We examine IT contracting in one particular segment of the IT outsourcing market - the market for large scale packaged software implementations such as Enterprise Resource Planning (ERP) systems. Using a small sample of actual outsourcing contracts in several industries and a review of the relevant outsourcing literature, we determined the common provisions and structural characteristics of these contracts. We then used this description to develop an analytical model of IT outsourcing, using principal agent techniques. Our model captures key characteristics of these IT contracts including a multi-stage project structure, vendor learning, probabilistic binary outcomes (success/failure), use of incentive contracting, and implementation risks. In addition to deriving the optimal IT contract; we specifically focus on how vendor learning affects the optimal structure of the contract. In particular, we find that with rapid vendor learning at the early stage of the contract, it is often more efficient to use a multistage contracting procedure, even if it is not technically required, as this enables stronger incentives to be given to the vendor and creates greater profit and surplus.