Two different problems, constantly conflated
Being findable and being chosen are separate. Ranking is about the first; the star average is almost entirely about the second, and a business can rank first in the local results and still lose to the 4.8 underneath it.
This matters for where you spend effort. A business with a 4.9 and forty reviews has a volume problem. A business with a 3.8 and two hundred reviews has a different problem entirely, and asking more people will make it worse before it makes it better.
How much one bad review moves a small average
The arithmetic nobody does. Starting from a 4.8 average, adding a single one-star.
| Existing reviews | Average before | Average after one 1-star | Drop |
|---|---|---|---|
| 10 | 4.8 | 4.45 | 0.35 |
| 25 | 4.8 | 4.65 | 0.15 |
| 50 | 4.8 | 4.73 | 0.07 |
| 100 | 4.8 | 4.76 | 0.04 |
| 250 | 4.8 | 4.79 | 0.01 |
| 500 | 4.8 | 4.79 | under 0.01 |
Source: Arithmetic, not research. Displayed averages are rounded, so the visible change is sometimes larger or smaller than the underlying one.
This table is the argument for volume, and it is more persuasive than any claim about ranking. At ten reviews a single bad day costs you a third of a star; at two hundred and fifty it costs nothing visible. Volume is the only defence, and it is the one that is permitted.
What the rating actually does to behaviour
Published consumer surveys consistently find that most people read reviews before choosing a local business and that a large majority set a minimum rating below which they will not consider one. Those surveys are self-reported and their exact figures vary by source and by year, so the honest use of them is directional: a threshold effect exists, and it sits somewhere in the low fours for most categories.
What that means practically is that the cost of a rating is not linear. Falling from 4.8 to 4.6 costs very little. Falling from 4.2 to 3.9 can cost a great deal, because it crosses the line where a share of readers stop considering you at all.
It also means the ordering of your problems is decided by where you sit. Above 4.5, volume and recency are the work. Below 4.2, the reviews are telling you something operational and collecting more of them faster will simply document it more thoroughly.
Objections
- Our customers do not read reviews.
- Some genuinely do not — trades that run on referral, businesses with a captive local market. The test is cheap: ask the next twenty new customers how they found you. Most businesses that believe this discover a third of new arrivals looked at the profile first, and a few discover they were right.
- The bad reviews are unfair.
- Frequently true, and it changes nothing about the arithmetic. What it does change is the reply — a calm, specific, non-defensive reply to an unfair review does more for the next reader than the review does against you, and it is the only lever available on a review you cannot remove.
- We would rather have fewer, better reviews.
- This is the gating instinct in polite clothing. The table above is the counter: fewer reviews means every future bad one costs more, so the strategy that feels protective is the one that leaves you most exposed.
- Reviews get removed anyway.
- Some do, through Google's own filtering, and the pattern is unpredictable enough that nobody should build a plan around it. Reviews disappearing is a known and poorly explained phenomenon, and the only response available is to keep collecting.
The three numbers worth holding
- stars lost to one 1-star review at ten reviews
- 0.35
- the same review's cost at 250 reviews
- 0.01
- where the threshold effect appears in most consumer surveys — directional, not precise
- ~4.2
Arithmetic, above
Arithmetic, above
Published consumer review surveys; self-reported and varying by source
The third figure is deliberately hedged. Precise click-through numbers by star rating circulate widely in this industry and most of them trace back to a single survey question being reported as behaviour.
What to do at each rating band
Five bands. The advice is genuinely different in each.
- Above 4.7 — collect volume and recencyNothing is wrong. The work is making the count robust and keeping the newest review recent, because a profile whose latest review is two years old reads as dormant.
- 4.4 to 4.7 — collect, and read the threesThree-star reviews are the most informative ones you will ever get: detailed, unemotional, and specific about a real problem. Almost nobody reads them because they are not alarming.
- 4.0 to 4.4 — fix something firstYou are near the threshold where readers start filtering you out. More reviews still helps, and it helps less than finding the recurring complaint and removing it.
- Below 4.0 — stop and look at the reviews as a datasetSort by theme rather than by date. Below four, the reviews are describing an operational fact, and a collection campaign will document it rather than dilute it.
- Any band, fewer than 30 reviews — volume is everythingAt this count your average is noise. One good week or one bad one moves it visibly, and no strategic conclusion drawn from it is reliable.
How far behind you actually are
Your count against the businesses ranking above you, converted into months of asking.