A deep-research pass on Meta's advertiser reputation systems — ACE percentiles, B-buckets, and the feedback-to-auction pipeline — with every claim adversarially verified against Meta's own live documentation.
Meta officially states that customer feedback — a global post-click survey generating 200M+ responses per year — "shapes the ad auction" as an input to the ad-quality term, in both directions: reward and penalty.
But the honest version is a three-system split, and evidence strength differs sharply by system. The ACE percentile and B-bucket layer specifically has zero public documentation — the rep-sourced intel from the Gonza meetings is ahead of anything public, including the practitioner reverse-engineering community.
The rep's claim that "the score doesn't affect performance" can be technically true of the displayed ACE percentile while the underlying feedback signals demonstrably feed the auction. And the "temporary auction mix shift" framing of the CPM hit is fully consistent with Meta's documented architecture.
0–5 score from post-purchase surveys of likely buyers. Documented enforcement: 1–2 = penalty band ("reach fewer people, cost more"), <1 = restricted from advertising.
Hide-ad feedback and low-quality-attribute detection sit structurally inside the total-value formula that decides every auction.
The percentile ranking (Bottom 40 → Top 60) and the B3→B4 bucket system Gonza surfaced. Nothing public documents its mechanics or auction linkage.
Each finding survived a 3-judge adversarial verification panel, checked verbatim against Meta's Business Help Centre on July 10, 2026.
Meta's help page section header is literally "How customer feedback shapes the ad auction." The global survey generates 200M+ responses annually across all ad types and verticals, and works in both directions:
"Advertisers with positive customer feedback can see improved performance… Conversely, customer feedback signals can help reduce the likelihood that people have negative experiences with ads."
Auction winner = highest total value = bid + estimated action rates + ad quality. Ad quality explicitly includes user feedback (viewing/hiding the ad) plus automated detection of low-quality attributes — withholding information, sensationalized language, engagement bait.
0–5 score from post-purchase surveys. 1–2 = "Under Penalty": "ads will reach fewer people and cost more" — a genuine reach/CPM hit. Below 1 = Restricted: the portfolio can no longer advertise or sell on Meta platforms.
"We may use data about you and your ad account to make adjustments that could affect auction outcomes and prices… enforce our terms, policies and other standards including fostering a safe and trustworthy environment."
This is the strongest primary-source basis for bucket-style account-level treatment — and it makes the "temporary auction mix shift" explanation of the CPM hit architecturally consistent. Adjustments are capped at the bid.
The score "represents the feedback received from people who have likely made a purchase from Shops or ads." This mechanically explains the conversion-volume data thresholds: Spa Dr. at $15K/day with few purchase events gets surveyed too little to have a score — structural, not fixable by spend. Secondary sources report a ~10-response minimum before a score displays. Also directionally consistent with (but does not confirm) the ad-comment → survey correlation, since both are engagement-based sampling triggers.
Quality Ranking diagnostics officially define "Average" as the 35th–55th percentile of ads competing for the same audience, with below-average tiers at bottom 35% / 20% / 10%. Methodological precedent for the ACE percentile framing — but it's an ad-level diagnostic; no public source maps it to B3/B4.
The Commerce Manager "Account Health" feature (seller-compliance grading) is officially "no longer available" as of July 2026. Don't conflate it with ACE — there are at least three distinct systems, and mixing them up is the most common error in practitioner content.
Treat these as industry speculation. Each was killed by an adversarial 3-judge panel.
"October 2025: Meta increased Customer Feedback Score weight in the auction" — the scoreify.net narrative, including the quoted Meta CSM statement, could not be corroborated anywhere.
REFUTED 0–3All specific CPM penalty mappings — "score 3 = 20–30% higher CPMs", "2.3 score = 50%+ CPM gap", "minimum 10% cost increase in the 1–2 band". Every concrete number circulating in practitioner content is unsupported or fabricated.
REFUTED 0–3"Feedback score UI removed since late 2024 but still secretly enforced" — failed verification; the feedback score remains advertiser-visible in Business Support Home.
REFUTED 0–3The two "Meta survey-to-auction patents" — US 12,321,983 and US 10,223,742 are assigned to Google LLC, not Meta (verified at USPTO). Anyone citing them as Meta ACE evidence is mis-attributing. They prove only that a rival platform patented survey-adjusted auction ranking.
MIS-ATTRIBUTION · VERIFIED 3–0"ACE = Asset Compliance Escalation" — one vendor's definition conflicts with the Advertiser Customer Experience program known from rep conversations; likely vendor-coined or a name collision.
UNSUPPORTEDThe public record cannot answer these. Structure them into the written change-log summary being prepared for Meta.
Or are they survey-sampling / account-management segmentation? Meta's documented account-level adjustment right (F·4) makes gating architecturally possible. The external ACE tool launch may expose the answer.
The post-click survey officially "shapes the ad auction." If the ACE percentile is computed from the same instrument, "doesn't affect performance" was a semantic dodge. This single answer resolves the core contradiction.
"Unexpected charges" is a canonical negative-feedback dimension — structurally salient for Club/Autoship, and the likely thread connecting the TrueKind page flag ("Unexpected charges after payment") to the score despite TK's setup matching Shapermint.
After the 2025–2026 help-page rewrite: is the banded model (1–2 penalty, <1 restricted) still enforced, or superseded by a continuous auction-integrated signal?
Meta's docs confirm the survey samples likely-purchasers about post-purchase experience — subscription friction is the highest-leverage survey dimension. UK's removal of Club + best-in-group improvement is exactly the causal evidence Meta can validate internally.
Purchase-event sampling, not spend, drives survey volume — no spend level fixes it. The group-level feedback ("most advertisers below our size will never have usable data") is well-aimed for the external tool feedback round.
The ~46th-percentile refund score sits exactly on a canonical survey dimension — the instant-refund and un-gated-return changes in the CX program are pointed at the right lever.
Secondary sources were used only where they corroborate Meta's current or historical wording. scoreify.net claims failed verification outright. No practitioner chatter (Reddit/X/LinkedIn) on ACE buckets survived verification.