I would suggest basing the cost of items on the relative skill/time required to gain them in your game.  Using a randomly chosen amount may work, but it may prove either too expensive(nobody will buy it, no profit) or too cheap(nobody will try to earn it in game) and your game will become ruled by pay to play gamers and eventually the bubble will pop and nobody will play or pay...

Just my 2c.

To expand a bit as to how to choose the value:

If you are using Gems and they are something you CAN earn in game, then you can base the cost on the time it takes to earn them. 

Most importantly, you should choose a value that you might be willing to pay, not just a value that you would like to earn.

If you are using Gems and Gems are NOT something you can earn in game directly.  Then you could base the cost of Gems on their relative value in exchange for game items you CAN earn in game.

So for example lets say my game uses gold pieces as currency, and I setup Gems on my store so people can buy them and exchange them for gold pieces and/or other gear/etc..

If it takes me 1 hour to earn 100 gold, it may be worth it to buy 80 Gems for .99 if 80 Gems will get me 800 gold pieces.. etc..

It has to do with time/effort vs. cost.

It has feel like the exchange is Valuable to the player or it won't be of any value to you, the game producer.

Store value gaps are generally calculated to reward the buyer for spending more money (thus encouraging them to do so).
80/$1 for base price
500/$5 = (base price * 5) + 100 worth of "thank you for spending more money"
1200/$10 = (base price * 10 + 200) + 200 worth of "thank you for spending even more money"
Of course, the final step is "eyeballing it."  A good-looking final number is ultimately more important than is a mathematically consistent formula.