Then, to tune precisely to the engine, you datalog the oxygen sensor or fuel ratio sensor in closed-loop and process that data to refine the fuel map.
I'm harking back to Megasquirt 1 here so things might be a little different now.
The MegaSquirt measures as its most basic inputs Manifold Absolute Pressure (MAP) and engine speed (RPM) and uses a Volumetric Efficiency (VE) [1] in a lookup table to calculate the fuel injection pulse width in order to combust at a stoichiometric air:fuel ratio (14.7g air:1g fuel) or maybe a bit rich (<14.7:1). A VE <100% is implied by all engines under vacuum because their displacement is constant while the atmosphere in the manifold is at a reduced density. Also good add an inexpensive intake air temperature sensor to adjust air density.
An oxygen sensor in the exhaust stream will "switch" between a lean signal —approximately 0.1 volts—and rich signal—approximately 0.9 volts. This acts as feedback to the engine management to add a correction to the theoretical VE.
More advanced is a "wide band O2" or "lambda" sensor that measures the air/fuel ratio over, as the name implies, a range of air/fuel ratios rather than just a nearly binary output at stoichiometry.
Many engine management systems use a Mass Airflow Sensor (MAF) to directly measure the mass of air taken into to the engine through a single point in the intake air piping and don't need to know much about the details of the VE of the engine.