The rapid growth of e-business in the last years made sponsored search a multi-billion dollar industry, which will continue to grow in the upcoming years. Approximately 50% of the total online advertising spending today is used for sponsored search, where search engine providers use sponsored search auctions for pricing the clicks and ranking the ads based on the bids advertisers submit for a search term. The bid determines not only the price per click, but also the position of the ad in the sponsored search results, consequently costs, revenue and finally profitability of sponsored search. As the advertisers do normally not know the relationship between bids and positions of the ads in the sponsored search results and can thus not calculate the optimal bid, it is particularly challenging for advertisers to know: which response model allows for the robust prediction of a position of the ad by a certain bid and is easy to use. Using real Yahoo! Search Marketing data for bids and resulting positions from diverse e-business sectors we conduct an empirical examination of the relation between bids and positions of the ads in sponsored search results by calibrating different response models. Our findings reveal that the semi-logarithmic model i) is the most robust function for predicting a position of the ad, and ii) provides clear assistance for advertisers in terms of decision making about the bid for a search term, which is necessary to gain a certain position of the ad in the sponsored search results.