The past year has produced evidence on some of the most successful anti-poverty programs in history. “Cash, livestock, and training” seems to be a simple and scalable way to help thousands and maybe millions out of poverty. This is great news. But while we should celebrate, we should pause before pouring millions into these programs. Paternalism and high price tags could mean that the charities are helping only one person when they could be helping two, three or four.

About two weeks ago, I opened the New York Times to see a full-page ad from a private charity called FXB (at left, click to expand).

“We know how to end global poverty” read the ad:

There’s now a proven approach. The journal Science just published a study proving what works to raise people permanently out of extreme poverty. It worked. And was cost-effective, paying for itself many times over.

The ad was talking about a program that gave extremely poor people livestock, training, a cash stipend for a few months, plus some supervision and advising. Nine scholars tested the program in six different countries, with thousands of people and several charities. (The charity with the Times ad was conspicuously not among them.)

The program did work. After two years, the households who received it were consuming an extra $120 per year — consumption being the most reliable measure of poor people’s income. Three years later, these same households were consuming an extra $80 per year. Since a dollar goes farther in poor countries, this is like getting $250 more purchasing power a year. When each person in your home is earning just a dollar a day, these gains are a big deal.

Given how many aid projects don’t pass a simple cost-benefit test, this “livestock-plus” approach is clearly a huge success. As a result, it’s getting scaled up in countries around the world. Bravo to the charities that did a rigorous evaluation, especially for opening their accounting books to inspection.

Cash-plus programs are showing promise, too. I studied one, in post-war northern Uganda, that gave some of the poorest women in the world $150 cash grants plus some business training and regular supervision and advice. It too paid for itself many times over. After two years, households who got the program were consuming an extra $202 per year (about $500 in local purchasing power).

As big a victory this might be (and it is), we shouldn’t pat ourselves on our backs just yet. The benefits are largely projections, while the large costs are not. Before we scale these programs to millions, we have more work to do.

First, consider the price tag. The cheapest livestock-plus program, in India, cost $413 per household. But the most expensive, in Peru, cost more than $3,000. The average cost of the six was $1,700. The cash-plus program in Uganda I studied was cheaper, but not by much. It cost $843. (Note: all the figures I use are adjusted to the same unit, 2014 U.S. dollars).

These are big sums. Fortunately, the estimated benefits to these programs are also big: about $2,000 for the average livestock-plus program, and more than $4,000 for the Uganda cash-plus one.

But these benefits are mainly based on estimates about what will happen in the future. None of the evaluations had more than three years of data. My cash-plus study had the shortest horizon — we measured impacts just 18 months after the grants.

To believe that the benefits of the cash-plus program exceed costs, you would need to believe that the increases in consumption will go on for at least two years past the final surveys. The same goes for the livestock-plus program in India.

Personally I think it’s likely that benefits go on at least a couple of years. But the other livestock-plus programs were much more expensive, and so the break-even point far further in the future. More than 85 percent of the benefits estimated by these papers (including mine) are in the future.

Indeed, on average, the livestock-plus programs take 18 years or more to break even. If the benefits of getting a goat and training tend to disappear after a decade, then the program doesn’t pass a simple cost-benefit test.

So while it’s reasonable to say these programs pay back more than they cost (indeed, this is some of the best quality evidence on poverty in the world) we shouldn’t forget this success is an extrapolation. In terms of cost-effectiveness, these cash and livestock programs are promising but not proven.

Fortunately, I think there’s a way for these programs to break even faster. The biggest expense across all the programs was staff time. Especially for supervision. Delivering training and cows takes skilled labor, and it’s hard to cut this back. But supervision? Charities feel a lot better when they can personally help their clients along. It’s not just paternalism, but also compassion. The staff care about these men and women and do not want them to fail. So they devote almost unlimited time to helping them. At, unfortunately, great cost.

Undoubtedly that investment of time helps the client. But should it cost 50 or 60 percent of the program? Is it more valuable than the cow or the grant itself? It’s hard to believe.

We tried to test this with cash-plus program in Uganda. Supervising the women cost about $377, about half the cost of the program and 2.5 times as much as the grant itself. After the 18-month impact evaluation, the charity brought the control group into the program. This time, however, the charity randomized the supervision: Some women got the training and grants with full supervision, some with only one or two visits, and some with none at all.

We surveyed the women a month and then a year later, and found that the supervision helped the women maintain the new businesses they started, but there was virtually no effect on consumption. We have no idea whether the supervision helps another year down the road. Maybe, eventually, it pays for itself. But the simple fact is this: taking away the most expensive part of the program had little effect on benefits after a whole year.

Indeed, if we cut staff time, and just gave training and cash to the very poorest, the Uganda cash-plus program could cost $300 a person or less. The low cost livestock-plus program in India is proof it can be done. Even if the benefits dropped by half, this program would still pay for itself in one or two years. It would get far more cost-effective.

Most important, three people could be helped instead of one.

That is the message I want you to walk away with: So long as aid is scarce, compassion and paternalism could mean helping only a million people out of poverty instead of three million. As the givers of all that aid money, I think we have a responsibility to do better.

Last week, the head of one of the world’s largest refugee and crisis response organizations, the International Rescue Committee, called on humanitarian organizations to make their crisis response twice as effective by cutting costs. Every charitable leader ought to publicly echo this call, yet almost none do.

To anti-poverty organizations around the world: Take the risk of putting your entire philosophy to the test. Put supervision and staff time under the microscope. Explore whether you should be in a different business: the business of handing out cash and low-cost services to millions of poor people instead of staff-intensive services to a few cherished thousands. The answer is not clear. But the lives of millions depend on it.