If each of those (on average) 800 matches were only 1% likely to donate if chosen, then the probability that none of them would donate would be very small, ~0.03%. The counterfactual expected impact would be much lower than just 1⁄800; it would be scaled down by a factor of 0.0003.
However, if each were only 0.1% likely to donate if chosen, then you’re close to a 50% probability of saving an extra life (or helping an extra person) than would have been saved (helped) otherwise, which would actually be very cost-effective.
If each of those (on average) 800 matches were only 1% likely to donate if chosen, then the probability that none of them would donate would be very small, ~0.03%. The counterfactual expected impact would be much lower than just 1⁄800; it would be scaled down by a factor of 0.0003.
However, if each were only 0.1% likely to donate if chosen, then you’re close to a 50% probability of saving an extra life (or helping an extra person) than would have been saved (helped) otherwise, which would actually be very cost-effective.