Direct method: decay distance between the production and decay points. But the lifetime is very short (10-16s) and so high spatial resolution is required, and only thanks to relativistic time dilation one is able to see it.
But since the pions cannot be produced monochromatically, it's also necessary a very good knowledge of the pion spectrum.
First experiment: CERN PS, 1963. It used targets (for the proton beam) of different thickness.
Second experiment: CERN SPS, 1985. It impinged the protons upon two tungsten foils whose separation was variable. The first foil served for pion production, the second for photon conversion into e+e-.
In both experiments, positrons were counted.
For a small foil separation, some of the pions decayed too late, while for a large separation essentially all pions decayed before the second foil. Thus, the positron rates where counted for different foil spacings.
The dominant systematic was the uncertainty on the π0 spectrum, which was not measured but was assumed to be the arithmetic mean of π+ and π- spectra.
Corrections had to be made for the Dalitz decay of the pion (e+e-γ, so providing extra positrons), conversions of the photons in the production foil, prompt positron and photon production, e+e- from η decay.
Method with γγ collisions:
At DESY, it was e+e- -> e+e-γ*γ* -> e+e-π0 -> e+e-γγ
The pion is identified from the γγ invariant mass. The cross section is proportional to Γπ0->γγ (squared?).
Method with Primakoff effect: other page.