In terms of bare parameter $\lambda$, the $\phi\phi\to\phi\phi$ scattering amplitude is $\lambda\phi^4$ theory is given by $$\mathcal{M}=-i\lambda+iC\lambda^2\Big[\ln\Big(\frac{\Lambda^2}{s}\Big)+\ln\Big(\frac{\Lambda^2}{t}\Big)+\ln\Big(\frac{\Lambda^2}{u}\Big)\Big]+\mathcal{O}(\lambda^3)\tag{1}$$ where $C$ is a computable numerical constant independent of $\lambda$ and $\Lambda$. In expression $(1)$, the parameter $\lambda$ is bare parameter and the cut-off $\Lambda$ is unknown.
According to Anthony Zee's book of QFT (See Eq.4, page 149 of Quantum Field Theory in a Nutshell) the whole point of renormalization is to measure the physical coupling $\lambda_P$ at a reference value of $s_0,u_0,t_0$ and express the bare parameter $\lambda$ in terms of the physical coupling $\lambda_P$ measured experimentally. To this end, he writes $$-i\lambda_P=-i\lambda+iC\lambda^2\Big[\ln\Big(\frac{\Lambda^2}{s_0}\Big)+\ln\Big(\frac{\Lambda^2}{t_0}\Big)+\ln\Big(\frac{\Lambda^2}{u_0}\Big)\Big]+\mathcal{O}(\lambda^3)\tag{2}.$$
Why does he write $-i\lambda_P$ in the LHS? More concretely, how did he know the LHS is linear in the physical coupling $\lambda_P$?