The supplied research points to useful changes in data and advertising. The plan keeps the actionable principles and treats vendor benchmarks as claims to test.
Our recommendation
Start with accurate first-party records, differentiated creative, dependable lead handling, and a reconciled sales ledger. Add increasingly complex enrichment, prediction, and causal measurement only when they improve a specific decision at an affordable cost.
What the source review changes
SUPPORTED PLATFORM CAPABILITY
Server-side events need deduplication.
Meta documents using matching event names and IDs when Pixel and Conversions API report the same event. Our plan includes that requirement, plus the agency’s own duplicate protection. This improves measurement integrity; it does not by itself prove incremental revenue.
Google’s current guidance describes a June 15, 2026 migration of offline and enhanced-conversion uploads toward the Data Manager API, with legacy-access conditions. New client implementations should validate the supported route and account permissions rather than assume an older Google Ads API workflow is suitable.
Google recommends connecting value data and meaningful business outcomes to bidding. We use actual contribution or calibrated expected value where appropriate. Campaign eligibility and data-volume requirements differ; the plan does not present a single conversion threshold as universal.
Meta’s October 28, 2026 engineering post describes a new retrieval architecture and reports platform-level gains. It supports the importance of effective creative and machine learning. It does not establish a universal four-to-six-ad quota, a fixed testing period, or a guaranteed gain for a McElroy client.
Identity precision is not a universal accuracy score.
Truthset’s Identity 3.0 page attributes its comparisons to commissioned CIMM benchmarking and Truthset estimates. Its 90–100% figure refers to a specified validated 1:1 matching modality, not all customer data, demographic accuracy, contact deliverability, or legal permission to advertise.
Meridian documentation explains that history, model complexity, geography, and independent variation affect information quality. Even two years of weekly data can be insufficient for a complex national model. Our small-business pilot starts with reconciled CRM cohorts and adds specialist modeling only when justified.
AI discovery still rests on useful, accessible content.
Google says existing SEO best practices apply to AI Overviews and AI Mode, with no special markup requirement or guaranteed inclusion. The agency therefore emphasizes original expertise, crawlable content, and qualified inquiries rather than promising AI rankings.
The FTC’s CAN-SPAM guide covers commercial B2B email, truthful sender information, a valid postal address, and an effective opt-out. The agency’s proposed operating policy is to honor suppression immediately. Platform terms and other applicable requirements must also be checked.
Vendor rankings and pricing: the supplied database counts, accuracy claims, package prices, and MCP availability were not independently benchmarked across all vendors. Test your niche and obtain current quotes.
Waterfall lift: fill rate is not accuracy. A richer match can still be wrong, stale, unusable, or duplicated.
Monthly decay: repeated 2–3% monthly decay compounds to about 21.5–30.6% annually. It does not support a single fixed annual rate for every dataset.
CPA uplift figures: broad platform averages and agency anecdotes are not forecasts for a new account. No guaranteed 17.8% improvement appears in the offer.
Clean-room costs and spend cutoffs: the quoted $879,000 average and fixed spend bands are not treated as universal thresholds. A specific use case and expected decision value should determine investment.
Robyn maintenance and universal timing rules: the supplied 2026 staffing claims and fixed creative refresh/learning windows are not relied on in the launch plan.
Privacy: hashing or server-side routing does not authorize collection, remove notice/consent duties, or make personal data anonymous.
This is a selective review of decision-critical claims, not a complete independent audit of every supplied citation. Sources describe their own products and methods; commercial interests and study populations matter.