Target-cost pricing has been a widely applied formula in defence contracting. If this type of pricing arrangement is chosen, the seller's ex-post profit consists of a fixed payment plus some share of the cost overrun, that is the difference between an ex--ante agreed estimation of the production costs and the actual production costs. In an incomplete--contract setting, where relationship-- specific investments have to be made prior to the production stage, the cost-reimbursement properties of target-cost pricing work against a first best. However, since costs are verifiable, the ex-- ante contract allows to condition the initial contract on costs, that is, to stipulate a separate trade price for each cost observation, plus a special price for the no--trade case. (If costs are non--verifiable, it is only possible to fix one price for trade and one price for non--trade.) This increase in the number of instruments available to the agents works in favour of a first best. The paper shows that the positive properties of target-cost pricing outweigh the negative ones: it is possible to find prices which induce the agents to invest efficiently into relationship-specific investments, thus avoiding Williamson's hold-up problem. This result is particularly important because fixed-price contracts a la Hart-- Moore (1988) fail to achieve the first best if they are applied in the same environment in which target-cost prices succeed in attaining the first best. Since any contract, which implies full cost-reimbursement, also fails to achieve the first best, this paper shows that the first best requires just that middle-of-the-road approach which is offered by target-cost pricing.