Diminishing returns (or diminishing returns to scale) is the decrease in marginal output of a production process as one factor of production is increased while the remaining factors are held constant.
Giving money to an altruistic project (for example, a charity) increases the amount of useful work that can be done by that project—this is a social return on the money donated. If an area exhibits diminishing returns, then the social return on each additional dollar will continuously decline as more and more money is given to the project.
The monetary difference between a project receiving nothing and it receiving $500,000 is obviously the same as the monetary difference between it receiving $4.5m and it receiving $5m. However, the theory of diminishing marginal returns suggests that in terms of the amount of good done, the difference between a project receiving nothing and receiving $500,000 is probably bigger than the difference between it receiving $4.5m and it receiving $5m. So if there are diminishing marginal returns, a project will get more good done with their first few dollars than with their last ones.
This point does not only apply to charities and money: it can also apply to other kinds of funding opportunities (e.g., individuals, universities, scientific research programmes, think tanks, nonprofit organisations) and other kinds of resources (e.g., time and labour). However, it is unclear to what extent diminishing returns apply in different areas. Owen Cotton-Barratt gives an overview of the variety of cases it seems to apply to. However, Michael Dickens, for instance, argues that there are no detectable diminishing marginal returns in animal advocacy or work on existential risk.
If charities face diminishing marginal returns, it may make sense to only fund them to a certain level: beyond that we say they have no additional room for more funding. Given that organizations and areas may face different returns to additional labour than they do to additional capital, it may be worth considering whether they are more talent- or funding- constrained (though there are also ways that framing could be misleading).
Dalton, Max & Owen Cotton-Barratt (2017) Selecting the appropriate model for diminishing returns, Centre for Effective Altruism, May 23.
Sets out some considerations for choosing between different models of diminishing returns.