You can't really say that a score is 'good' as it measures attention - which could be good or bad.
Furthermore the average score for journals varies: an article in Science or Nature will typically score much higher than one in a smaller journal, not least because more people have read it and are thus more likely to share it. A good score for one journal might be a low one somewhere else.
Bearing those two things in mind - in general if an article scores 20 or more then it's doing far better than most of its contemporaries.
When you click on the Altmetric donut score - you are taken to the Altmetric report of the article.
Within the Altmetric report you can choose the "Score in Context" tab which shows some different contexts to help you understand if the level of attention is typical compared to similar articles.
For example, you can see how the article's score compares to other articles from the same journal, or from the same journal and published within the same three month period.
Altmetric calculate these percentiles by looking at everything indexed in the Altmetric database, which you can browse using the Explorer. It's important to note that Altmetric don't include articles that didn't get any attention.
So, for example, if the Score tab says that an article is ranked #10 out of 220 published in the same journal it means the 220 articles that have been mentioned at least once on a data source we track, rather than that the journal has published 220 articles in total.