The Adequacy Gap: How Signal Loss is Creating a New Adequacy Risk in Digital Notice Programs
Reach and frequency remain the central evaluative metrics for digital notice programs, but the data, identifiers, and measurement methods that produce them have changed substantially over the last five years. On the page, those reach and frequency figures may look the same as they did five years ago. The infrastructure beneath them is materially different, and the implications for notice adequacy have not been adequately examined.
This matters most where digital notice is the mechanism rather than a supplement. In product liability, consumer fraud, false advertising, and similar matters where class members cannot be fully identified from defendant records, behavioral targeting carries the notice plan. The adequacy of those programs rests on an assumption hardened over years of practice: that digital targeting still delivers the precision it did when these methods became standard. Regulators, platforms, and the platforms' own AI products are actively dismantling that assumption.
What Actually Changed (and When)
The shift comes from three overlapping directions.
The first is platform-level. Apple’s App Tracking Transparency (ATT), enforced with iOS 14.5 in April 2021, required apps to obtain user permission before tracking users across other companies’ apps and websites or accessing the device advertising identifier.[1] Many industry benchmarks have shown ATT opt-in rates below half, often substantially below half, although reported rates vary by app category, geography, and methodology.[2] Google also reduced identifier availability on Android: beginning with a late-2021 Google Play services update, the Android advertising ID was removed when a user deleted it in Android settings, with apps receiving a string of zeros instead; the rollout began on Android 12 and expanded to all Google Play-supported devices on April 1, 2022.[1] Together, these changes reduced the availability of mobile advertising identifiers used for cross-app targeting and measurement.
Meta then consolidated detailed-targeting interest categories beginning in 2025 and removed detailed-targeting exclusions from key campaign workflows, while leaving some other exclusion tools available.[3] Ad sets still relying on deprecated interests stopped delivering on January 15, 2026.[3] The practical direction is clear: less granular manual audience construction and more reliance on Meta’s AI-optimized delivery tools, including Advantage+.[3] That may be acceptable for many commercial campaigns. For court-approved notice plans built around specific behavioral-interest assumptions, however, the methodology supporting the program has changed.
The second is browser-level. In April 2025, Google reversed course on its Privacy Sandbox cookie-replacement path, announcing that Chrome would keep its existing third-party-cookie choice model rather than introduce a new standalone prompt.[4] In October 2025, Google announced the retirement of many major Privacy Sandbox technologies, including Topics, Protected Audience, Attribution Reporting, Private Aggregation, and IP Protection.[4] Therefore, third-party cookies remain available in Chrome, subject to user settings. However, Safari and Firefox continue to restrict cross-site tracking by default, ATT still applies in the app environment, and major ad platforms continue to push advertisers toward AI-driven delivery.[1][4] Cookies remained, but the broader trend toward signal loss continued.
The third is regulatory. Eight additional state comprehensive privacy laws took effect in 2025: Delaware, Iowa, Maryland, Minnesota, Nebraska, New Hampshire, New Jersey, and Tennessee.[5] Three more took effect on January 1, 2026: Indiana, Kentucky, and Rhode Island.[5] Revised and new CCPA regulations also became effective on January 1, 2026, though some compliance obligations phase in later.[6] These laws do not operate identically, and they do not categorically ban behavioral targeting. They do, however, affect the conditions under which covered businesses use personal data for targeted advertising, profiling, sensitive-data processing, sale/sharing, and related measurement.[5][6]
Together, these three developments reduce observable user-level signals, limit manual targeting control, and increase legal constraints across the digital infrastructure that has historically supported digital notice.
The Class Action–Specific Problem
These developments affect consumer marketers as well, though largely by complicating performance measurement and audience attribution. In the class action context, the implications are different. The central question is not merely efficiency, but adequacy.
That does not make digital notice the wrong tool. In many cases, it remains the best tool available: more targetable than traditional media, more measurable than publication, and more scalable than most alternatives where class members are unknown. But the case for digital notice has to be made on today’s terms, not on assumptions inherited from an earlier targeting environment.
Unique reach is harder to prove. Frequency capping and deduplication, which is how a declaration arrives at a defensible unique-reach figure, depended on persistent tracking identifiers that have eroded. Deduplication that once relied on measurement now relies on modeling.
Targeting unknown class members has become less precise. Targeting unknown class members has become less precise. The same programmatic plan that isolated purchasers of a specific product three years ago will reach more out-of-class noise today — the targeting hasn't gone away, but the signals it runs on have eroded.
Notice plans can be functionally obsolete before a program even runs. The notice planning playbook hardened during a period when behavioral targeting was more precise, identifiers were more persistent, and reach could be measured rather than modeled. Much of it has not been rebuilt to reflect what has actually happened to that infrastructure since. Plans designed against the old assumptions are still moving through preliminary approval on the basis of reach calculations and targeting language that no longer describes the medium they will run on. Administrators should be asking, before any plan goes into market: Does this plan reflect the targeting environment that actually exists?
Geographic and demographic targeting remain essential tools, and for many programs, they are the right approach. They are also less precise than behavioral targeting, and that trade-off is often the best available option as long as the documentation matches what the targeting can and cannot do.
When programs are built around geographics and demographics, impression and spend levels need to scale upward to maintain in-class coverage. The right kind of scaling is quality inventory at greater volume. Padding with cheap banner display or low-visibility social placements adds impressions to the count without adding real exposure, and plans that lean on that math will deliver fewer in-class impressions than the headline number suggests.
What Administrators Are (Often) Missing
A few elements of how digital reach is produced tend to go unexamined in plan review.
Impressions are not unique individuals. The two numbers were never the same, but the gap between them has widened, and the industry's tools for closing it have weakened.
Vendor reach estimates today rely more heavily on modeling than on direct measurement. The methodology behind the model, including the panel used, the assumptions applied, and the overlap logic, deserves examination alongside the figure itself. How a reach number was derived carries as much evidential weight as the number.
Reach is typically quoted against a syndicated research proxy. These are panel-based research products, widely used across media planning, that provide a stable and externally verified basis for documenting reach against a defined class. Where direct notice (mail, email) and digital notice run against the same class, segments that cannot be cleanly deduplicated are calculated as separate, non-overlapping figures rather than summed under a single headline number. That methodology is defensible. It also has to be made explicit in the declaration, and the discipline of naming what a reach figure includes and excludes is what separates a defensible number from one that quietly aggregates across non-comparable measurement bases.
What Adaptation Looks Like
The adaptations that hold up to court scrutiny are the ones within an administrator's control.
Anchor reach to a defensible proxy, and balance it against the platforms in the plan.
Build the program against syndicated research panels. The proxy is what gets defended in a declaration; vendor numbers are what the program executes against. Documentation should be explicit about which is which, and the proxy alone is not enough. A defensible program reconciles the syndicated proxy with the actual targetable audience available on the platforms in the plan, and the two should sit as close to each other as the available data allows. Where a gap exists between proxy size and platform-available audience, it should be documented.
Account for overlap honestly, across all dimensions of the program.
Overlap exists within digital, between direct and digital, and across traditional and digital. Meta’s properties carry substantial cross-platform overlap on their own, and programmatic display routinely reaches some of the same users reached elsewhere in the plan. A defensible program accounts for that overlap in the planning model, rolls reach up carefully at the program level, and supports the final figure with clear assumptions. Where usable audience data is available for suppression or targeting refinement, it should be integrated to improve precision. Where overlap must be modeled, the methodology should be documented clearly.
Clear the threshold with sufficient volume of quality, not with impression padding.
Best notice practicable is a floor that requires high-quality reach against the right proxy at sufficient scale. A small quantity of premium impressions does not meet the floor any more than a large quantity of cheap ones does. The way to clear the threshold defensibly is the right kind of impression at the right scale. Over-reliance on cheap banner display and untargeted programmatic, or press releases counted as reach rather than as a separate disclosure path, looks impressive on a CPM line but does not move adequacy.
Diversify platforms because audiences are fragmented, not just to hedge platform risk.
Adults in 2026 are scattered across CTV, streaming audio, multiple social platforms, premium publishers, and contextual environments. No single channel, and no single walled garden, reaches a defined class on its own. Cross-channel construction is how a program finds class members where they actually are. The secondary benefit is that no single platform's mid-flight algorithm change can crater the program.
Document methodology as the burden rises.
As targeting precision drops, the documentation burden grows. Record the proxy used, how reach was modeled, what overlap assumptions were applied, and what the program did not count.
Define reach in the declaration.
Counsel and administrators should align on a working definition before the program runs, not after.
Taken together, these adaptations describe a coherent program shape rather than a checklist. Reach is anchored to a defensible proxy and balanced against the platforms in the plan. Overlap is calculated honestly and documented openly. The threshold is cleared with sufficient volume of quality inventory rather than padded with cheap impressions. Platforms are diversified to match where audiences actually are. Methodology is documented as the burden rises, and reach is defined before the program runs rather than after. A program built around those principles is one that can be defended on its own terms, regardless of what platforms or regulators do next.
The Adequacy Implication
Rule 23(c)(2)(B) requires "the best notice that is practicable under the circumstances," and that standard speaks to methods, not just impression volume. The Federal Judicial Center's longstanding guidance, which courts have leaned on for years to evaluate adequacy, describes notice programs reaching roughly seventy to ninety-five percent of class members as generally within the range considered reasonable.[7] That benchmark presumes a measurement environment in which reach can be verified with reasonable precision. The environment that produced those expectations no longer exists in the same form. Fifty million impressions delivered through degraded targeting signals and modeled reach estimates do not carry the evidential weight that fifty million carried in 2019. The number on the declaration may be the same. What sits behind it is not.
The legal field has not yet built shared standards for evaluating digital notice quality in a privacy-constrained, AI-mediated environment. Published scholarship connecting signal loss to Rule 23 adequacy is limited, and that absence is itself part of the picture. Administrators, counsel, and the courts that approve and review notice programs will need to develop those norms. At Élan™, we have built our notice planning around these principles for some time, and we expect the broader field to converge here as scrutiny rises. The administrators and counsel who help shape that conversation early will be better positioned to meet it.
References
Mobile app identifier changes (Apple ATT and Android Advertising ID). Apple Developer — App Tracking Transparency (link); Apple Developer News — User Privacy and Data Use: App Tracking Transparency requirements (link); Google Play Console Help — Advertising ID (link); Android Developers — Best practices for unique identifiers (link).
ATT opt-in rates and methodology variation. Adjust — ATT opt-in rates: 2025 data & benchmarks (link); Flurry — ATT Opt-In Rate Monthly Updates (link); FTC / Skiera — Economic Impact of Opt-in versus Opt-out Requirements for Personal Data Usage: The Case of Apple’s App Tracking Transparency (link); Purchasely — ATT Opt-In Rates In 2025 (link); AppsFlyer — Data Reveals Increase in User Opt-In Rates and Ad Spend on iOS Three Years After Apple’s App Tracking Transparency (link).
Meta detailed-targeting consolidation and exclusion changes. Social Media Today — Meta Is Consolidating More of Its Detailed Ad Targeting Options (link); MediaPost — Meta Further Consolidates Ad-Targeting Options (link); Social Media Today — Meta Removes Detailed Targeting Exclusions from Ad Campaigns (link); ThoughtMetric — What Meta’s Targeting Updates Mean for E-Commerce Advertisers (link); Brandwatch — Meta Changes to Detailed Targeting Interests in Advertise (link).
Google Privacy Sandbox reversal and API retirement. Google Privacy Sandbox — Next steps for Privacy Sandbox and tracking protections in Chrome (link); Google Privacy Sandbox / Google Help — Update on plans for Privacy Sandbox technologies (link); Adweek — Google’s Privacy Sandbox Is Officially Dead (link).
Comprehensive state privacy laws taking effect in 2025–2026. IAPP — US State Privacy Legislation Tracker (link); MultiState — All of the Comprehensive Privacy Laws That Take Effect in 2026 (link); Jenner & Block — New US State Privacy Laws Taking Effect in 2025 (link); Perkins Coie — Privacy Law Recap 2025 (link); Baker Donelson — Privacy Laws Ring in the New Year (link); Bloomberg Law — Which States Have Consumer Data Privacy Laws? (link).
CCPA regulations effective January 1, 2026. California Privacy Protection Agency — California Consumer Privacy Act Regulations, effective Jan. 1, 2026 (PDF link); California Privacy Protection Agency — CCPA Updates, Insurance, Cybersecurity Audits, Risk Assessments, and ADMT Regulations (link); Greenberg Traurig — Revised and New CCPA Regulations Set to Take Effect on Jan. 1, 2026 (link); Inside Privacy — California Finalizes Updates to Existing CCPA Regulations (link).
Federal Judicial Center class notice reach benchmark. Federal Judicial Center — Judges’ Class Action Notice and Claims Process Checklist and Plain Language Guide (2010) (link).
Alexander Stephens is the Client Director of Legal Notification Programs at Élan, where he serves as a strategic partner to some of the country's leading Class Action Administrators. He has directed more than 50 complex, multi-channel notice campaigns spanning digital, print, radio, and press across a broad range of class action and mass tort matters.
