Any statisticians here? What is the likelihood

[raccoon]

that someone will give a positive review of a business for a positive experience versus the likelihood that they will give a negative review for a negative experience?

My guess is they're far more likely to give the negative review.

[greymattermom]

1. One problem is that the expect you to score 5

for an experience that isn't horrible. Not everything is a 5. Maybe they should just ask if you would come back. One of the worst is the knot. Everything is 5. What good are the reviews then?

[MineralMan]

2. When a business tells me that they expect the top score,

I simply don't participate. It's rare when any business does a perfect job of anything. If they really wanted feedback, they wouldn't ask customers to rate them at the highest level only. Screw that.

[FSogol]

3. An old adage in the restaurant business was that if someone liked your restaurant, they'd tell 3

people, but if they had a bad experience at your restaurant, they'd tell 10 people.

That's why social media reviews are mostly crap.

[Wounded Bear]

4. Hyperbole reigns in the internet age...

nobody gives 2s or 3s. It's either 0 or 5 or don't bother.

[JudyM]

5. A social psychologist or behavioral economist would likely give a good idea.

My understanding... it would depend on the person's general world view and how the experience fit in that internal narrative/the gain they'd feel from publicly expressing their feeling about the experience, plus their personality "type" (e.g., introvert/extrovert, MBTI...). If it felt deeply personal/attached to their identity, whether positive or negative, the review would be more likely.

[genxlib]

6. I agree there is a lot of skewing in both directions

I think there is definitely a propensity for negative reviewers to be more likely to respond

And I agree with others here that there is a lot of hyperbole towards top ratings even when things are only OK.

However, I find the ratings incredibly useful. All things being equal, the results still matter.

While the scale might be thrown off by the above factors, the relative difference between comparable items/places is still valuable information. The beauty of crowdscourcing is that, if you have enough responses, the results will start to calibrate and normalize towards a reasonable approximation of ranking.

Perhaps a 5 star restaurant should only be 4 stars but I can still expect it to be better than the 4-star place that should be 3-stars.

This generally comes into play on Amazon (for stuff) and Yelp or Tripadvisor (for places). My rule of thumb is 4.5 should be good and 4.0 is worth trying.

It does not always work but it is far, far better than just guessing. Actual commentary on what is good and bad is also very helpful.

[L. Coyote]

7. We call that a probability (P), not a likelihood.

P = 1.0 if the person isn't a liar.