Executive Summary
Criteo’s latest revenue downgrade is not an isolated earnings miss — it is the financial surface of a structural collapse in probabilistic retargeting that has been building for three years. Cookie deprecation timelines, Apple’s App Tracking Transparency framework, and rising consumer opt-out rates have eroded the signal density that made cross-site retargeting economically viable. The retargeting platforms themselves acknowledge this: Criteo’s forward guidance now hinges on agentic AI ad placement, a product category that does not yet exist at scale and whose timeline is controlled by browser vendors, AI agent developers, and publishers with undefined incentive structures. For performance marketers accountable to ROAS, CPA, and acquisition targets this quarter, betting budget on a speculative product roadmap is not a defensible strategy. This whitepaper maps the structural reasons probabilistic retargeting is weakening, quantifies the signal loss driving that decline, and makes the case that programmatic direct mail — built on deterministic identity, first-party CRM data, and auditable matchback attribution — is the acquisition channel best positioned to absorb the budget that legacy retargeting can no longer justify. Every capability described here is operational today, measurable against the brand’s own transaction data, and independent of any browser, device, or platform policy decision.
The Probabilistic Retargeting Model Is Losing Its Inputs
The retargeting model that defined the last decade of performance marketing rested on three signal sources: third-party cookies for cross-site identity resolution, mobile device IDs for app-to-web graph stitching, and high opt-in rates that kept the addressable audience pool large enough to sustain frequency at efficient CPMs. All three are contracting simultaneously.
Safari and Firefox eliminated third-party cookies years ago, compressing the cookie-addressable browser population to Chrome’s share. Chrome’s own Privacy Sandbox APIs have fragmented that remaining pool, with the Topics API and Attribution Reporting API introducing noise and latency that degrade the signal quality retargeters depend on for real-time bidding decisioning. Even in the most optimistic cookie-survival scenario, the addressable universe for probabilistic cross-site retargeting is substantially smaller than it was in 2021.
On mobile, Apple’s ATT framework has held opt-in rates well below 50% since its 2021 launch, meaning retargeters can deterministically track a minority of iOS users. Android’s Privacy Sandbox is following a similar trajectory, with Google phasing in the Attribution Reporting API and limiting the device-level signals that underpin cross-app audience graphs.
Consumer behavior is compounding platform-level restrictions. Privacy surveys consistently show a majority of consumers have taken active steps to limit ad tracking, and that number is growing year over year. This is not a policy risk on a timeline — it is an already-realized contraction in the raw signal inputs that retargeting platforms need to function.
Criteo’s most recent earnings call made this explicit. Management guided investors toward a future revenue mix anchored by “commerce media” and agentic AI ad placement — neither of which is generating material revenue today. The company’s core retargeting business has declined year-over-year for multiple consecutive quarters. For advertisers, the translation is straightforward: the retargeting platform you are buying media from is telling you its own core product is structurally impaired, and its replacement depends on someone else’s development timeline.
The question for performance marketing teams is not whether probabilistic retargeting will continue to exist — it will, at reduced scale and efficiency — but whether the marginal acquisition dollar is better deployed in a channel whose targeting, identity resolution, and measurement are not degrading with every browser update and privacy policy change.
Why Deterministic, First-Party Channels Are Structurally Immune to Signal Loss
The channels least affected by signal loss share three characteristics: they resolve identity deterministically (a known individual at a known address, not a probabilistic device-graph match), they activate against first-party data the brand already owns, and they measure performance against the brand’s own transaction records rather than a platform’s self-reported conversion pixel.
Programmatic direct mail meets all three by design — not as a privacy-safe workaround, but as the foundational architecture of the channel.
Identity resolution is deterministic by default. Every mail piece is delivered to a verified physical address tied to a named individual or household. No probabilistic graph stitching, no cookie syncing, no device ID matching. The identity layer is the U.S. Postal Service’s address database — the most complete household-level identity graph in the country. This identity does not degrade when a browser updates its privacy policy, when a consumer opts out of app tracking, or when a walled garden restricts signal access. It is structurally independent of the digital signal chain.
Targeting activates against first-party CRM data the brand controls. Postie ingests a brand’s customer file — purchasers, subscribers, high-LTV segments, lapsed buyers — and resolves it to mailable addresses. From that first-party seed, ML-powered lookalike models build prospect audiences using transaction signals enriched with deterministic demographic and behavioral data. The enrichment layer is built on survey-sourced, public-record, and transaction-verified attributes — not on passive device tracking or SDK-harvested behavioral inference. The compliance profile is fundamentally different from the brokered device-graph data that regulators are actively restricting.
Matchback attribution is auditable at the household level. Matchback attribution — the standard measurement methodology for programmatic direct mail — works by matching the mail file (every household that received a piece) against the brand’s post-campaign transaction file. Every attributed conversion traces to a specific household, a specific mail drop date, and a verified purchase in the brand’s own records. No black-box conversion pixel, no platform-controlled identity graph between the exposure and the outcome, and no probabilistic lift model the brand cannot independently verify. The brand controls both sides of the measurement equation.
Four Strategies for Reallocating Retargeting Budget to Programmatic Direct Mail
The reallocation case is strongest when it starts with the retargeting spend that is already underperforming.
1. Flag retargeting campaigns where frequency is rising and conversion rates are falling.
Signal loss does not show up as a single-day failure. It manifests as a gradual increase in frequency against a shrinking addressable pool, paired with declining conversion rates as the remaining targetable audience is over-served. Pull 180-day trend data from your retargeting platform. Flag every campaign where frequency has increased meaningfully while conversion rate has declined. These campaigns are the clearest signal that the addressable audience is contracting — and they represent the budget most likely to deliver better returns through a channel with a structurally different reach model. Programmatic direct mail reaches households that are invisible to cookie-based retargeting: desktop users on Safari and Firefox, the majority of iOS users who opted out of ATT, and the growing share of consumers who have actively blocked ad tracking.
2. Build suppression-aware lookalike audiences from your highest-value CRM segments.
The most efficient direct mail campaigns start with the brand’s own first-party transaction data. Upload your top-decile purchasers by LTV, your highest repeat-purchase-rate cohorts, or your strongest CRM segments to Postie. The ML-powered lookalike engine identifies prospect households that share the demographic, behavioral, and transactional attributes of your best customers, while suppressing existing customers and recent purchasers to ensure every piece drives net-new acquisition. This is precision audience modeling built on the same first-party signals that the best digital acquisition campaigns use — activated through a channel that does not require cookie consent, device ID access, or platform opt-in to reach the recipient.
3. Run holdout-based incrementality tests against your remaining retargeting spend.
The most credible way to compare channel performance is a controlled holdout test where the brand defines both the exposed and control groups. With programmatic direct mail, this is operationally straightforward: randomly withhold a statistically significant subset of your mail audience, then compare conversion rates between the mailed and holdout groups using your own transaction data. The incrementality signal is clean because the identity is deterministic — you know exactly who received a piece and who did not, and you can validate conversions against your own purchase file without relying on a platform’s self-reported metrics. Run this test concurrently with your retargeting spend and compare incremental CPA across both channels. The comparison is especially informative in categories where signal loss has compressed the retargetable audience — DTC, home goods, financial services, and subscription commerce.
4. Integrate direct mail into trigger-based lifecycle flows alongside digital.
Programmatic direct mail is not limited to batch prospecting. Trigger-based campaigns — fired on behavioral or lifecycle signals like cart abandonment, subscription lapse, or post-purchase upsell windows — can run on the same CRM event triggers that drive your email and SMS flows. The difference is reach: direct mail lands at the household regardless of email deliverability, inbox placement, or notification opt-in status. For brands seeing declining email engagement and SMS fatigue driving higher unsubscribe rates, adding a physical touchpoint to lifecycle flows can recover conversions from audiences that digital channels are failing to reach. The cost per piece is higher than an email send — but the conversion rate differential on high-intent triggers often more than compensates, particularly when measured on an incremental-CPA basis.
Conclusion
The decline in retargeting revenue across the category is not a company-specific problem. It is the financial manifestation of a structural shift affecting every channel built on probabilistic, third-party signal resolution. ATT opt-out rates, Privacy Sandbox restrictions, rising consumer privacy behavior, and accelerating regulatory enforcement against the brokered data supply chain are all compressing the addressable audience for retargeting simultaneously. The platforms themselves are telling advertisers that the future lies in product categories that are speculative, early-stage, and dependent on ecosystem adoption timelines no single vendor controls.
Performance marketers do not have the luxury of waiting for speculative roadmaps. Acquisition targets are quarterly. Direct mail ROAS thresholds are auditable. Finance teams reviewing media budgets will not accept “we’re waiting for agentic AI” as a rationale for continuing to allocate spend to a structurally impaired channel.
Programmatic direct mail offers what the next generation of retargeting promises but cannot yet deliver: deterministic identity resolution at the household level, targeting built on first-party CRM data the brand owns, enrichment from consent-native demographic sources, and matchback attribution the brand can audit against its own transaction records. No browser policy, no device-level opt-out, and no platform algorithm change can degrade the signal chain — because the signal chain runs on the brand’s own data, resolved to physical addresses, and measured against the brand’s own sales.
The budget that legacy retargeting can no longer justify does not need to wait for the next platform innovation cycle. See how Postie’s lookalike modeling and matchback attribution work against your own data.
