AAAC/organic investigation

Section 06 · What's being considered, ranked by the data

Priorities & open questions

The strategic frame: reworking the PAA content is unlikely to restore calls, because it never drove them. The commercial gap and the local-pack presence are where call volume actually lives.

§ 6.1 — The action stack

Data-ranked, as it stands after Round 5

  1. Kill the Dallas tel-link hijack

    Delete/replace the HFCM "Bing Ads Body" snippet in wp-admin; fire UET from GTM instead. Restores call measurement dead since Oct 23 2024 and removes a plausible source of real lost calls. Cost: minutes.

  2. Fix the blind-analytics sites

    austin (GTM container lookup + remove the staging-URL UA snippet from production), pittsburgh & sc-ohio (no GTM at all), fort-pierce (zero analytics), and retire the orphan florence GSC property (the real franchise is centralky).

  3. Decide the Florida cluster

    fort-pierce's 42-town duplicate map + the shared three-site NAP: consolidate or genuinely partition. R5 reset expectations — this is a fragility fix (doorway-policy exposure, 3-way brand split), not a traffic-restoration lever on its own. Necessary-but-not-sufficient.

  4. Re-anchor Treasure Coast

    Pick a real anchor city consistent with the actual pin/GBP (schema says Palm Bay — verify where the GBP actually is), promote town pages, demote the region term to brand flavor. Validate with a GBP grid scan (Local Falcon or similar).

  5. Rebuild Dallas as a real Dallas site

    De-clone the *-fort-worth-slugged service pages, build real city pages for its 18 suburbs, expand areaServed beyond "Dallas; Irving; Coppell". The prize: 50.9k commercial-geo impressions in 5.5 months going unconverted.

  6. Content forensics on the survivors

    Diff columbus's surviving blog corpus (83% blog, +11% through the update) against the collapsed corpora (TC, phoenix, dayton) to learn what the June reassessment kept.

  7. Close the data gaps

    GBP insights for both markets, real call logs, GSC properties for dallas / fort-pierce / palm-beaches, and inventory of the four production sites found only in the rollup (columbia, kansas, clarksville, rockwall).

§ 6.2 — Quick fixes

Site defects fixable now

§ 6.3 — Open questions

What we still can't answer

The contradiction

All-time-high call events vs a franchisee complaint

TC total call events peaked (111, Jul 2026) while the franchisee reported a bad two weeks. Is it (a) phantom paid events (Bing CPC fires at 10–14%/session, ~40× organic), (b) a GBP/local-pack call decline invisible to GA and GSC, or (c) lower-quality paid clicks? Discriminators: real call logs, GBP insights, and a Bing UET/GTM double-fire audit.

Attribution

What exactly was Jun 11–25?

The May core update no longer fits (network peaked May 26); the Jun 24–26 spam update landed after the drops began. Unconfirmed mid-June volatility + AI-surface expansion remain the candidates — unresolvable from ranking data alone.

Separate phenomena

Dallas's April bleed and Tampa's September 2025 break

Both were disentangled from the June event by the seasonality audit and need their own explanations if pursued. Dallas has no pre-bubble baseline (GA starts Oct 2024); Tampa's level shift predates everything discussed here.

Anomalies on file

Unexplained but logged

The TC homepage call-rate step-up dating to May 2025 (unexplained site/instrumentation change); the possibility of bot traffic in the 2025 "yahoo" burst; a trailing-slash URL variant split that began mid-2025.

Data that would settle the biggest questions


Appendix

Methods & caveats

Sources

  • GA4 API: properties 390349966 (TC), 459755935 (Dallas), legacy rollup 387250340, 390759779 (centralky).
  • Search Console API: sc-domain properties; Dallas via the root domain property, page-filtered (no Dallas property exists).
  • Live-site HTML / WordPress REST / schema inspection; Google status dashboard; industry reporting.
  • ~12 research agents across five rounds, all on 2026-08-14.

Known limits

  • GSC retention is 16 months — history starts 2025-04-01, so the pre-PAA baseline is thin and 2024 comparisons lean on GA4.
  • 2025 is unusable as a seasonal baseline — bubble-contaminated (its summer/spring ratio of ~1.12 vs the 2024 norm of 0.75 is what misled Rounds 1–2).
  • Small n on calls: page-class calling-session counts are n≤21 (Poisson SE ~3.5–4.5); the session/impression denominators carry the statistical weight.
  • GA4 and GSC measure different things (sessions vs clicks); where both exist their seasonal ratios agree within 4 points.
  • Event counts overstate callers — one 2024 session produced 12 call events; calling sessions are used throughout.
  • Dallas GA history starts Oct 2024; Dallas call tracking is blind after Oct 23 2024 (the hijack).