| The chilly post dot-com online advertising climate has shifted the industry focus from impression-based performance metrics (CPM) to costper- click metrics. A variety of innovative,targeted advertising formats now purport to satisfy the advertisers'increased demand for performance accountability.In the second chapter this paper given the two different types of portals and their distinct appeal to online advertisers, it is important that advertisers understand the relative merits of each media outlet and decide accordingly the budget allocation between the two portals.Let n_G(t) and n_S{t) be the total number of visits to portals G and S, respectively, at time t. In reality, n_G{t) and n_S(t) can be measured in real time by using a combination of sophisticated traffic monitoring software tools such as WebTrends.Letk_G(t) andk_S (t) be the fractions of the corresponding users that type the trigger search keyword at time t. We assume that the portion of portal visitors that type the same trigger keyword at a specialized portal is higher than that at the generic portal, sok_S(t)>k_G(t).The justification of this assumption rests on the unique content offered by specialized portals and their concomitant unique user bases with very specific search interests.Let x_G(f)and x_S(t)be the average clickthrough rates for keyword ads at time t on eachportal. Then, the potential number of clickthroughs at time t at each portal will beLet A_G{t) and A_S(t) be the allocated advertising budgets at time t for G and S, respectively.We assume that the dollars spent on the advertising budget can be transformed into advertising efforts that can turn potential keyword ad clickthroughs into actual keyword ad clickthroughs.Letu_G(t)andu_S(t) denote the corresponding advertising efforts. Moreover, we assume that the advertising effort is a concave function of the advertising budget. This assumption captures the diminishing return of advertising expenditures.ThenLetρ_G > 0 andρ_S > 0 be each portal's effectiveness in converting potential clickthroughsinto actual keyword ad clickthroughs.Using an optimal control modeling approach, we have developed a dynamic decision model in the online advertising setting that captures the distinctive features of the online advertisingmedia , the specific mechanism of ad display (activated by a user keyword search), the dynamic nature of the new media, and lastly the action-based performance criterion (i.e., the actual number of keyword ad clicksthrough).The optimal control problem isThen we can get the ratio isIn the third chapter we give an adaptive algorithm for selecting profitable keywords for three cases under some assumptions, we develop a stylized model of search-based advertisingservices and formulate the problem of identifying profitable sets of keywords. Since the details of the auction process for determining ads placements vary by search engine, we will ignore the auction mechanism and any competitive responses by advertisers and assume that he bids for all keywords are constant, and thus, the costper- click of each keyword is fixed and known in advance.And through many assumptions,confirm the expected profitE[Z_A,] isThink about 3 cases1)A Static Setting: When Clickthru Rates and Expected Profits Are KnownThen2)Unknown Clickthru Probabilities with Known Expected Profits.In this case,by formulating the problem as an instance of the multi-armed bandit, we can apply existing adaptive algorithms developed for this class of problems. Unfortunately, existing multi-armed bandit algorithms have performance guarantees that deteriorate as the number of keywords increases, motivating us to develop an alternative adaptive approximationalgorithm that leverages the special structure of our problem.3)An Extension to the Case of Unknown Expected Profits. (Details for 2) and 3)are in the third chapter of this paper.)... |
PDF Full Text Request