Consider the following pair of surfaces:
$P\;$ is the plane with the origin removed.
$C$ is the cone $z = \sqrt{x^2+y^2}$ in $\mathbb{R}^3$, with the origin removed.
There are several possible structures we could put on these surfaces. For example, both of the surfaces support a differentiable structure. As differentiable manifolds, $P\;$ and $C$ are diffeomorphic. Thus the differentiable structure cannot "see" the conical singularity at the tip of the cone.
Each of these spaces also has a Riemannian metric. Both Riemannian metrics are flat (i.e. Euclidean), but the two surfaces are not isometric. In particular, a loop on $C$ that surrounds the origin will have a holonomy of $(2-\sqrt{2})\pi$ radians, while a loop on $P\;$ that surrounds the puncture has zero holonomy. Thus the metric can "see" the conical singularity.
You do not need the full metric structure to detect the conical singularity. For example, if we give each surface a Euclidean similarity structure, then it is still possible to define the holonomy of a closed curve, and we can detect the difference between $P\;$ and $C$.
However, it turns out that $P\;$ and $C$ are equivalent as conformal surfaces. For example, the map $P\to C$ that maps the point in the plane with polar coordinates $(r,\theta)$ to the point on the cone with cylindrical coordinates $\bigl(r^{1/\sqrt{2}},\theta,r^{1/\sqrt{2}}\bigr)$ is a conformal equivalence. Thus the conformal structures cannot "see" the conical singularity.