A lot of charitable watchdogs, such as Charity Navigator, divide the ways charities spend money into “program expenses” and “overhead.” Program expenses are money that the charity spends on its actual program: it’s the money that goes to buy malaria nets, stock the shelves with canned food, or pay the veterinarians who help the cute puppies with rare diseases. Overhead includes administrative and fundraising expenses. Administrative expenses are those associated with management and general operations.

It makes sense that people would care about overhead. In general, scam charities tend to spend very little on program expenses and a lot on overhead. If a charity claims to help cure rare diseases in cute puppies, and they’ve spend ten dollars on antibiotics for puppies, ten million dollars on fundraising, and twenty million dollars on the CEO’s salary and team-building trips to Tahiti, this is probably not a real charity.

However, concentrating too much on overhead can actually lead to charities becoming less efficient. Certain kinds of charities spend more on administration and fundraising than other kinds of charities do. For example, Charity Navigator notes, a food bank that takes donated canned goods will not spend very much on administration at all. Conversely, a charity that gives people cash might spend more money on administration, because they have to do accounting to keep track of all the cash. But it’s totally possible that the latter charity does better at helping poor people eat.

If charities are focusing on getting their overhead expenses as low as possible, it can lead to the charity actually being less efficient. For example, the office staff at a domestic violence shelter might use computers from 2008 because replacing the computers would count as overhead. Or they might underpay their managers, which means the managers burn out, quit, and take a bunch of institutional knowledge with them. Or they might avoid hiring an administrative assistant, which means that social workers spend time filling out forms instead of helping people.

Imagine that you were trying to buy a pair of shoes. You might look at how expensive the shoes are, or how well-made they are, or how good the conditions in the factory were for the employees, or whether they are fashionable; these are all reasonable things to take into account when you’re buying shoes. What you would not do is say “wow, I’m going to buy these shoes, the CEO only makes $13,000 a year and all the HR was done by unpaid interns and the office staff are all using out-of-date computers.” That is just totally uncorrelated with whether the shoes are good. Maybe it means the shoes are worse, because HR is actually kind of important in making a good pair of shoes, and you are unlikely to get good HR from a bunch of unpaid interns.

The same thing is true when you think about how to donate to charity. You should donate to a charity that, as best as you can tell, improves the world as much as possible, whatever that phrase means to you– just like you should buy the pair of shoes that fits the best. “Overhead ratio” is a good way of filtering out outright scams, but it is not a good way of separating the okay charities from the great charities. For that, you need to look at outcomes.