Naming drives rating. Occasion drives revenue.

The two games

Our earlier volumes established, three separate times, that a guest naming a staff member is the single strongest correlate of positive sentiment. It is a genuine finding and it replicated cleanly.

It is also nearly useless for predicting revenue. Comparing the highest-revenue venues against the rest, naming an employee separates them by a ratio of 1.20x. Essentially noise.

Here is what does separate them:

• Reservation, booked, reserved: 3.40x. The highest-ratio signal in the entire dataset.
• Celebration language: 2.81x. Birthday, anniversary, graduation.
• Group or party of N: 2.25x.
• Names an employee: 1.20x. The strongest sentiment signal, and almost flat on revenue.

The mechanism is not mysterious. Pre-commitment converts a discretionary visit into a scheduled obligation. A guest who has booked has already decided, usually brought other people, and will almost certainly arrive. A guest who is merely delighted may come back on some unspecified Thursday.

The operational consequence is uncomfortable for anyone who has invested heavily in floor training. A working online group-booking flow is plausibly worth more in revenue terms than a year of service coaching, not because service does not matter but because it is being measured against the wrong outcome.

Did negative reviews cost these venues money?

Indistinguishable from zero. We expected this to break under scrutiny and it did not: the result holds across both models independently and across the consensus label set.

The aggregate complaint rate carries no revenue signal. Broken out by aspect it is more interesting. Food complaints do track revenue, at r = -0.489 on consensus labels, and the number of aspects a review evaluates at all tracks it at -0.418. Staff attitude complaints run mildly positive against revenue at +0.262, which is the sort of result that should make you slow down rather than build a strategy on it.

What protects revenue is exposure, not perfection

Scoring a venue as having a food anchor when at least 30 percent of its reviews affirmatively praise the food, and an atmosphere anchor at 22 percent:

• 25 of 40 venues have a food anchor.
• 11 have an atmosphere anchor.
• 4 have neither.

Atmosphere turns out to be a floor rather than an anchor. It is the least complained-about aspect in the whole corpus, negative in only 6.4 percent of its 3,803 labels. Almost nobody fails at atmosphere, which is exactly why it does not differentiate anyone.

Is there usually a single problem to fix?

Counting negative mentions deterministically, the leading issue across the cohort breaks down as staff attitude at 12 venues, service attentiveness at 11, food quality at 10, value for money at 6 and service speed at 2.

But the more useful number is this: only 9 of 41 venues have a clear single leading issue. In the other 32 the top two categories sit within 25 percent of each other. Operators looking for the one thing to fix are, most of the time, looking for something that does not exist.

Did responding to reviews move the rating?

Whether management responds to reviews was uncorrelated with rating in two independent samples, at r = -0.10 and r = -0.008. We are not arguing against responding. We are reporting that in this cohort it did not move the number.

The limitation you should hold onto

Every venue here is a top-10 alcohol earner in its city. All 41 are already winning. Every correlation on this page is therefore a comparison among winners, it is descriptive rather than causal, and it does not transfer to venues outside that group.

That matters most for the negative-review finding. Saying complaints carry no revenue cost is true within this cohort. It says nothing whatsoever about a struggling venue with a 29 percent negative rate, because venues like that are absent from the sample precisely because they failed. Do not read this page as permission to ignore complaints.

How every number here was produced, including what we got wrong first.