Review sources
Analysing Facebook recommendations
Facebook has no star ratings — it has "recommends" or "does not recommend" plus text. That is a completely different metric, and the report treats it separately.
What recommendations are
Instead of a numeric rating, the author declares whether they recommend the place. The aggregate is a share of recommendations, not an average. For classifying text that changes nothing; for comparing with other platforms it changes everything, which is why we never set it against any average.
A share, not an average
Eighty percent recommendations is not a 4.0 rating. The report shows both numbers separately and does not try to convert one into the other.
A strong local effect
Recommendations arrive more often from people nearby and from the owner's own circle. It is usually the most biased of the four sources and we describe it that way.
Short and emotional
The average recommendation is shorter than a Google Maps review and more often speaks of impression than of specifics. Classification works, but a card resting on this source alone rarely clears the confidence threshold.
Step by step
How we read recommendations
-
1
You point at the page
A link to the business page. Recommendations have to be enabled — if they are off, it is worth turning them on, because for some industries this is the second source after Google Maps.
-
2
We compute the recommendation share
Instead of a star average you get the percentage of people who recommended. The report shows that number separately and does not try to convert it into a rating.
-
3
We classify the text
The text of the recommendation goes through the same classification as reviews from other sources. Statements are shorter, so a card resting on this source alone rarely clears the confidence threshold.
-
4
We flag sudden jumps
A jump in the recommendation share with no movement in the other sources is flagged in the report. That usually signals a push among acquaintances rather than a change at the venue.
In practice
The most sensitive barometer of local mood
Recommendations arrive more often from people nearby than on other platforms, so they react first to local events — changed hours, roadworks outside, new competition next door. Set against Google Maps they show whether a change concerns regulars or visitors.
local
recommendations come mainly from people nearby. The most biased of the four sources — and the fastest to react to what is happening next door.
Why it works this way
How to read recommendations so they lead somewhere
Recommendations are short and emotional, and the local effect makes them react faster than other sources — which is their value once you know how to read them. The report flags sudden jumps in the recommendation share rather than counting them as improvement, so you can tell a push from friends apart from real change. It is the most sensitive barometer of local mood you have.
Common questions
Do I need access to the Facebook page?
Not for the analysis — recommendations are public. Access is needed only for replying to them.
What if the page has recommendations turned off?
Then there is nothing to pull and the source stays empty. It is worth turning on if you serve customers locally — for some industries it is the second most important place after Google Maps.
Are comments under posts analysed too?
No. A comment under a post is not a review of the venue, and pulling it into the corpus would distort every share.
See also
Be first in when we launch
Join the waitlist. We'll write once, on launch day, with a link to create your account.