We study how the inaccuracy of a costly certification technology affects a monopolistic seller's profitability. We compare three scenarios: no certification, a 100% accurate certification, and a 50% accurate certification that produces accurate evaluations half the time. The noisy certification environment is never the most profitable and, depending on the buyers' loss aversion, can be the least profitable. However, a noisy certification can be more profitable than an accurate one, as it discourages the over‐certification that occurs in an accurate certification environment. Experimentally, the noisy certification is shown to be the least profitable treatment, whereas the accurate certification is shown to be the most profitable.