This paper proposes a model for a certification market with an imperfect testing technology. Such a technology only assures that whenever two products are tested the higher quality product is more likely to pass than the lower quality one. When only one certifier with such testing technology is present in the market, it is found that this monopoly certifier can be completely ignored in equilibrium, in contrast to the prediction of a model with perfect testing technology. A separating equilibrium is also supported in which only relatively high quality types (products) choose to pay for the certification service. It is true that in such an equilibrium having a certificate is better than not. The exact value of a certificate, however, depends both on the prior distribution of product quality and the nature of the testing technology. Welfare accounting shows that the monopolistic certifier's profit maximizing conduct can lead to under or over supply of certification service depending on model specication. Optimal certification fee is always positive and such that it makes all positive types choose to test. In the case of two competing certifiers with identical testing technologies, the intuition of Bertrand competition does not necessarily hold. Segmentation equilibrium in which higher seller types choose the more expensive certification service and not so high types choose the less expensive service can be supported. As an application, we argue that the fee differentiation between major and non-major auditing firms need not be a result of any dierences in their auditing technologies.