Why processes involving virtual loops (so-called indirect searches) allow to probe higher energy scale Beyond Standard Physics than those processes (so-called direct) that are typically of tree order ?
See for example page 3 of https://indico.in2p3.fr/event/13763/contributions/15263/attachments/12639/15522/3_AndreasCrivellin.pdf
Direct searches refer to the discovery of new particles via processes where they are produced. For instance, the Large Hadron Collider at CERN is currently looking for new heavy particles that can be created during the collision of high energy protons (the center of mass energy is around 13 TeV). Take one event of two protons with such center of mass energy colliding, the colliding partons will carry only a fraction of this energy, call it $\Lambda$. During the parton collision one can only produce a new particle whose mass is lighter $\sim \Lambda$ (in natural units). In the picture below you can see recent bounds obtained by the ATLAS collaboration on a new heavy charged boson (like a heavy W boson). 
.
In indirect searches, and in particular indirect searches in low energy process at the precision frontier, we can indeed be sensitive to new particles which are much heavier than those we can produce directly. This is due to a combination of factors: i) These processes can be measured with an extremely good accuracy in high intensity experiments. ii) The Standard Model (SM) contribution to these processes is very suppressed because of the flavour structure of the SM (GIM mechanism for example) and the processes also only receive loop level contributions in the SM so they also have loop suppression factors. These together, allow us to be sensitive to very small new physics contributions coming from very heavy particles, even if they also enter at the loop level.
