This paper analyzes Onbid car auction data by employing various methods, including structural estimation, to identify main factors which decides auction prices and figure out what effects those factors are making on the auction price. I then discuss on how to maximize sellers’ revenue in OnBid car auctions. The government and public institutes sell their assets through the OnBid auction, hence the optimal design of the OnBid auction is important. The paper’s main findings are as follows: (ⅰ) The independent private value model explains OnBid car auction data better than the correlated private value model or the interdependent value model; (ⅱ) Both the number of bidders and the ratios of the auction price to the evaluation value were lower in the auctions posted by the Kamco than auctions by institutes other than the Kamco; (ⅲ) Some auctions require that at least two bidders should submit a bid no less than the reserve price for sale. In those auctions, both the number of bidders and each bidder’s valuation on the auctioned object were lower than in auctions without that requirement; (ⅳ) The sum of sellers’ revenue would be decreased in the simulation with the reserve price higher by 5%, 10%, and 20% across auctions by institutes other than Kamco.