B2B retargeting after cookies: the Chrome playbook driving pipeline now

Protected Audience API remarketing

Short answers to the two big questions I hear most often:

• Can I still do retargeting after third-party cookies? Yes - via Protected Audience API remarketing, which keeps audience membership and ad auctions inside the browser.
• What's the ROI path for B2B services? Anchor to highest-intent site actions, cap reach and frequency, measure with Attribution Reporting, and compare cost per qualified form fill, SQL, and opportunity against your current retargeting baseline.

What it is and why it matters

The Protected Audience API (formerly FLEDGE) is a browser-based targeting API that lets advertisers form interest groups on a user's device and run ad auctions locally. In plain terms, my site can say "this visitor looked at pricing," and the browser can later compete my ad against other bids - without exposing the person's cross-site browsing. Instead of shipping user data to many parties, the browser keeps it sealed, renders creatives in fenced frames, and returns aggregate results.

A quick, practical pilot plan I use

• Stand up interest groups tied to clear intent, like pricing views, demo bookings, or whitepaper downloads.
• Run on-device auctions through a supported buying path that implements buyer worklets.
• Measure wins and conversions with the Attribution Reporting API, then compare cost per SQL and cost per opportunity against your legacy retargeting.

If that sounds new, it is. The goal is familiar: re-engage people who showed intent, in a way that respects privacy and still grows pipeline.

Remarketing without third-party cookies

Let me put it simply. With interest-group advertising, my site asks the browser to add a visitor to an interest group using joinAdInterestGroup. No cross-site cookie is needed. That group lives on the device for a time to live (TTL) I set. Later, when that person is on another site that sells ad space, the browser can bid and show my ad to that same person - primarily in Chrome, where the API is supported.

How I map this to common B2B segments:

• Visited pricing or services page: join "High-intent: pricing" for 7-14 days.
• Booked a demo but didn't attend: join "Demo no-show" for 14-30 days.
• Downloaded a whitepaper: join "Content engaged" for ~30 days.
• Started a form but bounced: join "Form starter" for ~7 days.

I keep consent front and center. I only add people to groups when I have the right consent. I don't pass PII into interest-group names or signals. I use generic labels and short TTLs. This mechanism sits alongside other cookieless remarketing strategies - like publisher first-party IDs and clean-room reach - but this one keeps the audience logic on the device.

A small twist I like: because groups sit in the browser, I can update them over time. If someone moves from "Content engaged" to "Booked demo," I remove the first and add the second to avoid noisy overlap.

On-device ad auctions in the Privacy Sandbox

Here's how an on-device auction flows. When a page loads an ad slot, the browser gathers eligible interest-group ads. Buyer logic runs in a buyer worklet to place a bid. The page's selling code runs in a seller worklet to score bids against other demand. The winning creative renders in a fenced frame so it can't see the page and the page can't see the ad's internals.

Privacy guards are built in. K-anonymity rules require a minimum audience size before an ad can run, so tiny groups don't leak identity. That's good for people, and it nudges me (and most teams) to plan broader segments rather than a tangle of tiny ones.

What about pacing and frequency? Trusted Bidding Signals and Trusted Scoring Signals let buyers and sellers pass small, privacy-safe hints at auction time. Buyers can reference trusted signals to apply frequency caps, budget limits, or simple priority. Sellers can use their signals for floors, blocks, and quality rules. It's not a 1:1 match with legacy cookie stacks, but it's close enough to manage campaign health.

Targeting precision shifts slightly. You still reach the right people, but the system avoids narrow micro-targeting that can create privacy risk. For B2B, that's fine. The best segments come from clear intent on your site, which maps well to interest groups.

Publisher controls and interest-group advertising

Publishers still hold the keys to their inventory. They choose where and when interest-group ads can run. Expect familiar controls like floors, brand-safety categories, and competitive exclusions. K-anonymity thresholds can also limit delivery for very small cohorts, which matters for niche B2B segments.

A few functional notes I confirm up front:

• Creatives render in fenced frames, which limits dynamic URL macros and some forms of third-party measurement. Clicks may open in a new tab by design.
• Creative review continues, and some formats have extra checks. In-stream video support exists in parts of the market, but feature parity varies by SDK and seller.
• Viewability and interaction metrics can lag for some formats while APIs catch up.

What to confirm with SSPs and publishers before you scale:

• Whether they allow interest-group ads across the placements you plan to buy, and which floors apply.
• How brand safety, category blocks, and competitive rules map in this mode.
• Which creative types they accept for fenced frames and any size or script limits.
• How they configure timeouts for on-device auctions and how that affects win rate.
• Which reporting fields they share on won auctions and impressions.

If you sell a B2B service with a narrow ICP, these checks reduce surprises. You may find that broad upper-funnel inventory is ready first, while niche pages come online later. That's normal in a shift to cookieless remarketing.

Protected Audience services deployment

I can roll out Protected Audience without turning my stack upside down. A clear, two to four week plan works well for most B2B teams.

Week 1: plan and pick partners

1. Select a DSP or buying tech that supports Protected Audience buyer worklets. Confirm budget pacing and frequency options.
2. Define interest groups and TTLs. Start with three to five high-intent segments, not fifteen.
3. Align roles. Marketing writes the intent rules and creative messages. Dev implements join logic and consent checks. Ad ops wires auctions and monitors delivery.

Week 2: implement and prepare

4. Implement interest-group join rules through your tag manager or site code with consent gating. Use short, clear group names and avoid PII.
5. Prepare creatives that are friendly to fenced frames. Keep tags light. Avoid heavy external scripts. Have fallbacks for click tracking.
6. Stand up a lightweight trusted-signals endpoint or use a supported service for budgeting and caps. Keep payloads minimal.

Week 3: measure and QA

7. Configure measurement with the Attribution Reporting API for event-level data (with reporting delays and privacy filters), and Private Aggregation for spend, reach, and conversion funnels. Map conversions to your CRM.
8. Run a QA flight on a small budget. Inspect auction debug reports, win rates, and creative render times. Compare to your benchmark CTR and CVR.

Week 4: expand and learn

9. Scale across exchanges that support this auction path. Keep pacing conservative at first. Raise caps as you confirm conversion quality.
10. Build a readout that connects spend to MQLs, opportunities, and pipeline sourced.

This is browser-based ad targeting. I keep it as platform-agnostic as possible while still using Google's Privacy Sandbox across multiple sellers that support it. If your team knew this as the legacy FLEDGE flow, the core idea is the same - the name changed and the privacy features are richer. Support and feature parity vary by seller, so I plan for incremental rollout.

Reporting and testing with non-Google sellers

Measurement shifts from cookies to APIs built for privacy. Two tools matter most for remarketing:

• Attribution Reporting API event-level reports. These link ad events to conversions with a short delay and privacy filters. Use them to track form fills, demo bookings, and early pipeline steps.
• Aggregated reports via Private Aggregation. These show spend, reach, and conversions in buckets. Use them to watch lift and cost trends by channel or interest group.

I bring a B2B lens. For each interest group, I track cost per MQL, cost per opportunity, and pipeline dollars created. I add a simple CRM flag to mark "seen in Protected Audience" so I can compare journey speed to other retargeting paths. You might see lower raw reach than legacy cookies. You can still beat cost per SQL if segments are clean and frequency is sane.

Testing design tips

• Run a holdout. Keep a geo or a slice of traffic as control for two to four weeks.
• Use a geo-split when possible to avoid cross-pollination that can happen with device-level splits.
• Aim for enough sample size to detect a 15-25% lift in form fills. That usually means a few hundred conversions per cell, though your baseline rate sets the bar.

What "allow testing on all inventory with non-Google sellers" means in practice

Include multiple SSPs and exchanges that can run interest-group auctions. If you use header bidding with Prebid, include a path that requests on-device auctions. Run the same creatives and bids across sellers to avoid bias. Expect some sellers to apply extra floors or timeouts while they tune their stacks. Note and adjust, but avoid one-off rules that skew measurement.

A small reality check. Viewability and reach reporting can look different across sellers while APIs roll out. Benchmark within each seller rather than across all at first, then unify once metrics line up.

Opting out and additional considerations

You always have control. At the site level, a Permissions-Policy header can block join-ad-interest-group and run-ad-auction. Users can also block interest groups in their browser settings. If you need to pause activity, partners can apply contractual opt-outs by site or by campaign.

Important practical notes for B2B advertisers:

• K-anonymity thresholds. Very small cohorts may not serve, so avoid slicing by too many conditions at once. Test group sizes and set TTLs that help scale without over-exposure.
• Frequency capping. Caps often apply per browser. Cross-device capping is limited, so consider tighter caps and shorter TTLs.
• Creative QA. Fenced frames restrict direct page access, so audit click tracking, landing-page redirects, and any dynamic macros before launch.
• In-stream video. Support exists, but features vary by SDK and seller. Some viewability or skip metrics may be limited during early phases. Keep separate reporting lines for video so display numbers stay clean.
• Latency and timeouts. On-device auctions have strict time budgets. Heavy scripts and slow endpoints can reduce win rate. Keep signals light.
• Regulatory and consent. Gate interest-group joins on consent. Keep group names generic and avoid encoding sensitive traits. Document your logic for audits.

If risk management helps your team align, sketch a simple grid: risk, likelihood, impact, and mitigation. For example, "Small segments may not serve" paired with "Merge segments, lengthen TTL slightly, broaden inventory." Keep it living, not static, as the stack matures.

Privacy Sandbox initiative

Protected Audience sits inside a bigger privacy plan. Privacy Sandbox includes multiple APIs: Topics API supports interest-based prospecting that does not trace people across sites. Attribution Reporting handles measurement for clicks and views with privacy filters rather than cross-site cookies. Protected Audience on Android extends the same approach into mobile apps.

Timelines are evolving, including the pace of third-party cookie deprecation and regulatory reviews, which is why steady testing helps. I phase my work so I learn while the market matures:

• Keep a small, always-on pilot that refreshes weekly.
• Compare one or two interest groups against your legacy retargeting each month.
• Review reach and cost by seller and format, then shift spend toward the most efficient paths.

This shift may feel like a contradiction at first. You lose a familiar tool, yet your pipeline can get steadier. Why? Because remarketing gets anchored to first-party signals, not a web of third-party cookies. That tends to pull in cleaner intent and more qualified forms. If I plan the rollout with clear segments, consent, and measurement, I keep the remarketing muscle while respecting people's privacy - and that's a win your board, your buyers, and your future self can live with.