What is the most precise physical measurement ever performed? Obviously some things, such as the speed of light in a vacuum, are defined to be a precise value.  The kilogram was recently defined to have a specific value by fixing Plank's constant to $6.62607015\cdot 10^{−34}\frac{m^2 kg}{s}$.
In particular, in the case of the latter, we held off on defining this value until the two competing approaches for measuring the kilogram agreed with each other within the error bounds of their respective measurements.
Which leads me to wonder, what is the most precisely measured (not defined) value that the scientific community has measured.  I am thinking in terms of relative error (uncertainty / value).  Before we defined it, Plank's constant was measured to a relative error of $10^-9$.  Have we measured anything with a lower relative error?
 A: A good candidate is the measurement of the fine structure constant $\alpha$.  This wiki article on precision tests of QED states that:

The agreement found this way is to within ten parts in a billion ($10^{−8}$), based on the comparison of the electron anomalous magnetic dipole moment and the Rydberg constant from atom recoil measurements...

There is also an upper bound on the mass of the photon, which is in the range of $10^{-27}eV/c^2$ although that's an upper bound rather than a measurement since the expected value is $0$.
A: To answer the question from a different angle, LIGO has measured gravitational waves several times over the last few years. To do so they have to observe a distortion in a 4km arm, in the order of $10^{-18}$m. In other words, it has to detect a fractional imprecision in its length of ~$2.5 * 10^{-23}$. That's pretty accurate.
A: Another quantity that is well known is the frequency of the 21 cm line of Hydrogen hyperfine splitting that has been measured to about 13 significant figures, 1420405751.7667±0.0009 Hz. Unlike the electron magnetic moment, mentioned in another answer, the QED calculation for hyperfine splitting is not as accurate. This calculation also depends on other fundamental constants, including electron/proton mass ratio and proton magnetic moment. which are not known to equivalent precision.
A: The accuracy of the latest atomic clocks is a good candidate for the prize, with an accuracy of less than 1 part in 10^20. One example has been used to confirm gravitational time dilation over a vertical height of a single millimeter, with even more precise confirmation possible with the latest atomic clocks.
An atomic clock measured how general relativity warps time across a millimeter
A: Carrying the fame of being one of the most precisely verified propositions in physics, the ratio of the gravitational to inertial mass was verified to be unity within $1$ in $10^{15}$ by the MICROSCOPE satellite in $2017$. The earlier best precision was $5\times10^{-14}$, obtained by Baessler, et al. in $1999$.
References: 


*

*https://en.wikipedia.org/wiki/MICROSCOPE_(satellite) 

*https://en.wikipedia.org/wiki/Equivalence_principle#Tests_of_the_weak_equivalence_principle
A: The magnetic moment of the electron has been measured to a few parts in $10^{13}$. (Source) This provides an exquisite test of quantum electrodynamics, and calculating the relevant Feynman diagrams has been a Herculean effort over decades.
Note that the more precise tests cited in other answers are basically null results: no difference between gravitational and inertial mass; no difference in magnitude of charge between proton and electron; no mass of photon. So I believe the magnetic moment of the electron is the most precise measurement that is non-null and thus “interesting”.
A: A few more candidates:


*

*The quantized Hall resistance is a great and surprising example, since it's an emergent property of rather complicated, "dirty" systems. As stated here (2013) one can measure the resistance to one part in $3 \times 10^{10}$. For this reason this effect is now used to define the Ohm.

*Equivalence principle tests using torsion balances reached precisions of about one part in $10^{11}$ in $1964$, see here. Another existing answer gives a more precise result from a more modern experiment. These can be thought of as either verifying the equality of gravitational and inertial mass, or placing bounds on the strength of long-range fifth forces.

*The electrical neutrality of bulk matter follows because the electron and proton have exactly opposite charges. Treating the charge of the electron as given, tests of neutrality effectively measure the charge of the proton, with one experiment achieving a sensitivity of one part in $10^{21}$ in $1973$. Arguments from cosmology can be used to set much larger bounds, though they require more assumptions. 

A: Aren't clocks the most precise devices, relatively speaking? Their uncertainty is 1 second in 33.7 billion years, or better than 10^-18.
Refs:
https://link.aps.org/doi/10.1103/PhysRevLett.123.033201
https://en.wikipedia.org/wiki/Quantum_clock#Accuracy
