ETTR is based on the following principle: a sensor is linear, out eyes are not. That means that the sensor data have to be ‘gamma corrected’ to give the same effect as our eyes do. Because a sensor is linear however, half the number of shades will be in the first (brightest) stop, 25% will be in the second stop, etcetera. If you shoot a relatively dark scene, you may lose 50% or more of what the sensor could record, because you record nothing in the first stop. At the same time you will be struggling to get enough meaningful information in the darkest stops (read: to get a signal that is higher than the noise).
That is when ETTR makes sense. You overexpose the image, in order to push all the information up to the right of the histogram. The result is an increased signal to noise ratio everywhere in the image. In post processing you correct for this overexposure, of course. The result is a cleaner image, which is most obvious in the darkest parts. Unfortunately, this is not as easy as it sounds. You may blow highlights if you are not careful and do not understand when to use this and when not to use this. Many people think that ETTR simply means you overexpose every image and correct that in post, but that is nonsense. You only overexpose those scenes that have headroom to do this.
That is when ETTR makes sense. You overexpose the image, in order to push all the information up to the right of the histogram. The result is an increased signal to noise ratio everywhere in the image. In post processing you correct for this overexposure, of course. The result is a cleaner image, which is most obvious in the darkest parts. Unfortunately, this is not as easy as it sounds. You may blow highlights if you are not careful and do not understand when to use this and when not to use this. Many people think that ETTR simply means you overexpose every image and correct that in post, but that is nonsense. You only overexpose those scenes that have headroom to do this.