The following is from Bott and Tu. If $f : M \to N$ is a map between two manifolds and $E$ is a complex bundle over $N$, then the pullback $f^{-1}E$ is a bundle over $M$. If the Chern classes of $E$ vanish, then the naturality property implies that so do those of $f^{-1}E$. Taking the Chern classes to be a measure of twisting of a bundle, we may assert that pulling back dilutes a bundle, i.e. makes it less twisted. One extreme is when $f$ is constant; in this case $f^{-1}E$ is trivial.
One may wonder if there exists a bundle which is so twisted that every bundle is a pullback of this universal bundle. Such a bundle indeed exists, at least for manifolds of finite type; it is the universal quotient bundle on the Grassmannian $G_k(\Bbb C^n)$ for $n$ sufficiently large.
I have some naive questions about this phrase. First, is there some simple example that would illustrate this "diluting" via $f$? If I consider for example $f:S^1 \to S^1$ and give the codomain the Möbius bundle, can I choose $f$ such that the pullback bundle $f^{-1}E$ untwists the Möbius bundle to $S^1 \times \Bbb R$ or $S^1 \times \Bbb C$?
Next, can I use this "twistness" to classify bundles? I.e. are bundles with the same twistness isomorphic? I guess that twistness means at least in the complex setting that the Chern classes are equal.