The performance indicators of Improved Cook Stoves (ICSs) for Developing Countries are commonly evaluated and compared using the arithmetic average of replicated tests performed using a standardized laboratory-based test, commonly the Water Boiling Test (WBT). Possibility theory is here employed to examine energy data retrieved from the WBT-based literature regarding the results of laboratory tests on ICSs and traditional Three-Stone Fire (TSF) stoves; fifty-seven comparisons of stoves are analysed. Chebyshev and uniform possibility distributions are employed to represent energy data affected by epistemic uncertainty. The extension principle of fuzzy set theory is applied to obtain possibility distributions of the saving of fuel use parameter for each comparison of cookstoves. The results indicate that at 90%, 95% and 99% degree of confidence, only 22.22%, 15.00% and 15.00% of all the supposed “improved” stoves emerged respectively as real ICSs at most, while the percentage of “improved” stoves obtained by considering the mean values of the WBT is among 3 and 6 times higher than the percentage resulted by taking into account the epistemic uncertainties. The work suggests how neglecting intrinsic uncertainties of tests’ results might lead to misinterpret and report non-comprehensive information about ICSs’ thermal energy performance, and to reveal some concerns about their effective improvements over traditional devices.@en