“Out of 80 experiments, 45 (56.3%) had a favourable outcome.” If you read this sentence in a manuscript, would you want to edit the figures?
I certainly would. There are too many digits in ‘56.3%’. The decimal 3 is meaningless; 56% is precise enough. If the number of favourable outcomes is 44, the percentage score is 55%; with 46 successes it is 58%. There is no uncertainty here.

But what should we do when we are dealing with 237 out of 623? Both 237 and 238 result in a score of 38%. Wouldn’t it be wise to distinguish these outcomes by writing 38.0% and 38.2% respectively? Well, if such precision is important, we can simply present the absolute values. Absolute values are always accurate; percentages and fractions are only approximations.

What might be the purpose of accurate percentages? I appreciate that percentage scores and fractions are better for comparisons than absolute values. With percentages I can see at a glance that 237/623 is more than 165/465 (38% and 35% respectively). Percentages are quick – and inaccurate, even with additional decimals.