Ebola is a deadly disease, with an estimated overall mortality rate of about 50%. Because the mortality rate is high, there is pressure to give experimental treatments—any treatments—to Ebola sufferers even without proof of efficacy. Interpretation of the resulting hodgepodge of data from ad-hoc allocation of treatments is nearly impossible. A randomized clinical trial is desperately needed to confirm efficacy of new treatments. We describe a flexible design allowing incorporation of promising new agents and facilitating interpretation of results even if the trial stops early because the epidemic wanes, the treatment supply is interrupted, etc. We dub the approach “barely Bayesian” because our non-informative uniform priors for the mortality probabilities in each arm are quickly overwhelmed by actual data. Focusing on posterior probabilities rather than type I error rate makes the conclusions depend only on results observed, not what action would have been taken had other results been observed. We will discuss this and other features of the design.