It isn’t always possible for manufacturers to make certain types of capacitors with all values with the lowest possible tolerance on the expected capacitance. Let’s say we choose an electrolytic capacitor with a 200uF, 20% tolerance with any voltage as a random example 1189-2375-ND; manufacturer part number 8AX200MEFC5X11. If the manufacturer truly attempts to make this capacitor each time guaranteed to meet a 20% tolerance, you could expect the following (keep in mind this is just a predictive model based on the assumption that the product is evenly distributed).
If you only buy a few, let’s say 2, you have a higher probability of getting values much further away from the original value. I made a quick spreadsheet of what you’d expect and ran some sample numbers. On average, if you only buy 2, you can expect around 2.9% tolerance (I attached a zip file of the spreadsheet to show my work). I basically set a random value between the minimum you can expect and the maximum you can expect. Then I took averages at the quantities purchased and let the numbers change ten times to get an overall average. An average can tell a lot, but the standard deviation can tell you what is usually “expected” based on the amount you purchase. In this case, for purchasing 2, the standard deviation is around 8.7% from the expected average of 2.9%.
Why does this sample analysis matter? It’s simply to show why capacitance tolerance (and any other tolerances) make a difference in some applications. For example, in order to even potentially obtain closer values to 200uF given a 20% tolerance, you’d have to effectively buy at least 10 to get around 3.2% (standard deviation) and 50 units to get even better chances at 1.1% tolerance based on quantity alone. This is not effective, as there is no guarantee when your application only requires a small amount, so a tighter tolerance (smaller percent) is recommended if you want the value closer to the expected rating. You may even have to consider different types of capacitors altogether that have better tolerance ratings since other types may be designed to be more accurate. We can’t control what values are received, this is another reason tolerance should be considered, you could end up getting several capacitors closer to the 20% tolerance rating.
The download below is for the Excel sheet, I also included a 2000uF example to see if there was a difference in values one may receive. I only considered whole number values in the data.
It’s impossible to know exactly what process manufacturers use when constructing these, it’s very possible that they try to avoid values on the “extremes” of the tolerance rating. I would assume that the extremes (full 20% tolerance either way) are not too common based on the statistics, however, nothing is guaranteed here. Don’t expect tight tolerance on parts rated with a high percentage.