We investigate to what extent the experimental information on B→ππ branching fractions and CP asymmetries can be used to better understand the QCD dynamics in these decays. For this purpose we decompose the independent isospin amplitudes into factorizable and non-factorizable contributions. The former can be estimated within the framework of QCD factorization for exclusive B decays. The latter vanish in the heavy-quark limit, m_b→∞, and are treated as unknown hadronic parameters. We discuss at some length in which way the non-factorizable contributions are treated in different theoretical and phenomenological frameworks. We point out the potential differences between the phenomenological treatment of power-corrections in the ``BBNS approach'', and the appearance of power-suppressed operators in soft-collinear effective theory (SCET). On that basis we define a handful of different (but generic) scenarios where the non-factorizable part of isospin amplitudes is parametrized in terms of three or four unknowns, which can be constrained by data. We also give some short discussion on the implications of our analysis for B→πK decays. In particular, since non-factorizable QCD effects in B→ππ may be large, we cannot exclude sizeable non-factorizable effects, which violate SU(3)_F flavour symmetry, or even isospin symmetry (via long-distance QED effects). This may help to explain certain puzzles in connection with isospin-violating observables in B→πK decays.