In a post-cookie environment, we've shifted our retargeting strategy to be built almost entirely around first-party data activation, using Privacy Sandbox signals and data clean rooms as validation and optimization layers rather than direct replacement tools. Practically, this meant unifying CRM data, on-site behavioral signals, and lead quality outcomes into privacy-safe audience definitions that could be modeled and reactivated through platform-native solutions, instead of relying on pixel-level user tracking. One tactic from our 2025 campaigns that delivered a measurable improvement in both CPA and ROAS was restructuring retargeting away from "who visited" and toward "who progressed," using clean room analysis to identify behavioral patterns that correlated with high downstream conversion value, such as content depth, tool usage, or repeat visits across specific time windows. By retargeting only those high-intent cohorts and excluding low-signal traffic early, we reduced wasted spend, improved message relevance, and saw more efficient conversion paths, proving that precision and intent modeling now outperform scale-based retargeting in a privacy-first ecosystem.
Post-cookie retargeting relied on first-party data combined with Privacy Sandbox cohorts and a retail data clean room. Customer email lists were matched anonymously to cohort segments, and purchase behavior was shared in aggregated form, ensuring privacy while keeping targeting precise. One tactic that worked in 2025 was creating micro-cohorts of repeat buyers by product type and showing tailored creative only to those groups, rather than blasting all past visitors. This approach lowered noise and wasted spend. Across four campaigns, cost per acquisition dropped by 18% and ROAS improved by 22% compared to previous cookie-based retargeting. The key insight was that privacy-safe segmentation with clear product relevance outperformed broad retargeting. Showing the right message to the right group, even in aggregated form, restored efficiency in campaigns while respecting evolving data rules.
Privacy Sandbox is overkill for most businesses. We're just getting smarter about the data we already own. We export our CRM contact list weekly--event planners who've inquired, downloaded speaker one-sheets, or attended webinars--hash the emails, and upload them as matched audiences to ad platforms. Then we layer in lookalike modeling to find similar decision-makers. The move that improved ROAS by 2.1x in 2025: we stopped retargeting everyone equally. High-intent prospects (those who viewed pricing or speaker availability) got aggressive retargeting with calendar booking CTAs. Low-intent got nurture content--speaker spotlight videos, industry insights. Matching ad intensity to intent signal made every dollar work harder. You don't need complex privacy tech--you need to respect your own data and use it with precision.
The big shift for us has been treating first-party data like an asset instead of an afterthought. With cookies fading, we're leaning hard into CRM, email engagement, and site behavior to build high-intent audiences inside clean rooms and Privacy Sandbox-friendly setups. Instead of chasing people around the internet, we're retargeting based on real signals like content depth, repeat visits, and product-level interest. One tactic from our 2025 campaigns that actually worked was creating narrow, behavior-based retargeting pools and pairing them with message sequencing instead of one-size-fits-all ads. That cut wasted spend fast because we stopped paying to re-educate people who were already warmed up. CPA came down not because of some magic tech, but because the targeting finally matched where people actually were in the decision process.
In 2025, our focus shifted heavily toward making first-party data actually work harder instead of trying to replace cookies one-for-one. At Timeless London, we leaned into consented customer signals like purchase history, on-site behaviour, and email engagement, then activated them through platforms like Google's Privacy Sandbox and clean rooms for media planning. The biggest unlock was building high-intent audience cohorts based on lifecycle stages rather than past clicks. One tactic that genuinely improved ROAS was syncing repeat-buyer and high-AOV segments into clean rooms and using them as seed audiences for modeled prospecting, while excluding them from retargeting entirely. That reduced wasted spend, lowered CPA, and pushed budgets toward users who looked like our best customers instead of chasing the same people again. The mindset shift was key: retargeting became about smart suppression and intent signals, not constant reminders.
To rebuild retargeting in a post-cookie world, I started using first-party data with Sandbox and data clean rooms when the third-party data stopped working. Let's understand with an example, where I'm running an online cosmetic business and one customer added beauty products, including shampoo and serum, to the card but not ordered. On that point, I use first-party data rather than tracking them through cookies which even helps in time management and take quick profitable steps. Using the privacy Sandbox tool helps the audience to show ads that they're actually interested in which maximise the possibility to purchase the required products. Also by the use of data clean rooms I match the customer list in the interested products with the platform's data securely to know which advertisements actually lead to purchase by customers.
In the absence of cookies, we have fundamentally changed our retargeting strategy, mainly by doing more with first-party data rather than trying to replace the lost data. It involved us being more rigorous in how we collect and segment data from the channels we own, such as user behavior on our site, engagement with our emails, and CRM activity. We used data from Privacy Sandbox signals and data clean rooms, not as magic solutions but as safety measures that allowed us to protect user data while still cooperating with platforms. The change in thinking was crucial. Retargeting was no longer about chasing users everywhere; it was about re-engaging people who had already demonstrated real intent. One tactic from our 2025 campaigns that really made a difference at Mad Mind Studios was using clean room insights to develop intent, based on audience tiers rather than broad retargeting pools. We combined anonymized conversion signals with platform data to identify which on-site actions had the strongest correlation with downstream value, and then tailored our messaging to those moments. Rather than delivering more ads, we delivered fewer, better, timed ones. This single change was a major factor in reducing CPA, as it helped focus spending on the highest-converting areas. What made this effective in the long term was accepting that precision now comes from patience and quality, not scale. By tightening feedback loops between first-party data, clean-room insights, and creative testing, we improved ROAS without increasing frequency or user pressure. The biggest lesson was that privacy-first doesn't mean performance-last. When first-party data is respected and used thoughtfully, it can outperform old retargeting models that relied on volume and shortcuts.
The most important advantage in a cookieless world is to view retargeting as both an audience-building issue and a measurement issue, not a tracking issue via third-party data. In building long-lasting audiences in this environment, we have relied very heavily on first-party signal sources, including, where available, authenticated or hashed identifiers and on-site intent signals. In addition to using first-party signals to create long-lasting audiences, we also use clean rooms to provide privacy-compliant measurement and to manage overlaps between platforms. One strategy we implemented in 2025, which consistently produced greater efficiency than previous strategies, was to implement an intent-based retargeting program with strict suppression of low-value traffic along with frequency controls. Rather than "all visitors," we created high-intent segments based on first-party behaviors (i.e., deep product engagement, comparing products, returning to the site within a short time frame, and key funnel milestones). Next, we added suppression of traffic that was low value (i.e., bounces, accidental visits) and users who had recently made a purchase or demonstrated strong post-purchase behavior. Finally, we added frequency limits and short membership windows to avoid continually advertising to users who were likely to convert. From a platform perspective, we focused on cleaning up our server-side event hygiene (using standardized event taxonomies, removing duplicate events, and ensuring that conversion events are consistent across platforms) to enable the platforms to effectively optimize based on the reduced signal set. From a measurement perspective, we ran hold-out/incremental analysis and reported lift via clean room reports to verify that we were demonstrating incremental lift and avoiding double-counts across partners. Overall, by concentrating budgets on true intent, reducing the wasted ad spend due to over-exposure, and providing clean conversion signals to guide the platforms' bids, we typically see a significant CPA reduction and increase in ROAS without relying on third-party cookies.
We're leaning on our own customer data to rebuild retargeting in a way that's privacy-compliant. Instead of relying on third-party cookies, we anchor campaigns in the behaviours and preferences people share directly with us. For example, we feed first-party event data from our website and app into Google's Topics API and use clean rooms to join that data with aggregated cohorts from partners. This lets us build high-intent segments while still keeping individual records anonymous. In our 2025 campaigns we saw the biggest impact when we combined this with predictive modelling. We scored leads based on their on-site actions and then built lookalike audiences around our best customers. By only retargeting the top quartile of high-intent visitors and pairing them with creative tailored to their previous interaction, we improved ROAS significantly and saw our CPA drop by more than 20% compared with broad cookie-based remarketing.
We've been running parallel tests with Google's Privacy Sandbox, specifically the Protected Audience API for retargeting. Instead of cookie-based remarketing, we're now adding users to interest groups based on on-site behavior, like product category views or cart adds. For one eCommerce client, we segmented audiences into interest groups by price point and product type. Their retargeting ROAS improved 38% because the ads were hyper-relevant without needing to track individual users across the web. On the data clean room side, we partnered with a client's media vendor to match our CRM data in a secure environment. This let us identify high-value lookalikes we couldn't reach before. First-party data quality has become critical. We've been obsessive about capturing granular behavioral signals on-site because that's what feeds these new systems effectively.
Our approach has been to rebuild retargeting around intent signals we actually own, not trying to recreate cookies in a new wrapper. First-party data became the spine, and Privacy Sandbox and clean rooms were used to extend it, not replace it. In practice, we leaned heavily on consented behavioral signals like product views, category depth, and repeat engagement, then activated those audiences through clean rooms with platforms like Google and retail partners. Instead of audience chasing individuals, we focused on high-intent cohorts and let the platforms handle reach within privacy constraints. One tactic from our 2025 campaigns that worked well was sequential retargeting based on action depth, not recency alone. For example, users who viewed pricing or added to cart twice but didn't convert were moved into a higher-value cohort and shown fewer, more specific messages. That reduced wasted impressions and decision fatigue. The result was a lower CPA and more stable ROAS, even though total retargeting volume was smaller. The big shift was accepting less precision per user in exchange for clearer signals and better psychological timing. When the message matched where people actually were in their decision process, performance followed.
Sy'a used first-party purchase and engagement data combined with Privacy Sandbox cohorts and a retail data clean room to rebuild retargeting without relying on third-party cookies. The tactic involved creating audience segments based on customers' preferred flavors, purchase frequency, and interaction with digital content. Ads were then served to similar cohorts across platforms, highlighting specific blends or seasonal offerings. This approach lowered the cost per acquisition by 31% compared to prior broad retargeting campaigns, while ROAS increased by 27%. The key insight was that connecting signals in a privacy-safe way allowed the campaigns to feel personal without tracking individual users. It also revealed that luxury audiences respond strongly to tailored experiences rather than generic ads. The experience showed that first-party insights, when used thoughtfully and safely, can maintain performance in a post-cookie world. Businesses can adopt this by mapping meaningful customer behaviors and serving content that resonates with intent, not just clicks.
Retargeting didn't die with cookies—unproven retargeting did. We run first-party data like a measurement system: consented events collected server-side, analyzed in a clean room, then activated through Privacy Sandbox-safe paths. The clean room's job is blunt: tell us where ads create lift versus where they just harvest credit. Our best 2025 move was cutting "free conversions." We found a segment of repeat pricing visitors who routinely converted via organic within a short window. We excluded them from retargeting and reallocated budget to the "stalled" users—checkout started, activation hit friction, then stopped—with hard frequency caps and a two-step sequence (remove the top objection, then make the offer). CPA dropped ~20% because we stopped paying to be present and started paying only to change outcomes.
We rebuilt retargeting by anchoring on first-party event data and activating it through clean rooms rather than identity stitching. With Privacy Sandbox limits, we shifted from user-level retargeting to cohort and intent-based sequencing using GA4 first-party events and platform clean rooms for overlap and lift analysis. One tactic that worked in 2025 was value-based audience reactivation. We fed high-intent, server-side conversion signals into clean rooms to build modeled audiences, then capped frequency and sequenced creatives by recency. CPA fell 18 percent and ROAS improved 22 percent versus cookie-based retargeting. Google has stated third-party cookie deprecation can reduce observable signals by over 20 percent, which matched our pre-migration benchmarks and validated the shift. Albert Richer, Founder, WhatAreTheBest.com
First party scheduling and intake data became the anchor when third party retargeting weakened. Instead of following up on previous visitors, campaigns were re-built around confirmed signals of intent such as appointment starts, abandoned intake forms and follow-up eligibility windows. Those events were triggered within Google Privacy Sandbox using audience segments aggregation instead of user-based tracking. Clean room analysis was not based on identity but sequence, which minimized noise. One tactic that reduced CPA was retargeting based on patients who paused in insurance verification and not site visitors in general. That shrunken pool converted at a higher rate and reduced wasted spend. Cost per acquisition fell some 22 percent while message frequency remained lower. Performance improved because ads reflected real points of friction and not things inferred to be of interest. Post-cookie retargeting worked insofar as it was based on operational behavior and not browsing history.
First-party data began to gain weight when the audience definitions changed from individual to moments. Consent based signals from forms on the site, chat transcripts and reorder behaviour were grouped by intent windows and not user profiles, then triggered via Google Privacy Sandbox cohorts and clean room analysis within Amazon Marketing Cloud. The goal stayed simple. Get people when the need was on, not after it had passed A tactic that worked came from suppressing audiences rather than chasing them. Patients who had completed an insurance verification or reorder in the past fourteen days were not considered for retargeting at all. Spend shifted toward lookalike intent cohorts constructed from those converters employing aggregate signals only. CPA fell nineteen percent quarter over quarter because ads ceased to follow people who already acted. ROAS improved due to closer impressions to decision timing. Measurement was on incrementality not clicks Clean room queries were used to compare exposed and unexposed groups through the use of matched conversion windows. The lesson was clear. Retargeting recovered once it respected consent, timing and restraint. First-party data did best when it helped to reduce waste and not increase reach.
First party data can only be useful when it reflects true intent over surface behavior. Health Rising moved away from the general retargeting pool and re-built audiences within the clean rooms using appointment timing, care gaps and service sequence data, not page visits. That structure fit very well with Privacy Sandbox signals because it was focused on patterns and not individuals. One of the tactics accounted for the greatest efficiency gain in 2025. Patients who completed an initial visit without a follow-up visit scheduled for fourteen days were placed into a delayed re-engagement cohort. Messaging was delayed for 7 days and then resumed in various contextual placements related to health education content instead of clinic advertisements. The pause mattered. Immediate follow-ups came across as intrusive and underperformed. That change helped CPA to be 22 percent lower over two quarters. ROAs improved as well due to the better value visits booked by returning patients and longer stay in care. Clean room analysis confirmed incremental lift by comparing rebooking rates against a matched group which was not shown any media at all. Retargeting is effective again when timing is based on clinical reality rather than ad platform urgency.
First party data was made useful once it was considered as an analysis asset as opposed to targeting list. Beacon concentrated on the consented CRM events, site behavior, and offline conversion data combined into clean room environments complying with the rules of Privacy Sandbox. That architecture enabled it to provide performance questions to be answered without revealing user level identities. Cohort patterns led to decisions, rather than individuals. One of the strategies that had quantifiable outcomes consisted in intent tiering. The short window interest clusters were defined by high signal actions in the clean room, including the revisits of pricing pages within seven days or downloading of implementation guides again. The clusters then relayed creative bid weighting and sequencing in platform native environments. There was no direct retargeting, but advertisements were in-line with readiness. CPA fell by 18 percent in ten weeks, in part due to the change in spending on low intent impressions which cookies previously covered. ROAS was also enhanced since there was matching of stage and not persona with messaging. The lesson was practical. When measurement is targeted at improving performance, the post cookie performance is enhanced. Efficiency is restored with exposure being directed by economics rather than identity even when privacy is tight.
Attempting to reconstruct pixel-based retargeting using data clean rooms and the Privacy Sandbox is merely accumulating architectural debt disguised as innovation. The premise of "rebuilding retargeting" ignores the fundamental shift in data sovereignty. In 2025, the tactic that actually lowers CPA is not a privacy workaround, but a strategic reallocation of resources toward AI-driven intent modeling and Product-Led SEO. We must stop treating user acquisition as an identity resolution problem and start treating it as a signal processing challenge. By leveraging LLMs to analyze search queries and on-site behavior, we can map content directly to specific user problems. This approach prioritizes Contextual Relevance over user identification. It filters out the noise of casual browsing and isolates the signal of active problem-solving. This drastically improves the Signal-to-Noise Ratio in our funnels. When we shifted our architecture to capture demand via high-fidelity technical content rather than chasing users across the open web, we saw a significant efficiency gain. We stopped paying a premium for stale third-party audience segments and started investing in the architecture of user intent, resulting in a sustainable, privacy-safe ROAS lift.
At the foundation of making post-cookie retargeting actually work is the robust infrastructure we're building. We start by taking a client's first-party data--CRM records like transaction history--and "move" it into the data clean room. Within this privacy-measure-driven space, they can form high-value audience segments that don't share raw user data with us. These privacy-safe segments can access the Google Privacy Sandbox APIs, allowing the right ads to be delivered without third-party cookies. A tactic that has produced lower CPA for our clients is high-intent look alike modeling. Rather than simply retarget all past visitors, we help them develop a model that scores users on deep on-site engagement, like watching a whole product video, or repeat visits to a category page. We'll take that high-propensity "seed" audience over to a clean room, where we can find look alike audiences based on a publisher's user base, effectively moving budget from broad retargeting to highly-qualified prospecting to win over a new customer at a lower CPA.