ContentsIntroduction . 3Research Objective . 3Need for Experimentation . 3Methodology.
5Experiment Set Up . 5Conclusion . 9Appendix . 10References . 10IntroductionThe technological know-how of Auctions have been well researched Krishna, V, 2009 through the agesAuctionTheory and there’s rich scholarly literature around it, in this contemporary issues study weare specializing in an extension of Auction Theory that hasnt acquired enough interest yet,Auction of auctions.
This is quite a gap field because the range of applications of this field is quitenarrow. Interestingly this topic is having a resurgence in the mobile advertisement space. This iswhy IIMB in collaboration with InMobi1, a mobile ads and discovery platform, isconducting a study on knowing the auction dynamics when distinctive stages of auctionsare mixed with distinct agents having unbiased goals. Need for ExperimentationThe need for growing this experimentation has stemmed from the public sale situations at InMobiand the requirement to optimize the public sale methodology at InMobi going forward. At InMobi,the public sale for commercial space is set at three levels; Advertisement Exchange AdEx,Advertisement Pools advertisement networks and Buyers. These entities perform two levels ofauctions, one at the adex level and an alternative at the adpool level; both the auctions are doneinternally within InMobi.
The buyers are external agents commercial agencies with adcontent to distribute. The adpool is a set of such buyers segregated on the premise of contenttype, purchaser type and company units. The public sale at the adpool level is on a contingent good,the contingent good being the ad space contingent on the fact that the adpool wins the auctionat the central ad trade. The layout on how the auctions are structured is can be seen below,The complexity of such a model has led to InModi needing modern research onunderstanding how this model can be optimized from an auctioneer point of view and avoidsystem inefficiencies such as double marginalizationDellarocas, C, 2012 etc. This has created aneed to extend existing analysis on auction of auctions with experimental to understand thesystem better.
This study has diagnosed a analysis paper Auctions with Intermediaries by Feldman EtalFeldman Et al , 2010 to lay the premise of this analysis. Feldmans analysis via mathematicalmodel stories concerning the ad auctions in the market place. The paper considers a atmosphere wherebuyers advertisers of a completely unique and indivisible good adverts try to buy the same froma central seller Ad Publishers. This is model is very comparable to the model at InMobi; this teamhas created an interpretation document of this cloth for extra understanding in Exhibit A. The mathematically proven theorems provided in the paper are: Theorem 1: Intermediaries use posted take it or leave it prices of their internal auctions asa dominant method when it comes to a single buyer.
Theorem 2: Intermediaries set the top-quality posted price for their inner auctionsrandomizing a cost from the range . They are detached about the actual value aslong as it falls in the period. Theorem 3: Posits in regards to the effect of changes in r and n on . Theorem 4: The greatest reserve price for Central Seller is inversely proportional to thenumber of intermediaries. Theorem 5: The facilities gold standard price will always be better than zero, irrespective of the no:of intermediaries.
MethodologyThe first task is to identify which all theorems truly are valid in a multi buyer situation. Thislead as to discounting theorems 1,2 and 3. This is because the first 3 theorems are about theposted take it or leave it price it’s implemented at the adpool level. This becomesirrelevant since a posted price methodology won’t work in a multi user scenario where anactual public sale has to take place and a posted price will limit the participation of buyers aswell as not meet the highest quality circumstances for the auctioneers. Hence we will be concentrating on the theorems 4 and 5 which speak in regards to the reserve price atthe adex central public sale.
This study will create an test setup drawing from secondary research onexperimentation reminiscent of Marketing analysis: An applied orientation by Malhotra, N. KMalhotra, N. K, 2008, Baird, D. C, 1962, Experimentation: an introduction to measurementtheory and test design etcExperiment Set UpThe aim of this experiment is to determine how auctioneers and buyers will behave when there aremultiple buyers current in a scenario with intermediary auctions. The test plans toidentify concomitant edition while decreasing the effect of different possible causal factors. Theoverall structure of the public sale is as indicated below;There are two levels of auctions during this test setup, Level 0 is at the Advertisementnetwork level which aggregates j buyers in each community.
Level 1 is at the Ad Exchange levelwhere a collection of n ad networks bid for the central ad stock for his or her respective winningbuyers. The public sale environment would contain at the center an ad trade that is a auctioning off anindivisible and unique ad inventory to a set of ad networks. The auctioneers themselves have novalue for the inventory whereas each of the buyers has their own inner most value for theinventory. The deepest values of the stock for the buyers are assumed to be uniformlydistributed in quite a number . The bidders aren’t conscious about the inner most values of different bidders. The public sale at the middleman ad n/w is on the contingent good of the ad n/w themselvesbeing able to win the next auction, if not the win at a community level is meaningless.
Asecond priced public sale format might be conducted at both the trade and network level. Thecenter will design an auction with a reserve price r. As part of the test, we will infer about the variation in the reserve price r with thechanging parameters within the auction setup. To summarize the auction situations:1. Indivisible and unique good at the center.
2. j buyers per ad n/w, can be chosen based on available impartial buyers in InMobi. 3. n ad networks in the centers auction can be chosen based on available ad pool inInMobi. 4.
Buyers have their own deepest value for the inventory with its distribution over . 5. Second priced auctions at central and ad pool level. 6. Central auction is run with a reserve price rAs a part of the experiment, InMobi will must keep the auction conditions same acrossexperiment trials and only tweak parameters that are discussed in the experiment.
Proposition 1 The choicest reserve price for Central Seller is inversely proportional to thenumber of intermediaries. The aim of this experiment is to identify the effect of the number of intermediaries on theoptimum reserve price of the vendor. There is currently no mechanism/mathematical model topredict the highest quality reserve price for a central auction with assorted buyers at level 0. HenceInMobi will ought to identify the ultimate reserve price via dissimilar rounds of auctions fora chosen set of circumstance as discussed in the test setup. Step 1: The experiment should be repeated distinctive times by altering only the reserve price inthe central public sale and through trial and error the highest quality reserve price for maximum revenuecan be diagnosed.
The next step is to identify the effect of the number of intermediaries n in the reserve price. This consists of altering the cost of n to n+/ and re executing Step 1 to find most advantageous reserveprice. Using this test we can identify the dating of n to order price. e. the most advantageous reserve price.
Let the Optimum Reserve price be Z. Let it be X. Let it be Y. Expected Result: According to our Proposition, X must be lower than Z and Y have to be greaterthan Z. Proposition 2 The centers finest price will always be better than zero, regardless of theno: of intermediaries. This proposition, is about knowing even if any eventualities exist in which a zero reserveprice at the central public sale is a good idea.
Theorem 5 proves that reserve price should bepositive no matter how big the number of intermediaries. This theorem would also rely on the outcomes of the experimentation in Proposition 4, ifwe can prove that the relationship is one of inverse nature then the reserve price shoulddisappear at large no: of intermediaries. It would not be pragmatic to collapse the ad poolsto create an extremely large number of smaller ad pools. In reality the ad pool size of Inmobi isnot going to be that high. References Feldman, J. , Mirrokni, V.
, Muthukrishnan, S. and Pai, M. M. , 2010, June. Auctions withintermediaries.
In Proceedings of the 11th ACM conference on Electronic commerce pp. 2332. ACM. Kirchkamp, O. and Moldovanu, B. , 2004.
An experimental evaluation of auctions withinterdependent valuations. Games and Economic Behavior, 481, pp. 54 85. Krishna, V. , 2009. Auction theory.
Academic press. Dellarocas, C. , 2012. Double marginalization in performance based advertising: Implicationsand answers. Management Science, 586, pp. 1178 1195.
Baird, D. C. , 1962. Experimentation: an creation to size theory and experimentdesign. Prentice Hall.
Malhotra, N. K. , 2008. Marketing research: An applied orientation, 5/e. Pearson EducationIndia.