It is normal to have a small discrepancy between the audio count and the final count. Usually, it's less than 3, but in the case of small numbers, it can be off by a little more as it has trouble finding the pattern with fewer data. The more reps you do, the better it should get at detecting the number.

Did this answer your question?