This is a typical thread, where the original poster asks for help but isn't willed to give us all details about the setup he intends to use. Ok, you don't want to get fired by your boss, but without any numbers this discussion is nearly senseless. So, yet again, my reply starts with the phrase: "Assume you want to use a ... "

Assume you want to use a high quality, temperature compensated ratiometric linear Hall effect sensor, which is similar to this one

http://web.mit.edu/6.302/www/SS490.pdf

Then you will notice that there's a certain temperature caused Null drift, in this example of about +-0.05%/°C. Further assume, that the supply voltage is 5V, both Null potentials are 2.5000V initially and a change of ambient temperature of +-5°C is to be expected. Then, worst case situation will mean a Null potential of 2.4875V for the one Hall sensor and 2.5125V for the other, yielding an offset of 0.025V, which is more than 1LSB of an 8-bit ADC!

So, if you really want to be as precise as a 14bit ADC (not even mentioning 24bit precision!), then you need to take into account the temperature caused Null offset drift between the both Hall sensors. This means that you additionally have to measure the ambient temperature near the Hall sensors and to correct the Null offset drift by the help of a Null offset drift calibration curve.

If you additionally expect relevant common mode signals, then you also have to correct the temperature caused sensitivity drifts, which will become the real challenge then...

Kai