Query words
===========

    * Car brands from autobytel.com.
    * Generic terms such as car, cars etc.
    * Brand with location.
    * Brand and model.
    * Brand with typos.

Local markets are important
===========================

Query Word(s)       #1       #2        #3        #4        #5    #6        #7    #8
------------------------------------------------------------------------------------
Car               5.00(gm)  2.80      1.01      0.98      0.97  0.96      0.94  0.79
Cars              5.00(gm)  2.80      1.01      0.98      0.97  0.96      0.94  0.79
Cars_Boston       5.00(gm)  4.62      4.00(gm)  3.99      3.00  2.23      2.22  2.00
Cars_NewYork      5.01      5.00(gm)  4.99      4.00(gm)  3.00  3.00      2.50  1.16
Cars_Philadelphia 6.51      6.50      6.01      5.00(gm)  4.02  4.00(gm)  3.01  3.00
Cars_Seattle      6.44      6.01      5.00(gm)  4.00(gm)  3.00  3.00      2.99  2.41

* Yahoo seems to treat 'Car' and 'Cars' as the same (stemmed word).
* Overall, localized online ads tend to be more expensive.
* GM has a bid for $5 regardless of what others are bidding (100k Warranty).
* GM also has a bid for $4 for its regular advertisement (Oct 13) targeting
  local market. But that does not shown up today (Oct15).

Potential business opportunity: 
    * Advertisement consultation/management for big advertisers.
    * Organize group buying against Google/Yahoo (time division ads placement).
        Suppose companies A and B are interested in the search word K. 
        Now the ad management agency can divide time into X-minute slot,
        and places bids for A and B alternatively every X minutes. Once
        enough companies sign up as clients, they can enjoy significant
        reduction on advertisement cost.


Bid on typos (why not every thing?)
===================================

Query word   #1    #2    #3    #4    #5    #6    #7    #8
---------------------------------------------------------
Cadilac    1.03  1.03  0.95  0.89  0.70  0.62  0.61  0.60
Cadillac   1.03  1.03  0.95  0.89  0.70  0.62  0.61  0.60
Cadillak   2.00  0.29  0.26  0.25  0.10  none  none  none

* Yahoo seems to treat certain typos as if they are correct,
  but not others. 

Potential opportunity for advertisers:
    * Bid on every thing (including machine learning) for 
      just 1c. This could get free ad impression at no cost 
      most of the time and pay only if the ad is clicked.


Brand/model based bidding
=========================

              #1   #2   #3   #4   #5   #6   #7   #8
----------------------------------------------------
Cadillac     1.03 1.03 0.95 0.89 0.70 0.62 0.61 0.60
Cadillac_CTS 1.01 0.86 0.77 0.77 0.75 0.75 0.70 0.69
Cadillac_DTS 0.84 0.84 0.80 0.79 0.76 0.75 0.67 0.59
Cadillac_SRX 0.78 0.76 0.75 0.75 0.75 0.68 0.67 0.57
Cadillac_STS 0.75 0.75 0.61 0.59 0.59 0.55 0.54 0.52
Cadillac_XLR 0.75 0.75 0.57 0.56 0.55 0.54 0.51 0.50

* Competition isn't much lower on a specific brand/model
  than on a specific brand.


Statistical analysis - bids by brand
====================================

It is interesting to see how much more advertisers 
pay to get in the #1 slot. The ratio between #x and
#1 per brand is calculated and then averaged over
all brands. People are paying 20-45% more to be the
#1, instead of #2 to #4 on the same page.

I am not sure about the benefit of paying significantly
more (15-20%) to move from #4 to #2.

Percent of advertisement cost comparing to the #1 slot
------------------------------------------------------
#2   83.4%
#3   75.8%
#4   70.1%
#5   65.8%


It is also interesting to find out 
  * The less expensive brands spend more on online advertisement.
  * The special named brands tend to spend the least.

#1 Bids by brand
----------------
 [1,] "Lamborghini"  "0.11"
 [2,] "Panoz"        "0.21"
 [3,] "Ferrari"      "0.27"
 [4,] "Maserati"     "0.32"
 [5,] "RollsRoyce"   "0.36"
 [6,] "AstonMartin"  "0.37"
 [7,] "Bentley"      "0.37"
 [8,] "Morgan"       "0.4" 
 [9,] "MINI"         "0.41"
[10,] "Suzuki"       "0.52"
[11,] "Isuzu"        "0.66"
[12,] "Pontiac"      "0.66"
[13,] "Porsche"      "0.68"
[14,] "LandRover"    "0.7" 
[15,] "Saab"         "0.73"
[16,] "Audi"         "0.82"
[17,] "Buick"        "0.85"
[18,] "MercedesBenz" "0.86"
[19,] "BMW"          "0.87"
[20,] "Chrysler"     "0.87"
[21,] "Lotus"        "0.91"
[22,] "Jaguar"       "0.92"
[23,] "Volvo"        "0.93"
[24,] "Acura"        "0.96"
[25,] "Hummer"       "0.96"
[26,] "Infiniti"     "0.99"
[27,] "Cadillac"     "1.03"
[28,] "Mitsubishi"   "1.03"
[29,] "Dodge"        "1.06"
[30,] "Volkswagen"   "1.07"
[31,] "GMC"          "1.08"
[32,] "Honda"        "1.09"
[33,] "Mercury"      "1.09"
[34,] "Saturn"       "1.12"
[35,] "Chevrolet"    "1.13"
[36,] "Subaru"       "1.17"
[37,] "Scion"        "1.18"
[38,] "Lincoln"      "1.19"
[39,] "Jeep"         "1.22"
[40,] "Ford"         "1.28"
[41,] "Nissan"       "1.28"
[42,] "Hyundai"      "1.40" 
[43,] "Mazda"        "1.50" 
[44,] "Toyota"       "1.58"
[45,] "Lexus"        "1.67"
[46,] "Kia"          "2.22"

#4 Bids by brand
----------------
 [1,] "AstonMartin"  "0.1" 
 [2,] "Lamborghini"  "0.1" 
 [3,] "Panoz"        "0.1" 
 [4,] "RollsRoyce"   "0.1" 
 [5,] "Maserati"     "0.19"
 [6,] "MINI"         "0.23"
 [7,] "Morgan"       "0.23"
 [8,] "Bentley"      "0.24"
 [9,] "Ferrari"      "0.26"
[10,] "Suzuki"       "0.3" 
[11,] "Porsche"      "0.38"
[12,] "Isuzu"        "0.5" 
[13,] "Mitsubishi"   "0.52"
[14,] "Pontiac"      "0.56"
[15,] "Jaguar"       "0.59"
[16,] "LandRover"    "0.59"
[17,] "Buick"        "0.6" 
[18,] "BMW"          "0.62"
[19,] "Lotus"        "0.66"
[20,] "Audi"         "0.69"
[21,] "Saab"         "0.69"
[22,] "Subaru"       "0.71"
[23,] "Volkswagen"   "0.71"
[24,] "Dodge"        "0.73"
[25,] "Scion"        "0.73"
[26,] "Mercury"      "0.74"
[27,] "GMC"          "0.75"
[28,] "Hummer"       "0.75"
[29,] "Lexus"        "0.75"
[30,] "Saturn"       "0.76"
[31,] "Chrysler"     "0.78"
[32,] "Honda"        "0.81"
[33,] "MercedesBenz" "0.82"
[34,] "Infiniti"     "0.84"
[35,] "Acura"        "0.85"
[36,] "Chevrolet"    "0.86"
[37,] "Volvo"        "0.86"
[38,] "Cadillac"     "0.89"
[39,] "Lincoln"      "0.89"
[40,] "Jeep"         "0.93"
[41,] "Kia"          "0.95"
[42,] "Toyota"       "0.98"
[43,] "Ford"         "1.00"   
[44,] "Mazda"        "1.00"   
[45,] "Nissan"       "1.02"
[46,] "Hyundai"      "1.11"


Appendix - Splus script and raw output
======================================

all_scan("C:/users/h/cis620/brand.txt",what="")
all_matrix(all, ncol=3, byrow=T)
brand_all[,1]
bid_as.numeric(all[,2])
pBrand_'begin'
idx_bid
for (i in 1:length(brand)) {
    if (brand[i] == pBrand) {
	idx[i]_idx[i-1]+1
    } else {
	idx[i]_1
    }
    pBrand_brand[i]
} 
skip_((brand != "Maybach") & (brand != "Saleen"))
brand_brand[skip]
bid_bid[skip]
idx_idx[skip]
> print(mean(bid[idx==2]/bid[idx==1]))
[1] 0.8341862
> print(mean(bid[idx==3]/bid[idx==1]))
[1] 0.7582417
> print(mean(bid[idx==4]/bid[idx==1]))
[1] 0.7014977
> print(mean(bid[idx==5]/bid[idx==1]))
[1] 0.6583173

print(cbind(brand[idx==1][order(bid[idx==1])],bid[idx==1][order(bid[idx==1])]))
print(cbind(brand[idx==4][order(bid[idx==4])],bid[idx==4][order(bid[idx==4])]))

 [1,] "Lamborghini"  "0.11"
 [2,] "Panoz"        "0.21"
 [3,] "Ferrari"      "0.27"
 [4,] "Maserati"     "0.32"
 [5,] "RollsRoyce"   "0.36"
 [6,] "AstonMartin"  "0.37"
 [7,] "Bentley"      "0.37"
 [8,] "Morgan"       "0.4" 
 [9,] "MINI"         "0.41"
[10,] "Suzuki"       "0.52"
[11,] "Isuzu"        "0.66"
[12,] "Pontiac"      "0.66"
[13,] "Porsche"      "0.68"
[14,] "LandRover"    "0.7" 
[15,] "Saab"         "0.73"
[16,] "Audi"         "0.82"
[17,] "Buick"        "0.85"
[18,] "MercedesBenz" "0.86"
[19,] "BMW"          "0.87"
[20,] "Chrysler"     "0.87"
[21,] "Lotus"        "0.91"
[22,] "Jaguar"       "0.92"
[23,] "Volvo"        "0.93"
[24,] "Acura"        "0.96"
[25,] "Hummer"       "0.96"
[26,] "Infiniti"     "0.99"
[27,] "Cadillac"     "1.03"
[28,] "Mitsubishi"   "1.03"
[29,] "Dodge"        "1.06"
[30,] "Volkswagen"   "1.07"
[31,] "GMC"          "1.08"
[32,] "Honda"        "1.09"
[33,] "Mercury"      "1.09"
[34,] "Saturn"       "1.12"
[35,] "Chevrolet"    "1.13"
[36,] "Subaru"       "1.17"
[37,] "Scion"        "1.18"
[38,] "Lincoln"      "1.19"
[39,] "Jeep"         "1.22"
[40,] "Ford"         "1.28"
[41,] "Nissan"       "1.28"
[42,] "Hyundai"      "1.40" 
[43,] "Mazda"        "1.50" 
[44,] "Toyota"       "1.58"
[45,] "Lexus"        "1.67"
[46,] "Kia"          "2.22"

 [1,] "AstonMartin"  "0.1" 
 [2,] "Lamborghini"  "0.1" 
 [3,] "Panoz"        "0.1" 
 [4,] "RollsRoyce"   "0.1" 
 [5,] "Maserati"     "0.19"
 [6,] "MINI"         "0.23"
 [7,] "Morgan"       "0.23"
 [8,] "Bentley"      "0.24"
 [9,] "Ferrari"      "0.26"
[10,] "Suzuki"       "0.3" 
[11,] "Porsche"      "0.38"
[12,] "Isuzu"        "0.5" 
[13,] "Mitsubishi"   "0.52"
[14,] "Pontiac"      "0.56"
[15,] "Jaguar"       "0.59"
[16,] "LandRover"    "0.59"
[17,] "Buick"        "0.6" 
[18,] "BMW"          "0.62"
[19,] "Lotus"        "0.66"
[20,] "Audi"         "0.69"
[21,] "Saab"         "0.69"
[22,] "Subaru"       "0.71"
[23,] "Volkswagen"   "0.71"
[24,] "Dodge"        "0.73"
[25,] "Scion"        "0.73"
[26,] "Mercury"      "0.74"
[27,] "GMC"          "0.75"
[28,] "Hummer"       "0.75"
[29,] "Lexus"        "0.75"
[30,] "Saturn"       "0.76"
[31,] "Chrysler"     "0.78"
[32,] "Honda"        "0.81"
[33,] "MercedesBenz" "0.82"
[34,] "Infiniti"     "0.84"
[35,] "Acura"        "0.85"
[36,] "Chevrolet"    "0.86"
[37,] "Volvo"        "0.86"
[38,] "Cadillac"     "0.89"
[39,] "Lincoln"      "0.89"
[40,] "Jeep"         "0.93"
[41,] "Kia"          "0.95"
[42,] "Toyota"       "0.98"
[43,] "Ford"         "1.00"   
[44,] "Mazda"        "1.00"   
[45,] "Nissan"       "1.02"
[46,] "Hyundai"      "1.11"