If no one donated blood, a lot of trauma/hemorrhage victims would die, and the world would be a lot worse off. The average unit of blood donated goes pretty far, in terms of the expected value of the good it does. However, when considering whether or not to donate, we need to evaluate the counterfactual difference our actions make. That is, rather than looking at the average donation, the relevant measurement is of the marginal donation. This brings up the following considerations:
1) As it stands, a substantial number of people already donate regularly, and will continue to do so whether or not the (comparatively tiny) EA community does too.
2) Truly life-or-death situations are a minority of transfusions, and these are pretty much already covered by the existing supply. In fact, hospitals almost always keep an emergency reserve of O- specifically for these cases, so it's very rare that someone directly dies for lack of compatible blood. Because a large number of transfusions/donations happen each day, and blood product can often be transported to different hospitals to meet local shortages, projected supplies are relatively easy to forecast within a given margin of error, so it is possible for hospitals to maintain this emergency supply to handle urgent cases.
Thus, the effect of an additional donation to the existing supply is to help cases where the patient wouldn't be directly saved from death, but a transfusion would improve the quality of their recovery. Nailing down exactly how many QALYs this typically adds is very difficult to track, and probably hasn't been done in a rigorous way. However, there is reason to believe this number is not that high.
In the UK, a unit of red blood cells (RBCs) costs about 120 pounds. While financial incentives don't translate seamlessly into extra donations, this is roughly this price at which more supply can be obtained, so it roughly reflects the medical field's impression of how valuable it would be to do more outreach per unit. Furthermore, the typical cutoff for whether to fund treatment is ~20,000 pounds/QALY, which is much less efficient than the ~130 pounds/QALY one can get by donating to the AMF. (For more detail on these numbers, see this guesstimate and this explanation of it.)
Thus, for blood donation to be anywhere near as effective as the AMF (in terms of paying 120 pounds/unit for more product), the medical field would have to be undervaluing the effectiveness of blood donations by 2 orders of magnitude. Despite the lack of rigorous calculations done in the literature, a collective miscalculation of this magnitude seems implausible given the feedback mechanisms which exist in medicine, not to mention the tacit knowledge hematologists have developed from making these tradeoffs.
The role of effective altruism is to look for, and seize upon, moral opportunities that have been unfairly passed over by society at large. GiveWell-recommended charities, for instance, may sometimes get positive comments from economists, but receive insufficient funding to fully exploit the ethical gold mine that is their cause area. In the case of blood donations, the medical field generally has ways of spotting and filling in the cheap and obvious ways to save more lives, so our time is better spent on causes that aren't being watched over as carefully.
That said, there are occasional cases where emergency supplies dwindle. When this happens, specific appeals are made, and in these cases it probably is effective to lend some helping hemoglobin. Less crucially, regular donors often drop out on holidays and during the winter (due to colds/flu), so if one is inclined to donate, those are the best times to do so.