Change log

Improved Google connection-status guidance and authorization messages.

Added guidance for renewing Google connections. Connections requiring renewal remain clearly identified.

Tested newer AI models on a consistent 15-question panel, comparing response quality and source coverage. No production model change.

Explored alternative processing options to improve efficiency. These options have not been activated.

Compared 15 consumer ChatGPT responses with API-generated answers to evaluate customer-experience fidelity. Results remain separate.

Prototyped richer response capture with formatted text, tables, source cards, screenshots and images. Full citation coverage remains under evaluation; this is not yet a production feature.

Completed an 80-question consumer-response capture trial to evaluate coverage and presentation. Trial data remains separate from production.

Scheduled two further shadow-trial passes for October 11 and 12. The existing production collector remains unchanged.

Scheduled consistent sentiment and wording comparisons across the baseline and both trial days.

Added the footer-linked change log, then replaced the short milestone summary with this searchable daily history. Future entries distinguish delivered work, fixes, tests and scheduled work.

Scheduled regular release-history maintenance to keep documented improvements current.

Improved Yelp listing-level count validation and separated public totals from imported-review counts.

Validated Yelp count reporting against historical captures and selected live listings.

Added explicit coverage indicators for verified Yelp listings. Unknown coverage is not counted as zero.

Restored validated Yelp count history from August 12, retaining supported observations through collection gaps and clearly indicating older data.

Improved Trustpilot historical-count validation while retaining original observations.

Strengthened URL, brand, business-header and count validation to reject unrelated listings and conflicting headers.

Verified ConsumerAffairs counts without changing them, and verified both company report copies and chart scales after count repairs.

Enabled automatic analysis of new and edited reviews within minutes of import.

Strengthened duplicate protection, retries and result validation, with automatic dashboard refresh and company-report synchronization.

Consolidated automatic analysis into a single processing pipeline and verified queued results were synchronized.

Improved analysis-status reporting to reflect current queue activity rather than historical run timestamps.

Expanded ConsumerAffairs date-format support and corrected 45 historical publication dates from saved evidence. Review text and history remain unchanged.

Verified the first complete multi-page ConsumerAffairs daily collection for both brands.

Added Total reviews / Star rating controls to the existing trajectory widget. All sources show aggregate counts, never a blended public star rating.

Improved All brands and Google filter handling for consistent chart coverage.

Added automatic and manual daily, weekly, monthly, quarterly and yearly time scales based on the selected date range. Verified both dashboards.

Replaced first-page-only daily capture with all available-page traversal, capped at 50 pages per brand, with overlap deduplication and new/edited counts. Tests passed; first live daily verification followed on October 6.

Added collection-completeness checks and source-evidence validation. First live verification followed on October 6.

Completed a verified Google inventory refresh with full-page coverage, preserving earlier captures.

Completed the official Google API transition across 178 agreed website-scope entries and verified company synchronization. Retained the former collection method for recovery.

Strengthened review-availability confirmation: two complete daily inventories and two independent source checks at least 24 hours apart.

Added evidence-reset and replay protections. Confirmed unavailability does not imply a known removal reason.

Added controlled automatic recovery for temporary collection issues, with safeguards for access and identity errors.

Verified current reviews, historical versions and status events across company reporting. Existing removal history remains intact.

Reconciled 17 verified location aliases while preserving their reviews and historical snapshots.

Added verified Springfield coverage and clarified alternate-location records.

Validated Birmingham listing identity across public, official and historical evidence.

Improved Seattle listing selection and completed a verified inventory, retaining prior source evidence.

Registered the company-accessible Minneapolis listing and collected its first official inventory.

Documented five locations outside connected-account API coverage. Their website and review history remain retained.

Added a visible API-coverage exceptions section with dated observations and clearly identified unknown coverage.

Expanded verified historical matching and official review imports while retaining safeguards for unresolved identities.

Verified collection scope and company reporting before the final API transition.

Completed both company-owned OAuth connections and account/location discovery, then activated fully paginated daily official review collection at 6:30 AM Chicago.

Fixed setup-page rendering and made the exact shared authorized redirect URI selectable and copyable in the walkthrough and brand cards.

Added durable review/page capture, stable native IDs, edit tracking for review text, stars and owner replies, protected ingestion and inventory completeness gates.

Published a side-by-side reconciliation screen for exact matches, possible edits, official-only reviews and uncertain identities; ambiguous cases remain held.

Expanded evidence-based review matching using stable identities, publication details and historical content; uncertain cases remain separate.

Improved historical reviewer and owner-response attribution using official source evidence, retaining prior versions.

Enabled verified identity matching when minor source-count differences exist, without relaxing import or removal safeguards.

Improved synchronization of historical reviewer names, ratings and dates.

Separated active unverified legacy Google records from API-verified current metrics without deleting them or interpreting unlinked history as removal.

Launched secure, company-owned Google connections with a guided brand-by-brand setup.

Added a four-step connection walkthrough and a Settings/Data Status entry point.

Simplified the connection setup sequence and highlighted the next required action.

Built the company-owned Google connection flow with secure, separate credential storage. Production launch followed on October 1.

Completed integration-readiness checks while preserving existing reporting data.

Designed official Google collection to use the existing validated reporting pipeline.

Restored the historical review baseline and corrected public score extraction to use the visible matching profile rather than unrelated structured data.

Made the ConsumerAffairs score chart honor the selected full date window instead of only recent points. Genuine no-snapshot dates remain gaps.

Improved ConsumerAffairs profile-score validation against the corresponding public listing.

Changed empty period wording to “No published reviews” rather than implying that the entire source has no data.

Added a default combined Google AI Overviews + ChatGPT view with source-labelled evidence, an executive snapshot and clearly framed detailed metrics.

Replaced stacked sentiment trends with separate recommendation, positive, mixed and negative lines plus brand-reading volume. Missing capture periods are gaps, not neutral sentiment.

Aligned headline and trend denominators to all brand readings. Explicit-call-only recommendation ratios remain supporting context, not the primary rate.

Added visible KPI help, structured hover cards, independent On/Off trend controls and the definition of mixed language. Improved responsive layout and chart labels.

Published the updated company Worker and verified both report slots, counts, live pages and APIs against the source.

Launched independently managed company reporting with access controls and a validated fallback report.

Enabled hourly incremental synchronization with validation before activating updated reports.

Added a weekly full-report fallback alongside incremental synchronization.

Verified report consistency, relationships and live acceptance checks. Unsuccessful transfers preserve the last validated report.

Validated staged reporting snapshots in preparation for the company-reporting launch.

Prepared validated reporting packages for the company migration. Live activation followed on September 24.

Established the independent company-reporting architecture with validated transfers and a shared collection and analysis pipeline.

Verified recurring cloud-based ChatGPT collection. The exact original activation date is not established by this verification.

Enabled and verified daily Google AI Overview collection for the fixed 80-question panel.

Improved page headings, filter reset, category comparison deltas and navigation consistency. Excluded malformed citation cards lacking a usable domain while retaining raw evidence.

Corrected Google monthly growth to compare only locations present in both periods, separating actual comparable growth from older reviews introduced by newly tracked locations.

Made YTD Google sentiment use stored January–August monthly history rather than only the current capture.

Launched the protected AIO dashboard and separate permanent storage for prompts, raw provider output, answer text, citations, brand/competitor mentions and run history.

Validated a five-prompt test, then a 60-prompt baseline. Stored exact answer wording and did not invent citations absent from provider output.

Expanded to the fixed 80-prompt panel: 30 branded reputation prompts and 50 non-branded category/comparison prompts, with no daily prompt rotation.

Added branded/non-branded filters, category visibility and sentiment, historical trends, clickable brand/sentiment/competitor drill-downs and readable response formatting.

Added review-dashboard date and comparison options to AIO, linked source cards and consistent KPI/chart direct-brand mention counts.

Repaired incomplete generated review links to open the correct profile. Preserved genuine review-share links and differentiated “Open this Google review” from “Open Google profile.”

Added a clear “no comparable score snapshot” label when historical comparison data is unavailable.

Improved BBB recurring-collection reliability and verified successful recovery.

Improved reconciliation-record storage and dashboard availability.

Added selective BBB collection recovery, preserving completed captures and importing validated evidence.

Expanded recovery of complete Google review text while retaining strict eligibility checks for removal decisions.

Staggered local collectors at hourly intervals to reduce browser contention. Empty profile-only captures remained queued for later text recovery, not imported as empty reviews.

Hardened Zippy directory-sync handling to retain the last verified 95-location roster and warn when the website unexpectedly returned no usable roster.

Added persistent reconciliation history and freshness indicators, including no-change collections. Live validation followed after this initial build.

Added login-triggered missed-collection catch-up with a startup delay and serial execution of only that day’s unsuccessful collectors.

Replaced routine unlimited full scrolling with profile-first checks and bounded newest-first reconciliation; unchanged profiles skip text extraction, mismatches stay explicit.

Added priority retries for incomplete captures and reduced normal Google batch size from 70 to 25 to fit the task window.

Enabled independent import of validated listing captures for more resilient collections.

Improved directory-sync retries and timeout handling while preserving the last verified roster.

Added a read-only scheduled audit for stable-ID collisions, cross-profile conflicts, analysis variants and same-text clusters. It reports candidates; it does not automatically merge or delete.

Expanded selected-period metrics to the complete filtered database while retaining paginated review browsing.

Added clearly labelled review-count-weighted Google location scores and a separate imported-review average when a BBB public profile snapshot is unavailable.

Filled the complete 90-day weekly range, including zero-volume weeks. Prior-year comparison no longer limits chart coverage.

Ignored stale responses during rapid filter changes and verified full-database search and filter combinations.

Optimized analysis-queue performance and improved resilience to temporary service errors.

Refactored shared helpers, analysis queue, collector state, integrity audit and dashboard query/cache handling into focused modules in separate verified passes.

Standardized 21 historical publication dates and improved timeline consistency for future imports.

Added provider-specific prefilled search shortcuts and reversible listing confirmation. Undo removes the future collection association but preserves existing reviews/history.

Allowed manual BBB approval after a normal browser check when automated verification receives 403, while still rejecting malformed/non-BBB URLs.

Introduced Claimed, Unclaimed and No public signal labels with supporting evidence and improved request handling.

Added 110 evidence-supported listings to the recurring roster, leaving ambiguous or missing profiles unresolved instead of attaching regional/HQ profiles by guesswork.

Added profile-first BBB and Yelp daily checks with targeted review retrieval for changed/new listings and periodic audits. Partial captures remain ineligible for removals.

Captured expanded review text, corrected brand filtering and differentiated profile links from genuine review-share links.

Added Today and Yesterday date options and clarified that historical backfills are newly imported records, not newly published reviews.

Removed dashboard build/deploy work from the critical daily location-sync task so it could finish before review collectors.

Launched the official-website-based 183-location coverage matrix with Google, Yelp and BBB links, brand/platform/search filters and gap prioritization.

Added candidate confidence, rationale, persistent confirmation, manual missing-listing URL entry and verification before adding to collection.

Strengthened geographic validation, improved Albany BBB mapping and simplified confirmed-listing views.

Added separate address, phone and ownership evidence with tooltips showing official versus listing values; unavailable evidence is not treated as a mismatch or unclaimed ownership.

Improved official-address normalization, verified location details and interface typography.

Added deduplicated listing registration, pending Place ID resolution and automatic activation after verified collector discovery.

Reconciled discovery against the official roster and BBB branch directory, deduplicating address aliases by review pool rather than counting every URL separately.

Imported 288 unique historical reviews with stable IDs, dates, ratings, text and preserved source evidence; added BBB to filters, search and analysis.

Installed daily review checks and monthly discovery/reconciliation. Metadata-only exports were explicitly rejected as review backfills.

Validated a complete Houston fallback inventory against the exact public count and native IDs; retained the historical confirmation evidence. Later October removal rules supersede this older policy.

Imported the historical Yelp backfill with owner responses, stable IDs and duplicate protection; recovered 13 records omitted by a 500-review export cap. All new text entered analysis.

Replaced preview/sample-based KPIs and topic/star totals with complete database aggregates while retaining paginated detail rows to avoid resource limits.

Added full-database search across text, title, reviewer, location, topics and AI summaries, with combined filters and full filtered CSV export.

Added complete all-time yearly/monthly history, zero-volume periods, horizontal monthly timelines, readable labels and inward-opening edge tooltips.

Added YTD, brand legends, multi-topic hover counts and explicit positive/neutral/negative sentiment presentation with comparison deltas.

Added location search, View all, detailed location sorting, current/prior counts and ratings, rating-growth toggle and an unrated/no-reviews band.

Separated new-to-history locations from comparable growth and public totals from imported text and analyzed reviews. Unrated profiles no longer lower averages or appear as poor-rated locations.

Recovered 60 missing historical dates from exact published/update evidence rather than estimates.

Improved Google fallback collection and Trustpilot capture-to-import validation.

Accelerated the historical backlog runner with larger serialized batches and automatic continuation. This historical runner is superseded by the October 6 cloud worker.

Introduced source-specific score reporting and an Executive Brief with momentum, strongest signals and analysis coverage.

Added combined topic, sentiment and star drill-downs, clickable distribution rows and review-card topic tags.

Made topic sentiment mention-based across primary/secondary topics while keeping overall sentiment review-based; one review can praise drivers and criticize billing.

Added global Processing & Collection metrics with Waiting/In progress separation, saved-result counts and refresh timestamps.

Added current/prior month and custom dates, dynamic daily/weekly/monthly review velocity, numeric scales, responsive labels and hover detail.

Improved deployment consistency and reduced dashboard payload size while preserving full-database metrics.

Restored the daily Google task, updated changed selectors, enforced newest sorting/full text expansion and rejected invalid captures.

Established the official US brand websites as the master roster and retained ambiguous platform matches separately. Added 12 distinct supported Google profiles.

Completed resumable Yelp discovery, deterministic vetting and 242-profile enrichment with saved address/contact/status evidence for historical backfill. Uncertain or address-hidden profiles were disclosed.

Validated both historical page sets and imported 1,440 PACK-RAT and 289 Zippy unique visible reviews. Preserved replies and raw records; older screenshot duplicates were retained but hidden.

Imported 466 PACK-RAT and 69 Zippy historical reviews, with duplicate fallback pages excluded. Configured the tested homepage-first daily collector with local raw HTML retention.

Configured and tested daily capture for both brands with scores, totals, star distributions, exact dates and replies.

Added an automatic hash-keyed backlog queue with bounded batches, separate analysis storage, retries and no duplicate completed analyses.

Repaired missing/future-date filtering; made brand/platform/date selection affect KPIs, charts, topics, sentiment and exports, then aligned search-specific cards and comparisons.

Added explicit Search/Clear controls, Enter support, highlighted matches and previous-period/year comparison options.

Imported 174 locations and 3,412 monthly snapshots from the existing workbook, then 21,549 unique Google reviews, preserving rating-only records and owner responses.

Added adaptive profile collection, count/score trajectories and exact-removal history with complete-inventory gates. The original scraped policy is superseded by October’s native-API safeguards.

Added daily collection and sentiment-quality monitoring without automatically spending credits or rewriting production data.

Launched Review Monitor with a hosted dashboard, persistent review storage, source-evidence retention and CSV support.

Added the protected Trustpilot paste/manual importer with preview and duplicate checks, followed by screenshot and saved-HTML import options.

Added clear empty states and source-coverage indicators.

Made tall screenshot parsing sequential and memory-bounded with visible progress and errors; excluded Trustpilot’s AI summary from customer reviews.

Improved saved-page extraction and made import errors easier to understand.

Built resumable Trustpilot historical capture with pagination, source validation and retained capture evidence.

Validated the initial historical Trustpilot capture and source-page checks.

Established review sentiment and topic-analysis configuration.