It's not clear if you're asking for details of how the CMB power spectrum is analysed, or whether it's a general question about how this sort of measurement is made. I'll assume the latter - hopefully this will be of interest to others even if it isn't what you intended.
There is a good discussion of the procedure in the Wikipedia article on the Lambda-CDM model, which you'll often hear described as concordance cosmology.
It's very common in Physics that you can't directly measure a quantity. Instead you construct a mathematical model to describe your experimental results, in which the quantity you're trying to determine is an adjustable parameter. Then you pass your results to a big computer that adjusts the parameter to get the best fit to experiment. The result is your measurement, but it's an indirect measurement. Pretty much everything measured by the LHC is actually a best fit obtained from a mathematical model.
This is how the proportion of baryonic matter etc are determined. A minimal description of the universe requires six parameters:
- physical baryon density
- physical dark matter density
- dark energy density
- scalar spectral index
- curvature fluctuation amplitude
- reionization optical depth
(No, I don't know what the last three of these are either, but as always Wikipedia is your friend.)
Planck didn't measure the CMB in the entire universe, however it did measure the CMB as it looks from Earth in all directions. We assume the universe is homogenous and isotropic, so the results measured from Earth should be basically the same as those measured anywhere else in the universe.
Anyhow, a mathetical model is constructed to describe the CMB using these six parameters, then a computer adjusts the values to get the best fit. The result is calculated values for the baryonic matter fraction etc. As dmckee mentioned, a similar fit can be done for other measurements like the baryon acoustic oscillation, and reassuringly the results are in line with the CMB fit.
The values can and do change as more data becomes available. Before Planck the calculation was done using the WMAP data, and the (much more accurate) Planck measurements revised the values determined from the WMAP data. Future measurements may cause further revision, however this is likely to be just increasing accuracy rather than a wholesale overturning of our current models.