For example, nobody saw this coming:
Aug 08 2014 19,729,645,941 5.30% 141,230,307 GH/s
Jul 25 2014 18,736,441,558 8.08% 134,120,673 GH/s
Jul 12 2014 17,336,316,979 3.08% 124,098,191 GH/s
A lot of the first ASIC mining equipment (65nm) was preordered in 2012-2013. It was put into production during the last half of 2013. I think the 24.93% increase on June 29th got a lot of people's attention, including mine. I had a Butterfly Labs Jalapeno, and I decided it was time to sell it. I felt the equipment was near the end of its useful life, and I still wanted to sell it for a nominal value. I held on to it so long, because I never broke even with it.
Difficulty is determined by the amount of new mining equipment put into production versus the amount of old mining equipment taken out of production. I think we had three small increases in a row as many of the initial ASICs, such as the Jalapeno, were removed.
Also the price of Bitcoin is another factor that determines difficulty. If the value of Bitcoin drops, it may no longer be profitable for some mining equipment to mine. It makes sense at that point to turn off the equipment and just buy the Bitcoins directly. If the value of Bitcoin begins to skyrocket, you might see a lot of older equipment brought back online.
Perhaps a month ago, I determined that the average difficulty increase this year was about 16%, so that is what I use in my calculations. For contracts purchased near the end of July, they need average future increases to be around 12% to break even.
When you buy a contract, the purchase price and the first few difficulty increases have the most impact on whether or not you break even.
I've also noticed that we also tend to underestimate difficulty increases. In May, BFL Josh said that having 30B difficulty in September was ludicrous, but that will happen. The ASIC market is relatively competitive and there is money to be made by designing and selling more efficient hardware. It's Moore's Law at work.