setup. The roas glossary explains the terminology used in those guides.

