Measuring frequency digitally is an interplay between hardware, the microphone, and software, the sound processing engine. Microphones are not all alike and you can get different readings running the same software with different mics. Not a big difference, generally, but enough take a bit of an edge off the illusion of exactness.
On the software side the program needs to isolate frequencies and then process them individually. Lots of phone aps have problems with this and jump from note to note. There is software out there designed for composers that has very good frequency isolation. Once the software isolates the frequency, that is tracks each individual note and overtone, the next step is to generate a number of data points for each tone. By data points I mean individual readings over time, sometime hundreds of readings per second. When you examine a data dump on a bowl what you don’t see is the same frequency over and over. Instead the readings spread over a range. This range might be narrow, under 1 Hz or wide 5 Hz and even more. The software will take an average or a most frequent occurrence (mode) and display that as the bowl’s frequency.
Best Singing Bowls has employed software engineers to develop a program that uses multiple sound processing engines to generate frequency readings. We also use third party software to analyze sound clips and generate a frequency report. We get pretty good agreement among the various programs, but not perfect, again adding nuance to the concept of exactness.