The first step might be to realize that most of the observed particles have spin-0, spin-1/2 or spin-1. Knowing this we would like to write down Lagrangians for these particles. Renormalizability is of great help here, since it essentially brings down the set of allowed theories to a small number.
For purely scalar field theory, you can only have $\phi^4$ interactions. If you now add spinors, the only allowed interaction is the Yukawa coupling. Finally, if you add vector bosons, the allowed interactions is that of QED or scalar QED. Another kind of interaction allowed by renormalizability of gauge fields includes interactions of the type $A^2(\partial_\mu A^\mu)$ or $A^4$. Gauge symmetry restricts the coupling constants of these interactions a lot, and the only allowed theory is Yang-Mills. And, we're done! Nothing else is allowed.
So, we have already reduced the potentially infinite number of theories to
- $\phi^4$
- Yukawa ${\bar \psi} \psi \phi$
- QED ${\bar \psi}\gamma^\mu \psi A_\mu$
- Scalar QED $A_\mu \left( \phi \partial^\mu \phi^* - \phi^* \partial^\mu \phi\right)$
- Yang-Mills
One could go further from here, and reduce it a little bit more. This is still an infinite class of theories. For example, the Lie group of Yang-Mills is unknown at this point.