Several Yelp questions ask for the “average rating of businesses,” but their golden answers pool all individual reviews. This weights businesses according to their number of reviews.
A natural alternative is:
- Calculate the mean review rating for each business.
- Average those business-level ratings, giving every business equal weight.
The two calculations differ in all three affected questions:
| Query |
Pooled-review mean |
Equal-business mean |
Golden |
| Indianapolis businesses |
3.547008547 |
3.497493225 |
3.547008547 |
| Businesses in top-review state (PA) |
3.699395770 |
3.648005824 |
3.699395770 |
| WiFi businesses in PA |
3.483985765 |
3.582654975 |
3.48 |
The third pooled value matches the golden after rounding.
Affected questions include:
What is the average rating of all businesses located in Indianapolis, Indiana?
Which U.S. state has the highest number of reviews, and what is the average rating of businesses in that state?
Which U.S. state has the highest number of businesses that offer WiFi, and what is the average rating for those businesses?
Could the intended weighting be clarified?
If pooling reviews is intended, wording such as “average rating across all reviews for those businesses” would make the expected calculation explicit. If businesses should be weighted equally, the golden answers should instead use the business-level means.
Without clarification, semantically reasonable solutions can produce different answers.
Several Yelp questions ask for the “average rating of businesses,” but their golden answers pool all individual reviews. This weights businesses according to their number of reviews.
A natural alternative is:
The two calculations differ in all three affected questions:
The third pooled value matches the golden after rounding.
Affected questions include:
Could the intended weighting be clarified?
If pooling reviews is intended, wording such as “average rating across all reviews for those businesses” would make the expected calculation explicit. If businesses should be weighted equally, the golden answers should instead use the business-level means.
Without clarification, semantically reasonable solutions can produce different answers.