Check out this article, "The Lies Readers Tell," from Webcomics.com.
Perhaps you are among those who have attempted to poll readers on various things, and been disappointed by the results. The article above talks specifically about readers saying they would buy a proposed t-shirt design, only for the creator to discover that it sells poorly.
"It's one, immutable fact of webcomics: Readers lie about what they'd be willing to buy," says the author.
Actually, the problem is usually in the polling. I call it the Applause Effect. When you are in a theater, you have an investment in being there. The entertainers lead the audience along, prompting reactions. An event starts people clapping. You find yourself clapping too.
Observation #1: It's a lot easier to go with the idea than to scrutinize it for flaws and be critical.
Another problem lies in t-shirt design. There is a huge glut of fair-to-good shirts out there, but great shirts are less common. First glance at a decent concept may lead to a positive reaction, but setting it aside for review later, before opening your wallet, may reveal it's not as great as you thought.
Observation #2: Good seems good because it is an improvement over average.
I haven't done exhaustive research, but I hear from people that most shirts sell best when they are first introduced. Unveiling a proposed work eliminates the grand unveiling, when impulse buys are more likely. By the time a proposed shirt hits your store, everyone has had their laugh.
You can observe this with certain popular comics that play fast and loose with the truth: they do pre-orders on a shirt at a tremendous mark-up (often $20/shirt), brag about how they are selling like crazy, but quickly drop them from their line. If this formula worked as anything more than a once-in-a-while grift, people would be doing it all the time.
Observation #3: It's better to surprise than to ask.
T-shirts are a different kind of art than comics, though some comic art translates well onto shirts with little alteration. T-shirts are graphic design, and most comics art is cartooning, not graphic design, though skills often overlap.
Some t-shirt producers never seem to throw away a single idea, and trudge from show to show with bags of boring shirts. This suggests that t-shirt conceptualizers are another skill set. I like to discard 20-30 ideas for every final design we make. I get the impression that some people produce every design they conceive, especially if it works in bacon or coffee.
Observation #4: The majority of cartoonists would greatly benefit from the help of a graphic designer before finalizing a shirt design.
There are a lot of people out there who identify themselves as graphic designers. Having the degree is not the same as having talent and experience.
Observation #5: Examine their portfolio before hiring.
Graphic design work requires a lot of messing around with aspects of the design until it either clicks or is abandoned. Most of this work takes place out of view, so you might think a good graphic designer isn't being productive.
Observation #6: Ask the graphic designer if they think your design ideas have legs before you pay them to refine them. Let them be a creative partner, as long as they are respectful of your vision.
The graphic designer should be able to give you camera-ready art separations for the printer. If not, something is wrong. Find out what.
Observation #7: It's just a matter of having the right kind of printer, and they should have one or have access to one. I like Canons.
A printer who doesn't take their time with you and hold your hand doesn't think you are going to make it. They are reading you as an amateur who doesn't do your homework and lacks perseverance to develop an online shirt business. Either they mostly make shirts for athletic teams, in which case you are in the wrong store, or they are sending you a message about your project.
Observation #8: A designer and a printer can tell you much more than your readers about the prospects for a design, but smart ones will avoid predictions because that's a fool's game.
There is a tendency for people who take a poll, even a short one, to only take it if they are supporting something they believe in. People who feel less passionate usually skip the poll.
Observation #9: Readers don't "lie." They just get over-enthusiastic and are the wrong people to ask anyway. Poorly designed polls, however, lie like crazy.
I don't enjoy contesting someone's conclusion, especially if they feel they are trying to be helpful. But once again, over-simplification of complex issues makes it seem that the webcomics business is more within grasp than it really is. It's a slander against readers, who are justified at feeling miffed for being classified as liars.
It's also hypocritical, hauling ass all over the country to kiss ass and "hail the fans" while blaming them for your own failings.