Often, we look at the survivors and make our decisions, but we forget about the casualties.
The famous example is that the military put extra armor where planes were being shot, but they were only thinking of the returning planes (survivors as the dataset), not those that were shot down. They should have put additional armor on everywhere except where the bullet holes existed on returning planes: those areas signified lethal hits, as no aircraft returned.
This “in your face” or “availability” bias is the same as when seatbelts or advanced military helmets are introduced: injuries go up, causing us to correlate the new technology with more injuries. However, this limited view does not consider that deaths decrease: an unseen factor.
We are left confronted with the pain of the survivors, while neglecting the more significant implication: they are still alive.