FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

20
FLOWER AUCTIONS IN AMSTERDAM
  • date post

    20-Dec-2015
  • Category

    Documents

  • view

    219
  • download

    2

Transcript of FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Page 1: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

FLOWER AUCTIONS IN AMSTERDAM

Page 2: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Ad Auctions

March 16, 2007

Page 3: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.
Page 4: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Overture slide

Page 5: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Motivation

• Market inherently interesting– 98% of Google’s and ~50% of Yahoo’s revenues– “Future of advertising”

• Unusual auction rules– Multiple units, but only one bid. Continuous time.

• Structured market– Rules. Almost like a lab. Good data.

• Purely electronic market– No goods ever shipped anywhere.

• Flexibility to change auction rules from time to time

Page 6: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Market history & evolution

early banner ads (circa 1994)

Overture (Goto.com) (1997)

per-impression pricing per-click pricing

limited targeting keyword targeting

person-to-person negotiations

automated acceptance of revised bids

generalized first-price auction rules

Page 7: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Generalized first price auctions

Problem: Generalized first price auctions are unstable.

No pure strategy equilibrium, and bids can be adjusted dynamically. Bidders want to revise their bids as often as possible.

Page 8: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Edelman and Ostrovsky, 2007

Yahoo data from June 15, 2002 to June 14, 2003

1000 top markets

10,475 bidders

18,634,347 bids

Observe bids at the quarter-hour

Page 9: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

CyclingTime Market Bidder Bid6/17/2002 6:30 AM 24 13 $5.916/17/2002 6:30 AM 24 810 $5.926/17/2002 6:30 AM 24 14 $5.936/17/2002 6:30 AM 24 13 $5.946/17/2002 6:30 AM 24 60 $5.956/17/2002 6:30 AM 24 14 $5.966/17/2002 6:45 AM 24 810 $5.976/17/2002 6:45 AM 24 13 $5.97 …6/17/2002 11:30 PM 24 13 $9.986/17/2002 11:30 PM 24 14 $9.986/17/2002 11:45 PM 24 14 $10.006/17/2002 11:45 PM 24 60 $10.006/17/2002 11:45 PM 24 13 $10.006/17/2002 11:45 PM 24 810 $10.016/17/2002 11:45 PM 24 14 $10.026/17/2002 11:45 PM 24 13 $5.126/17/2002 11:45 PM 24 14 $5.13

Page 10: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Cycling

4

5

6

7

8

9

10

11

7/1

8/2

00

2 0

:00

7/1

8/2

00

2 6

:00

7/1

8/2

00

2 1

2:0

0

7/1

8/2

00

2 1

8:0

0

7/1

9/2

00

2 1

:15

7/1

9/2

00

2 7

:15

7/1

9/2

00

2 1

3:4

5

7/1

9/2

00

2 2

0:0

0

7/2

0/2

00

2 2

:00

7/2

0/2

00

2 8

:00

7/2

0/2

00

2 1

4:0

0

7/2

0/2

00

2 2

0:1

5

7/2

1/2

00

2 2

:15

7/2

1/2

00

2 8

:15

7/2

1/2

00

2 1

4:1

5

7/2

1/2

00

2 2

0:1

5

7/2

2/2

00

2 2

:15

7/2

2/2

00

2 8

:15

7/2

2/2

00

2 1

6:3

0

7/2

3/2

00

2 6

:45

7/2

3/2

00

2 1

2:4

5

7/2

3/2

00

2 2

0:0

0

7/2

4/2

00

2 2

:45

7/2

4/2

00

2 9

:00

7/2

4/2

00

2 1

5:1

5

7/2

4/2

00

2 2

1:1

5

Page 11: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Alternative mechanisms

• Generalized first-price

• Generalized second-price– Pay the bid of the next-highest bidder– First implemented by Google (2002),

later adopted by Yahoo

• VCG– Each bidder pays the externality he

imposes on others

Page 12: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Generalized second-price auctions

Position Bidder Bid

1 A $7

2 B $6

3 C $5

Payment

$6.01

$5.01

$0.10

Page 13: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Computing VCG payments: example

Position # clicks

1 100

2 80

Bidder Valuation

A $8 B $5

C $10

C’s payment:

C pushes A from 1 to 2

C pushes B out completely

So C should pay $160+$400=$560

loss of surplus (100-80)*$8=$160

loss of surplus 80*$5=$400

Page 14: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

GSP in use

Adv Bid Payment

A $3.01 $3.01

B $3.00 $2.81

C $2.80 $1.11

D $1.10

Page 15: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

GSP versus Vickrey and VCG

• With only one slot, GSP is identical to standard second price auctions (Vickrey, VCG)

• With multiple slots, the mechanisms differ– GSP charges bidder i the bid of bidder i+1– VCG charges bidder i for his externality

“[Google’s] unique auction model uses Nobel Prize-winning economic theory to eliminate … that feeling that you’ve paid too much.”

- Google marketing materials

Page 16: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Truth-telling is not a dominant strategy under GSP

Intuition: Sometimes, bid below your true valuation. You may get less traffic, but you’ll earn greater profits.

bidder bid

A $8

B $5

C’s valuation: $10

C bids $10, pays $8 → payoff ($10-$8)*100 =$200

C bids $6, pays $5 → payoff ($10-$5)*80 =$400

Suppose there are 3 bidders but 2 positions.Positions have click-through rates 100 and 80.

$400>$200. So C should place a bid below its valuation.

Page 17: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Bidders’ actual strategies

Page 18: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

Optimal reserve prices

• What reserve price maximizes search engine revenue?

• How do outcomes differ from optimal reserve price? From the reserve price that maximizes advertiser surplus?

Method: Simulate a set of vectors of valuations. Use equilibrium bid formula to compute bids. Compute outcomes under each minimum bid rule.

Page 19: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

SE revenues and adv. surplus

si ~ lognormal ( 1, 0.12 )

maximum search engine revenue

maximum total surplusand advertiser surplus

Page 20: FLOWER AUCTIONS IN AMSTERDAM. Ad Auctions March 16, 2007.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 20

0.2

0.4

0.6

0.8

1

1.2

1.4

Minimum Bid

Searc

h E

ngin

e R

evenue &

Advert

iser

Surp

lus (

per

clic

k)

Total Surplus Search Engine RevenueAdvertiser Surplus

With fewer bidders

K=5 si ~ lognormal ( 1, 0.52 )

Max SE revenue

Max total surplusand advertiser surplus

loss in total surplus if search engine chooses reserve price

gain in SE revenue if search engine chooses reserve price

loss in adv surplusif search engine chooses reserve price