RatingMinder/Numbers
Every number we've released, with its count and its date. Wins and losses both go here as they come in.
Reviewed by Doug Baker, founder · Updated September 23, 2026
Here's the short version. We have numbers from a pilot and no removal outcomes yet. Of the reviews our analysis flagged as having grounds for removal, 93% held up against Google's written policies. We measured that on a set of 164 pilot reviews, so it's how often our flags held up on that set. It isn't how often Google removes reviews. We don't have that number yet. As outcomes come in, they go in the table below, losses included.
Four one-star reviews on one profile. One has grounds for removal. The other three don't, and they stay up. The count is what this page keeps.
Two one-star reviews. One has grounds for removal. The other doesn't, and it stays up. The count is what this page keeps.
Outcomes
Has a review come down after an appeal we prepared?
Not yet. We have no removal outcomes so far, and as of September 24, 2026, the table is empty. When owners file the appeals we prepare, each one lands here under the grounds we argued.
| What we argued | Filed | Removed | Lost | Pending | Updated |
|---|---|---|---|---|---|
| Off-topic | 0 | 0 | 0 | 0 | — |
| Rating Manipulation (conflict of interest) | 0 | 0 | 0 | 0 | — |
| Fake engagement | 0 | 0 | 0 | 0 | — |
| Any other policy | 0 | 0 | 0 | 0 | — |
We record a removal only when we see the review gone from the profile, and an appeal counts as a win only once the review has stayed gone through at least 30 days of watching. Google can put a review back, so we watch every removal for 30 days. A loss counts when Google's decision email says no and you confirm it on your appeal page. It also counts when the appeal is in, 30 days have passed since the report, and the review is still up. Pending is everything in between.
If Google puts a removed review back during the 30-day watch, we count it in a line under the table. Filing again after that is a new report on the same review, not a second appeal.
A loss here is also where the refund comes in: the appeal fee only, never subscriptions. How the refund works.
We won't call any of this proven until 50 filings are in. Below that, a run of wins or losses tells you more about which reviews happened to come in than about the method. We picked 50 before there was a single result, so the line can't move to suit one.
The pilot, August 4, 2026
What did the pilot find?
Every number we've released comes from one pilot. On August 4, 2026 we pulled 1,637 negative Google reviews with text, one and two stars, from 120 businesses in 10 trades across 10 US metros. Then we analyzed every one of them against Google's policies. It flagged 86 of them, about one in 20.
| Measure | Figure | Counted from |
|---|---|---|
| Reviews analyzed | 1,637 | Negative (1 and 2 star) reviews with text, from 120 businesses, 10 trades, 10 US metros |
| Reviews flagged with grounds | 5.3% | 86 of the 1,637: 18 tier 1, 68 tier 2 |
| Businesses with at least one flagged review | 41% | 48 of 117 |
| Flags that were off-topic | 65 of 86 | 30 of the 65 were road or neighbor encounters |
Pulled August 4, 2026.
The other 94.7% weren't flagged. 65 of the 86 flags were off-topic, and 30 of those described a run-in on the road or with a neighbor rather than time spent at the business. The analysis never flagged personal information, hate speech, impersonation or extortion anywhere in the pilot. That tells you what we flagged. It doesn't prove none of those reviews were out there.
Tier 1 and tier 2 are the two levels of grounds our method uses. What each tier means.
What does the 41% mean?
It's the figure people misread most, so here it is in full. 41% is a share of businesses, not reviews. Among map-pack businesses, the few Google shows next to the map in a local search, with at least three negative ratings and at least 20 reviews, 48 of 117 had at least one flagged review. It is not 41% of reviews. That figure is 5.3%. And it is not a removal rate. It means the analysis found something worth arguing at about four businesses in ten.
Why 117 and not 120? Three of the 120 businesses had no negative review with text, so there was nothing to analyze.
The other side of it:
About six in ten profiles we analyze have no review with grounds for removal. We say so plainly, with a reply playbook attached.
In the pilot, 69 of the 117 businesses had none.
How often are we right?
93%
Of reviews we flag as having grounds, 93% were confirmed by expert adjudication against Google’s written policies (65 of 70 flags, n = 164 adjudicated reviews). Our assessment accuracy, not Google’s removal rate; removal outcomes publish as they accrue.
Here's where the 70 comes from. The analysis flagged 86 of the 1,637 reviews. To grade it, we built a set of 164 reviews with a known right answer. The set has the 114 reviews an earlier version of the analysis flagged, plus a sample of 50 from the 1,523 it cleared. A second reading went through them blind, set up to be skeptical. On the 82 where the two readings disagreed, an expert settled each one by hand against Google's written policies. 70 of the 86 flags fall inside that set, and 65 of them held up.
We built that set around the flags, so it isn't a random slice of the 1,637. The 93% and the recall below are measured on those 164 reviews, not on the whole pilot.
What did we miss?
The same 164 reviews measure the other direction. In that set, 71 reviews have grounds, and we flagged 65 of them. So in that set we caught 91.5% of the reviews that had grounds. That's our recall, and it's a ceiling, not a measurement of all 1,637 reviews. The six we missed would have reached an owner as no grounds. Across the whole set, 71 have grounds and 93 don't.
Those 164 answers never change. Every change to the analysis gets scored against them before it ships, and a change that drops that 93% below 80% doesn't ship. That 80% is a floor on precision, the share of our flags that hold up, measured on the same 164 reviews. The full method, step by step.
Outside numbers
What does anyone else know about removals?
Less than you'd hope. We haven't found anyone who publishes removal success rates by category, Google included. Google does publish two figures worth knowing, and each has a limit.
| Source | Figure | The limit |
|---|---|---|
| Google, Maps 101 post, February 2021 (2020 figures) | 960,000+ reviews removed after user reports | No denominator, so it shows reports sometimes work, not how often |
| Google, EU transparency report, first half of 2025, §3.2.3 | 88% sampled accuracy for automated Maps moderation (95% CI 81–92%) | Google grading its own automation, in the EEA only, on a sample whose size it doesn't publish. It says nothing about how one report will go |
Checked September 23, 2026.
How long does Google take?
Google doesn't publish a timeline for appeal decisions, so we never promise a date.
What stays off this page
What won't we ever publish?
Never published here
- A forecast. Google doesn't decide on odds. It decides on policy, one review at a time, so we never print a number for the chance of removal.
- One owner's result as proof. One appeal isn't a rate. The numbers here are counts, with what they're counted from.
- A rating change multiplied by odds. Your verdict shows what your rating would be if the reviews with grounds come down. That's arithmetic, and it stays arithmetic.
- A rating threshold. We've found no support for a rating threshold, so we don't claim one in any form.
More on how this works
- How the 164 reviews were graded
- When your fee gets refunded
- How we prepare the appeal, and who files it
The only number that matters to you is the one on your own profile. We'll analyze your bad reviews free, and we don't need your email.
