[citricsquid]

That's true (re: quality), I've spent $50 on a few t-shirts from my favourite quality companies (http://jinx.com t-shirts wear like a dream)

An idea that might fit with your site would be to offer a shirt of the month deal (bustedtees example: http://www.bustedtees.com/shirtofthemonth).

$200 upfront for 12 months of shirts (works out at about $13/m), every third week of the month you generate a random shirt for me from that weeks shirts and if that shirt matches the least popular I also get a bonus shirt (fits with the theme of the main weekly sale) or maybe if I am a subscriber I get double the chance of a bonus shirt (maybe least popular and most popular or least popular and second least popular) or something.

You could even do it like a loss leader type deal, have a yearly subscription for $100 (which would be ridiculously low and pull lots of people in) and then use the marketing from that (people telling their friends about the cool shirt that just arrived, tweeting about it etc. every month).

Although I guess as you're doing t-shirts on demand there's probably not much financial outlay so cash up front might not be worth potentially losing profits.

If your shirt quality is good (I'm going to order a shirt later to find out) then I'd sign up to that sort of deal ($200, or $100) without a second thought because there's nothing better than getting a surprise t-shirt.

[seekely]

The t-shirts are of quality, I promise :)

The subscription suggestion is great and might be a better fit for Nifty. I'll have to give some thought.

Another direction I am thinking of going is more clearly separating the purchase from the pick. So you pick your shirt to buy and then make a separate pick for the least popular. I worry the additional UI may complicate things, but it is probably worth testing.

[kerneis]

No, regardless of UI, I think the dilemma of having to buy yourself the least popular one to get an extra t-shirt is important. And what would happen to your buisness model if everybody liked (and ordered) t-shirt A but nominated t-shirt B? You would have to ship two t-shirts to everybobdy, whereas your current plan guarantees that no more than 20% of customers get a free t-shirt --- which also makes it more challenging to play the game.

[CaveTech]

Not to be nit-picky, but $200/12 is closer to 17 than 13.

[thedufer]

Maybe he calculated assuming a free shirt if you randomly get a least-popular? This means 14.4 instead of 12 per year, and truncates to $13 (rounds to $14).