Last week saw the end of an ambitious experiment: Panera Bread's pay-what-you-want turkey chili, which the sandwich chain started offering at 48 St. Louis-area stores to raise awareness of food insecurity and offer meals to the needy. Customers could pay the $5.89 "suggested" price, or more or less, or nothing. After an initial burst of interest in March, company executives said that customers just stopped realizing the option existed, and they pulled the option for "retooling."

This isn't Panera's first time at at the pay-what-you-can rodeo. In 2010, the company's charitable foundation opened its first "Panera Cares" cafe in a St. Louis suburb, and now has five of them across the country. Overall, they bring in about 70 percent of a traditional store's revenue, which is enough to run job training programs for the needy. The pricing scheme has grown more popular over the years, with high-profile record releases, theater performances, and donation-based app downloads.  It's the kind of thing that looks appealing in a world where people expect more content for free, but free doesn't pay the bills.

So what works and what doesn't with write-your-own price tags? University of California  San Diego marketing professor Ayelet Gneezy is the subject-matter expert, having published several influential studies on how consumers behave when given a sliding scale of options. The less-than-smashing success of Panera's turkey chili, she says, probably had to do with the fact that uninitiated patrons were confused about why that menu item was different from the rest. "It might be that communication is lacking," she says. "Panera knows this--you need to be in touch with people all the time. It should be, I go in, I get it, I do what I want, I leave."

Here's what else the science says about what makes pay-what-you-want systems successful:

- Charity: In a 2010 study, Gneezy tried selling people photos of themselves at an amusement park under four different conditions. In one, the photos were priced at a flat fee of $12.95, and only 0.5 percent of people bought one. That increased to only 0.57 percent when people were told half the price went to charity. When people could pay what they want for the photos, many more were sold, but at unsustainably low prices. When people could pay what they wanted and half the price went to charity, Gneezy made the most profit, and generated lots of money in donations as well.

- Nudging: If people have an idea of what the offeror of the good or service thinks is fair, they're likely to stick near it.  "People tend to anchor on whatever number they're given," Gneezy says. That also allows those with means to feel validation for paying more than asked. However, it's got to be a reasonable price. In a 2012 experiment with selling photos on a cruise, everyone was told that the value of the photo was $15. Some were asked to pay $5, and some could pay what they wanted. Ultimately, fewer people bought a photo in the pay-what-you-want scenario than when they were asked to pay $5, because they didn't want to pay $15, but were afraid of appearing too parsimonious if they paid less.

"If $5 seems unfairly low, people find it easier to maintain self-image by foregoing the purchase altogether," she wrote. "However, when the company sets the price at $5, there is no ambiguity about fairness, self-image concerns disappear, and people are happy to pay."

- Privacy: In another experiment in a pay-what-you-want buffet in Vienna, Austria, Gneezy tested whether people paid more if the owner knew or if the donation was sealed. The  anonymous donations were higher, potentially because people pay more when they don't think it's because they "have to," for fear of judgment.

- Snap decisions: "If you ask people to think about how much you want to pay before they actually do pay, they pay less," Gneezy notes.

- Requiring people to pay something: The dynamics of pay-what-you-want are somewhat different with digital goods, but it can still work, as the makers of the "Humble Bundles" of indie video games found when they packaged their wares with donations to like-minded organizations like the Electronic Frontier Foundation. This works best when people have to at least register their credit card information, because laziness will win out over a sense of responsibility more times than not.

- Adjusting offers based on what buyers pay: The "Fairpay" mechanism allows a seller to make different offers to an individual buyer based on what they choose to pay each time, building in an incentive structure to paying a fair price without forcing people to do so (and potentially losing customers who might pay something more than nothing).


- Making reputation matter: I'm reminded about the pay-what-you-want ridesharing service Lyft, which allows passengers to pay for their transportation with a discretionary donation--but also for drivers to rate passengers, which influences who's willing to pick them up in the future. It's a less formal incarnation of the Fairpay idea, with pricing driven by community norms.

Overall, it's probably not a good idea to undertake a pay-what-you-can system if you can't afford to lose a lot of money, because lots can go wrong--but it's possible to minimize the risk.