Display Remarketing and Dynamic Remarketing: Lists, Feeds and Sequencing

Display remarketing and dynamic remarketing in 2026: lists, feeds and sequencing

Remarketing was once the easy lever in paid media. You dropped a pixel, built a “all visitors” list, ran a static banner with a 10% coupon, and watched return on ad spend climb. That world is gone. Between iOS 14 (2021), the deprecation of third-party cookies in Chrome announced and walked back several times by Google between 2020 and 2024, the rollout of Consent Mode v2 in March 2024, the sunset of Similar Audiences in May 2023, and the shift of Google Ads spend toward Performance Max (which according to Google handled more than 80% of new Display budgets by late 2025), the remarketing playbook of 2018 simply does not run anymore.

What still works in 2026 is a layered architecture: clean first-party data, granular audience segments, a properly governed product feed, frequency caps that respect human attention, sequencing that matches awareness, decision and win-back creative to where the user actually is, and a Performance Max audience signal layer that feeds Google’s machine learning the right examples. This guide walks through the entire stack, with the numbers, the limits and the audit cadence we use on LaFactory.com client accounts.

The distinction between display remarketing (the same banner to a behavioural segment) and dynamic remarketing (the exact product the user looked at, with current price and stock) still matters, but in 2026 both live inside a wider ecosystem: Customer Match for CRM lists, video remarketing on YouTube, and audience signals on Performance Max. Treating them as one toolbox, not as separate channels, is what separates a 1.5x ROAS account from a 6x ROAS account.

Why remarketing still matters in the 2026 audience-attrition landscape

Three forces shrunk classic remarketing audiences between 2021 and 2026. iOS 14.5’s App Tracking Transparency cut mobile pixel match rates roughly in half on iOS traffic, according to the Tinuiti Q2 2021 Digital Ads Benchmark Report. Apple’s Intelligent Tracking Prevention in Safari capped first-party cookie lifetime at 7 days as early as 2019, well before Chrome’s slower moves. And Consent Mode v2, which Google made mandatory for EEA traffic on 6 March 2024, means a meaningful share of European visitors land on your site without consent for advertising cookies, which excludes them from standard remarketing lists unless you implement the “advanced” Consent Mode that allows conversion modelling from anonymous pings.

The practical effect: a French e-commerce site that had a 90-day “all visitors” list of 600,000 users in 2020 now typically sees 150,000 to 250,000 in 2026 for the same traffic volume. That is not a tracking bug. It is the new baseline.

Remarketing still matters, and arguably matters more, because the cost of cold prospecting climbed in parallel. Search Engine Land’s 2025 Q3 PPC benchmarks showed average CPCs up 19% year over year on commercial intent queries in retail. When prospecting gets more expensive, every visitor you already paid for becomes a more valuable asset. The teams winning in 2026 are not the ones running more remarketing. They are the ones running cleaner remarketing on smaller, better lists.

WordStream’s 2024 Google Ads benchmarks also reinforced a long-standing pattern: remarketing campaigns convert at 3x to 10x the rate of cold display, depending on vertical, and cost per conversion runs 30% to 50% lower. That delta is the budget you are leaving on the table when remarketing is treated as a checkbox rather than a discipline.

Audience list types in Google Ads 2026

Website visitors (the foundational list)

Built from the Google tag (gtag.js) or Google Tag Manager firing on your domain. The tag captures page URL, optional event parameters, and (in dynamic remarketing) the product ID viewed. In 2026 the standard configuration sends a Google Ads remarketing tag plus a Google Analytics 4 property tag, with audiences defined inside GA4 and exported to Google Ads through the linked-account flow. According to Google Ads Help, audience exports from GA4 take up to 24 hours to populate, and ads will not serve to a Display audience until the list reaches the minimum of 100 active users on the Display Network.

For Search remarketing (RLSA), the minimum is 1,000 cookies on the Search Network. For YouTube, the minimum is 1,000 active users on YouTube. These thresholds, published in the Google Ads audience requirements documentation, have not changed in 2026 and they are a hard floor, not a soft signal. A list with 800 users on Display will simply not deliver impressions.

App users (Firebase and Google Analytics 4)

For mobile apps, audiences are built in Firebase or in GA4’s app data stream and pushed to Google Ads through the linked Firebase project. You can target users who installed but did not complete onboarding, users who opened the app more than 7 days ago, users who reached a specific in-app event (added to cart, completed level 5, started a free trial). App remarketing runs through App campaigns or, increasingly, through Performance Max for app inventory.

The 2026 wrinkle is that on iOS, audience size depends entirely on ATT consent. Apps that do not explicitly request and receive ATT permission cannot pass the IDFA, which means the user is invisible to Google’s app remarketing graph. Most serious app businesses now rely on SKAdNetwork conversion modelling for iOS rather than deterministic app remarketing, with deterministic targeting reserved for Android.

Customer Match (the CRM-powered list)

Customer Match lets you upload hashed customer data (email, phone, address) directly to Google Ads. Google matches it against signed-in Google accounts across Search, YouTube, Gmail, Display and Discover. As of 2024, Google removed the previous policy minimums on advertiser spend for Customer Match access, making it available to virtually any account in good standing.

Match rates depend on data quality. Google’s documentation reports that well-formatted email lists typically achieve 40% to 70% match rates. Hashed phone numbers in E.164 format usually outperform email in B2C verticals where customers gave their phone for delivery. The richer the input (email + phone + name + postal code, all hashed with SHA-256), the higher the match.

Customer Match is the most resilient audience type in 2026. It does not depend on third-party cookies, it survives ITP and ATT, and it carries first-party signal that Google’s bidding algorithms weight heavily. Tinuiti’s 2024 retail benchmarks showed Customer Match audiences delivering 2.4x the ROAS of cookie-based website lists in Q4 2024 holiday campaigns.

Video remarketing (YouTube)

You can remarket to people who watched specific videos, watched any video on your channel, subscribed, liked or commented. The list is built automatically once you link your YouTube channel to your Google Ads account in the “Linked accounts” section. Audiences include “viewed any video”, “viewed certain videos”, “viewed any video as ad”, “subscribed to your channel”, “visited your channel page”.

Video remarketing is underused. According to Search Engine Journal’s 2024 video advertising trends report, fewer than 30% of advertisers running YouTube ads also run a remarketing layer back to video viewers. That is leaving warm intent on the table. A user who watched 50% of a 60-second product demo is meaningfully more qualified than a user who clicked a banner.

Similar Audiences (deprecated, what replaced them)

Similar Audiences, Google’s lookalike feature for remarketing seeds, were deprecated on 1 May 2023 and fully removed from campaigns by 1 August 2023, as announced in Google Ads’ April 2023 Help Center post. The replacements are optimized targeting on Display and Discovery, and audience signals on Performance Max. Both work the same way conceptually: you provide a seed (your Customer Match list, your converters list, your high-LTV segment), and Google’s algorithm finds users who behave similarly across its inventory.

The difference from the old Similar Audiences is that you no longer get a discrete, addressable audience to bid on or report on. The audience is implicit inside the campaign’s optimization. This shifts the workload: you spend less time managing similar-audience lists, and more time curating high-quality seed lists.

List sizes, minimums and the 100-user trap

Display Network minimum: 100 active users in the past 30 days. Search Network (RLSA): 1,000. YouTube: 1,000. Customer Match: 100 matched users for Search and Shopping, 1,000 for Display and YouTube.

“Active” means the user has been seen in the network in the last 30 days. A list of 50,000 cumulative users that has had only 80 active in the last 30 days will not serve on Display. This is a common cause of “my campaign has zero impressions and I do not understand why”. The Audience Manager in Google Ads now shows both list size (cumulative) and active size per network, but you have to expand the columns. Most accounts we audit do not have those columns visible.

The 1,000-user minimum on Display for Customer Match (and the 1,000 on Search for RLSA) is the practical floor for any meaningful campaign. Below that, even if the list “works”, optimization is unreliable because conversion volume is too thin for Smart Bidding to learn. Search Engine Land’s October 2024 PPC column on remarketing health-checks suggested 5,000 active users as a working minimum for any list you want to optimize with Maximize Conversions or Target ROAS.

Membership duration is the other lever. The maximum is 540 days for website lists and 540 days for Customer Match. The default is 30 days. Most lists should not run at the maximum: a user who visited 18 months ago is a different person, in a different financial situation, possibly with a different need. Set duration to match your real consideration cycle: 7 to 14 days for cart abandoners, 30 to 60 days for product viewers, 90 to 180 days for past purchasers, 540 days only for very high-LTV B2B sales cycles.

Dynamic remarketing setup with Merchant Center

Dynamic remarketing for retail uses your Google Merchant Center product feed as the source of truth. The Merchant Center feed already exists if you run Shopping or Performance Max for retail, so the marginal cost of activating dynamic remarketing is low.

The setup, per Google Merchant Center documentation, is four steps: link Merchant Center to Google Ads in the Linked Accounts section; install or update the Google tag with the dynamic remarketing event parameters (ecomm_prodid, ecomm_pagetype, ecomm_totalvalue); confirm the dynamic remarketing event flow in the diagnostics tab; build a dynamic Display campaign or Performance Max campaign that references the feed.

The ecomm_prodid value must match exactly the “id” attribute in your Merchant Center feed. This is where most implementations break. Sites with multilingual catalogs often pass a localized SKU on the page (FR-12345) while the feed uses a master ID (12345), and the dynamic ad falls back to a generic placeholder. The diagnostic in Google Ads under Tools > Audience Manager > Dynamic Ads shows the match rate. Anything below 90% is a tagging defect to fix before optimizing creative.

The product feed itself should follow Google’s Merchant Center specification: id, title, description, link, image_link, availability, price, sale_price, brand, gtin, mpn, condition, google_product_category, custom_label_0 through custom_label_4. The custom labels are the hidden lever. Use them to tag products by margin tier, by stock urgency, by seasonality, by hero status. Then in the dynamic remarketing campaign you can structure ad groups that prioritize, for example, products with custom_label_0 = “high_margin” for the cart-abandoner audience.

Image quality matters. Google’s specification calls for at least 1200 pixels on the longest side for product images used in Display creative. Feeds with low-resolution or watermarked images render poorly in responsive display ads and lose CTR by 20% to 40% in the tests Tinuiti published in 2023.

Custom business data feeds for non-ecommerce verticals

Dynamic remarketing is not retail-only. Google Ads supports custom business data feeds for travel, real estate, jobs, education, hotels and rentals, and a generic “custom” type. The feeds live in Tools > Business data, are uploaded as CSV or scheduled from a URL, and follow vertical-specific schemas.

For travel: destination ID, destination name, image URL, price, sale price, formatted price, contextual keywords, final URL. The image is typically the destination photo, the contextual keywords drive the matching when the user looked at “Lisbon weekend” on your site.

For real estate: listing ID, listing name, address, price, image URL, contextual keywords, final URL. The contextual keywords might be “3-bedroom Madrid Salamanca” and the dynamic ad surfaces the exact listing the user opened.

For jobs: job ID, job title, location, image URL, salary, contextual keywords, final URL. Indeed and LinkedIn both run this pattern at scale, retargeting candidates with the exact role they viewed.

The custom feed unlocks dynamic remarketing for any business that has a catalog of pages with structured attributes: course providers retargeting course pages, SaaS companies retargeting feature pages, B2B firms retargeting industry-specific landing pages. The mental model is the same as retail: tag the page with an ID, ensure the feed has that ID, and the dynamic ad will pull the right image, title and price.

Layered targeting: cart abandoners, product viewers, recent buyers

The 2018 best practice was to build one cart-abandoner audience and one product-viewer audience and run them in separate campaigns. The 2026 best practice is to layer them, because Performance Max and optimized targeting on Display weight the seed audience differently when it is precisely defined.

The minimum useful segmentation is five lists: all visitors (broad), product viewers (mid intent), cart abandoners (high intent), checkout abandoners (very high intent, often 5x to 10x conversion rate of cart abandoners), recent buyers (cross-sell and upsell). Each list has its own duration: 30 days for all visitors, 14 days for product viewers, 7 days for cart abandoners, 3 days for checkout abandoners, 90 days for recent buyers.

The granularity that matters most in 2026 is the checkout abandoner. A user who reached the payment page but did not complete is signalling either friction (a payment method failed, shipping cost was too high, the form rejected their address) or doubt (cold feet at the last second). Both are recoverable with the right message, often a transactional one (“your basket is still here, complete in one click”) rather than a discount. WordStream’s 2024 e-commerce benchmarks showed checkout-abandoner remarketing converting at 12% to 18%, against 3% to 6% for cart abandoners.

For B2B and high-consideration B2C, add a “demo viewers” or “pricing page viewers without conversion” segment. These users are deep in evaluation. They respond to social proof (case studies, ratings, testimonials) more than to discounts. Keep this segment in a 30-day to 60-day window.

The cross-sell layer is recent buyers. Exclude them from prospect campaigns (you are paying to re-acquire your own customers) and run a parallel campaign with cross-sell creative. The exclusion is an explicit “Customer Match: customers” or “Recent buyers (90 days)” set as audience exclusion at the campaign level. We routinely find e-commerce accounts where this exclusion is missing and 8% to 15% of remarketing budget is being spent re-targeting people who bought yesterday.

Frequency capping: respecting human attention

Frequency capping limits how often a single user sees your ad. In Google Ads, you set caps at the campaign or ad-group level, by impressions per day / week / month, per ad / ad group / campaign.

The classic guidance, supported by WordStream’s frequency studies and by Search Engine Journal’s 2023 display fatigue analysis, is 3 to 5 impressions per user per week per campaign for awareness, and up to 7 to 10 per week per campaign for high-intent retargeting (cart abandoners). Beyond that, click-through rate decays and unaided brand recall actually declines (banner blindness), per the Nielsen Norman Group’s recurring research on banner advertising.

The 2026 reality is that you also need a global frequency cap across campaigns, because a user remarketed by your “all visitors” Display campaign and your “cart abandoner” Display campaign and your “video viewer” YouTube campaign will see your ads multiple times even if each campaign respects its individual cap. Google Ads supports account-level frequency caps for Display through the “shared budget” and “campaign group” features, but the implementation is partial. The clean solution is to assign each user to one cohort at a time, through audience exclusions: a cart abandoner is excluded from the all-visitors campaign, a recent buyer is excluded from cart abandoners, etc.

For Performance Max, frequency capping is not exposed as a direct lever. Google manages it inside the campaign’s optimization. The only way to influence it is through the audience signal and the asset rotation: a tighter audience signal naturally limits how often the same user is hit.

Sequencing: awareness, decision and win-back

Sequencing is the choreography of which message a user sees, in which order, over which time window. The objective is to match the message to the user’s psychological state at each touch.

Days 1 to 2: direct product reminder (decision)

The user just visited. Memory is fresh. Show the exact product, current price, current availability. Dynamic remarketing creative pulls from the feed. No discount, no education. The job is to bring them back to the same decision they were already making.

Days 3 to 7: social proof and reassurance (decision)

Memory is fading, doubt is rising. Shift to creative featuring star ratings, customer testimonials, “10,000 customers in 2025”, “free returns within 30 days”, “secure payment”. The product is still in the ad, but the surrounding copy answers the unspoken objection.

Days 8 to 14: urgency and incentive (decision)

Conviction has wavered. Now you can introduce a measured incentive: “free shipping on your basket”, “the last 5 in stock”, “your wishlist expires soon”. This is the only stage where a discount is on the table for most categories, and even there the discount should be modest (5% to 10%) to avoid training the audience to wait for promotions.

Days 15 to 30: education and content (consideration)

For high-consideration purchases (B2B SaaS, premium goods, services), the user often needs more information rather than more pressure. Serve content: a buying guide, a comparison page, a case study, a webinar. The conversion event for this stage is content engagement, not purchase. The user moves into a “high-intent content reader” segment which then re-enters the decision flow.

Days 31 to 90: win-back (re-engagement)

If the user has not converted in 30 days, treat them as cold. A different creative, often bolder, with a stronger offer, sometimes a category change (“you looked at running shoes, our new trail collection just launched”). Frequency drops to 1 to 2 impressions per week. The objective is to surface back into consideration without burning the impression budget.

Past 90 days: lapsed (Customer Match)

Past buyers who have not repurchased move into a Customer Match win-back list. The creative is fundamentally different: thank you for being a customer, here is what is new, exclusive access, loyalty offer. Frequency is even lower. This is brand maintenance more than direct response.

The video sequencing equivalent on YouTube is the Ads Sequencing format inside Video campaigns: a 30-second skippable ad on first exposure, a 15-second non-skippable on second exposure to users who watched the first, and a 6-second bumper on third exposure to users who watched both. Google’s 2019 case studies on Ads Sequencing showed efficiency gains of 20% to 35% versus single-format video remarketing, and the format remains supported in 2026.

The 540-day retention maximum

Google Ads caps website-list membership duration at 540 days. The same cap applies to Customer Match lists: a user uploaded today stays in the list for up to 540 days unless re-uploaded.

This cap is a regulatory and operational compromise. Regulators (CNIL in France, ICO in the UK) have repeatedly flagged that consent for advertising cookies cannot be assumed to last forever. 540 days is Google’s outer bound, not a recommendation. Most lists should run at 30 to 180 days. Lists at 540 days serve a specific purpose: long sales cycles (B2B enterprise, real estate, automotive), or seasonal businesses where the consideration cycle is annual (winter sports gear, garden furniture).

The operational discipline is to refresh Customer Match lists. A list uploaded once in January 2025 will start expiring in mid-2026, with users dropping silently. Schedule monthly or weekly Customer Match uploads from your CRM or your customer data platform. Tools like Hightouch, Census and Segment all support direct Customer Match destinations now, replacing the old manual CSV upload.

GDPR, CCPA and Consent Mode v2

Consent Mode v2 became mandatory for advertisers serving the European Economic Area on 6 March 2024. Without it, conversions and audience signals from EEA users are not eligible for Google Ads personalization features, including remarketing and Smart Bidding optimization.

Consent Mode v2 has two implementations: basic and advanced. In basic mode, no tags fire when consent is denied. The user is invisible to Google. In advanced mode, anonymous “cookieless pings” fire when consent is denied, allowing Google to model conversions in aggregate without identifying the user. Google’s own documentation and the IAB TCF v2.2 guidance recommend advanced mode for any account where EEA traffic is more than 10% of the audience.

The impact on list sizes is real and structural. CMP (consent management platform) consent rates in the EEA average 60% to 75% in 2026, varying by country (lower in Germany and Netherlands, higher in southern Europe and the UK). That means 25% to 40% of EEA traffic is excluded from your standard remarketing list. Advanced Consent Mode partially recovers that loss for conversion modelling, but it does not put the user back into a remarketing list.

For CCPA in California (and now CPRA), the same logic applies through the IAB GPP (Global Privacy Platform) signals. Google Ads honors these signals: a user with the GPC (Global Privacy Control) header set is not added to remarketing lists. As of 2024 California’s enforcement of GPC was active, and Search Engine Land reported audience-list shrinkage of 5% to 12% for California-heavy retailers in the months following.

The compliance posture in 2026 is non-negotiable: a properly implemented CMP, Consent Mode v2 in advanced mode, IAB TCF v2.2 string passed to Google tags, GPP signals honored, and a written record of consent for any Customer Match upload. Accounts that skip this lose ad personalization eligibility and, increasingly, face enforcement actions from the CNIL, the Italian Garante, and the Spanish AEPD.

Remarketing inside Performance Max via audience signals

Performance Max replaced Smart Shopping in late 2022 and steadily absorbed Display, Discovery and YouTube budgets through 2024 and 2025. By the end of 2025, Tinuiti’s Q4 retail benchmarks reported that Performance Max accounted for the majority of Google Ads spend in retail, with classic Display campaigns shrinking to specific use cases.

In Performance Max, you do not target audiences directly. You provide audience signals, which Google uses as starting points for its optimization. The signals are: your Customer Match lists, your remarketing lists (website visitors, app users, video viewers), custom segments (built from search terms, URLs, apps), and demographic / interest data.

The mental shift is from “target this audience” to “show this audience as an example of what conversion looks like”. The richer and cleaner the signal, the faster Performance Max learns. A signal stitched together from a dirty all-visitors list with no exclusions teaches the algorithm to find more dirty all-visitors. A signal built from your top 10% LTV Customer Match list teaches the algorithm to find more high-LTV users.

The 2026 best practice for Performance Max remarketing is to run at least three asset groups in the campaign: one anchored on the Customer Match high-LTV signal, one on the cart-abandoner remarketing signal, and one on the high-intent custom segment built from competitor and brand search terms. Each asset group has its own creative, which lets Google match the right creative to the right user implicitly.

Search Engine Journal’s June 2024 deep dive into Performance Max audience signals found that accounts with three or more well-curated signals outperformed accounts with a single signal by 18% to 27% on ROAS at scale. The signal is not a targeting filter, it is a learning accelerant.

The 90-day remarketing audit playbook

A remarketing operation drifts. Tags break, feeds desynchronize, audiences age out, exclusions go missing, creative fatigues. A 90-day audit cycle is the minimum cadence to keep the system honest. Here is the playbook we use on LaFactory.com client accounts.

Day 1 to 7: tagging and feed integrity

Run Google Tag Assistant on the homepage, a category page, a product page, the cart, the checkout and the thank-you page. Confirm the Google tag fires, the dynamic remarketing event parameters (ecomm_prodid, ecomm_pagetype, ecomm_totalvalue) are populated, and the IDs match the Merchant Center feed. Pull the Audience Manager > Audience Sources > Google Ads Tag report and confirm the tag has fired in the last 24 hours from each page type.

For the feed: in Merchant Center, check the Diagnostics tab for disapprovals, item-level errors, and image issues. For dynamic remarketing specifically, check Tools > Audience Manager > Dynamic Ads in Google Ads for the match rate. Anything below 95% needs investigation.

Day 8 to 14: audience inventory

List every audience in Audience Manager. For each, capture: list size (cumulative), active users on Search, active users on Display, active users on YouTube, membership duration, last update, source. Flag any list with fewer than 1,000 active users on its primary network. Flag any list older than 18 months that is still set to 540-day duration without a documented reason. Flag any Customer Match list whose last upload is more than 30 days ago.

Day 15 to 21: campaign and audience mapping

For each Display, Discovery, Video and Performance Max campaign, document which audiences are used as targeting (Display, RLSA) or signals (Performance Max), and which are used as exclusions. Build a matrix: rows are audiences, columns are campaigns, cells are “targeting”, “signal”, “exclusion” or empty. The matrix immediately reveals: missing exclusions (recent buyers in prospect campaigns), missing signals in Performance Max, and overlapping targeting that produces the cross-campaign frequency problem.

Day 22 to 28: frequency and creative fatigue

Pull the Reach > Frequency report for each Display campaign. Anything above 10 weekly impressions per user is fatigue territory. Pull creative-level CTR over the last 30, 60 and 90 days. Creative whose CTR has dropped more than 25% from its launch baseline is fatigued. Refresh.

Day 29 to 35: sequencing audit

For each user cohort (cart abandoner, product viewer, etc.), document which creative variant is shown at which day-since-visit. Most accounts have no sequencing at all: every cohort sees the same creative for 30 days. Build at least the three-stage sequence (decision / urgency / win-back) per cohort.

Day 36 to 49: Customer Match refresh

Refresh Customer Match lists from CRM. Validate match rates are stable (40% to 70%). For lists where match rates dropped, investigate input data quality: missing country prefix on phones, stale emails, incorrect hashing.

Day 50 to 63: Consent Mode and compliance

Confirm Consent Mode v2 is in advanced mode and the IAB TCF v2.2 string is being sent. Check the Consent Mode diagnostics in the Google Ads Conversion settings. For Customer Match lists, audit the consent record: every uploaded user must have documented consent for advertising contact.

Day 64 to 77: incrementality testing

For one major audience (typically cart abandoners), run a geo-experiment or a conversion-lift test in Google Ads. Hold out 30% of the audience as control, expose 70% as test. After 14 days, measure incremental conversions. If incrementality is below 15% (test conversions only 15% higher than control), the audience is over-saturated and budget should be redirected.

Day 78 to 90: documentation and roadmap

Write up findings, decisions made, decisions deferred. Update the audience inventory, the campaign-audience matrix, the sequencing map. Schedule the next audit at day 180. The audit is not a project, it is an operational cadence.

Common mistakes

The recurring failures we see in 2026 audits, in rough order of frequency:

• Recent buyers not excluded from prospect or cart-abandoner campaigns. Direct budget waste, often 8% to 15% of spend.
• Dynamic remarketing tag firing on category pages without a product ID, which means the campaign falls back to generic creative for those users.
• Customer Match lists never refreshed after the initial upload, decaying silently.
• Consent Mode v2 in basic mode (or not implemented at all) on EEA traffic, which silently disqualifies a large share of conversions from optimization.
• Remarketing campaigns running on Maximize Conversions when the audience is too small for Smart Bidding to learn, producing erratic CPA.
• 540-day duration left as default on every list, which inflates list size on paper without adding qualified users.
• Performance Max audience signals left empty or filled with a generic “all visitors” list, which gives the algorithm no useful learning seed.
• Cross-campaign frequency unmanaged, with users hit by 4 to 5 different campaigns simultaneously.
• Single creative variant per cohort, no sequencing, no refresh schedule.
• Similar Audiences still referenced in account documentation and team training, two years after deprecation.

Conclusion

Remarketing in 2026 is not the same lever it was in 2018. Audience attrition, consent regulation and Performance Max have changed the inputs, the controls and the reporting surface. What has not changed is the underlying logic: a clean first-party signal, segmented by intent, refreshed on a known cadence, sequenced to match the user’s psychological state, and frequency-capped to respect attention, will outperform a bigger, dirtier, undisciplined operation by a factor of three to six on ROAS.

The teams winning are the ones treating remarketing as a system: tagging, feeds, audiences, exclusions, sequencing, audits. Not as a campaign you launch and forget. The infrastructure is more demanding than it was, but the payoff is also bigger, because the cost of cold prospecting climbed in parallel.

Start with the audit. Most accounts we look at lose 20% to 35% of their remarketing budget to fixable defects: missing exclusions, broken tags, stale lists, consent gaps. Recovering that budget is usually a 30-day project, and it pays back the entire 90-day audit cycle several times over.

Sources

• Google Ads Help, audience requirements and minimum list sizes (support.google.com/google-ads).
• Google Ads Help, Consent Mode v2 implementation guide (support.google.com/google-ads, March 2024).
• Google Ads Help, Similar Audiences sunset announcement (April 2023).
• Google Merchant Center documentation, product feed specification and dynamic remarketing setup.
• Search Engine Land, “Q3 2025 PPC benchmarks: CPC inflation in commercial intent” (October 2025).
• Search Engine Land, “Remarketing health checks for the post-cookie era” (October 2024).
• Search Engine Journal, “Performance Max audience signals: what works in 2024” (June 2024).
• Search Engine Journal, “Display fatigue and frequency capping: 2023 evidence” (2023).
• Search Engine Journal, “YouTube remarketing trends 2024” (2024).
• Tinuiti, Q2 2021 Digital Ads Benchmark Report, ATT impact on iOS match rates.
• Tinuiti, Q4 2024 Retail Benchmarks, Customer Match versus cookie-based ROAS comparison.
• WordStream, “Google Ads Benchmarks 2024” (2024 industry averages by vertical).
• WordStream, “Frequency capping research and creative fatigue” (2024).
• Nielsen Norman Group, banner blindness research, recurring updates 2007 to 2023.
• IAB Europe, TCF v2.2 specification and Google integration guidance.
• CNIL, deliberation on cookie consent durations and audience list retention (2021 to 2024).