Electronic Auctions for Perishable Goods : Lessons Learned from a Decade in the Dutch Flower Industr

1
Electronic Auctions for Perishable Goods
Lessons Learned from a Decade in the Dutch
Flower IndustryEric van HeckAUEB, Athens,
June 30, 2003e.heck_at_fbk.eur.nl
2
Menu

• Motivation and Focus
• First study Reengineering Dutch Flower
  Auctions
• Second study Screen Auctioning
• Third study Buying-At-A-Distance (KOA)
• Fourth study KOA Bidder Analysis
• Conclusions

3
Focus talk

• Central question of electronic market theory how
  does Information and Communication Technology
  (ICT) change market behavior?
• Focus this talk on traditional vs. electronic
  markets, not on the (electronic) markets vs.
  hierarchies debate.
• We are moving from place to space!

4

• Many changes in switching from traditional to
  electronic markets occur often simultaneously
  varieties of traditional markets and electronic
  markets occur. Consequently, many differences
  between traditional and electronic markets as
  well.
• Which differences make a difference?
• Methodological challenge in separating them!
• This talk presents several analyses aimed at
  separation

5
First study Reengineering the Dutch Flower
Auctions

• what are the characteristics and effects of the
  four electronic auction initiatives in the Dutch
  flower industry?
• what are the reasons for the failures and the
  successes of these electronic initiatives?
• what can we learn?
• Four case studies in Dutch flower industry
• (Kambil van Heck, Information Systems
  Research, 1998)

6
Dutch flower industry

• Holland is the worlds leading producer and
  distributor
• Flowers 59 market share
• Potted plants 48 market share
• VBA in Aalsmeer and BVH in Naaldwijk/Bleiswijk
  annual turnover of 1,5 billion each
• Growers are the sellers, wholesalers/retailers
  are the buyers

7
Flower auction hall
8

• Flowers transported from cold-storage warehouse
  to auction hall on carts.
• Through auction hall below the respective clock
  (2-3 clocks per hall), sample shown by raiser
  to buyers.
• Buyers bid using Dutch auction clock price
  starts high and drops fast. First person to stop
  the clock wins and pays that price. Invented in
  1887.
• Extremely fast! On average on transaction every 3
  seconds.

9
Dutch auction clock
10
Distribution to buyers
11
Four Case Studies

• Vidifleur Auction 1991
• Sample Based Auction 1994
• Tele Flower Auction as new entrant 1995
• Buying At a Distance Auction 1996

12
1. Vidifleur Auction (VA)

• BVH / Potted plants / 1991
• real time video images displayed at a screen in
  the auction hall
• product representation real lot on site and
  video image on screen
• buyers bid in the auction hall and on-line

13
Why was VA a failure?

• no new efficiencies for the buyers
• quality of the video display was poor
• trading from outside the hall created an
  informational disadvantage (no social interaction)

14
2. Sample Based Auction (SBA)

• VBA / Potted Plants / 1994
• Logistics directly from growers to buyers place
• Quality grading on sample
• EDI technology
• Product representation sample of lot

15
Why was SBA a failure?

• Buyers didnt trust the sample
• Slower auction because of specification of
  packaging/delivery by buyers
• Next day delivery was for some buyers difficult
• SBA became in a dead spiral decreasing supply -
  lower prices

16
3. Tele Flower Auction (TFA)

• East African Flowers / Flowers / 1995
• Buyers can search supply data base
• Logistics from storage rooms to buyers place
• Product representation real time digital image
  on screen
• Buyers bid on-line via ISDN connection

17
(No Transcript)
18
Tele Flower Auction
19
Why is TFA a success?

• Buyers trust the quality of the flowers
  (indicated on their screen)
• After-sales process is fast delivery within 30
  minutes by EAF
• Use of Dutch auction clock no learning barriers

20
4. Buying at a Distance auction (KOA)

• BVH / Flowers / 1996
• Buyers can search supply data base
• Logistics via auction room to buyers place
• Buyers can bid off-line and on-line
• Real lot on site, digital image on screen

21
TFA and KOA
22
(No Transcript)
23
Why is KOA a success?

• Better overview and communication between
  purchase and sales people of the wholesale firms
• Lower travel costs for on-line buyers
• Amount of buyers (physically or electronically
  connected) will be stable or increase expect
  increasing prices

24
Critical factors

• Vidifleur Auction product representation on
  screen, information disadvantage of online buyers
• Sample based auction product representation by
  sample, slower auction, unequally distributed
  benefits for sellers and buyers
• Tele flower auction digital product
  representation, logistics, ISDN technology, only
  way to get African products, low learning costs
• Buying At a Distance More reach for buyers and
  auctioneer

25
A model of Exchange ProcessesUpdated version
(2002)
trade context processes
product representation
risk management
regulation
influence
dispute resolution
communications computing
authentication
search
valuation
logistics
payment settlements
basic trade processes
in Making Markets" Kambil Van Heck (2002).
Harvard Business School Press. June 2002
26
Two hurdles to value

• New electronic markets challenge the status quo
  and the existing relationships between buyers and
  sellers.
• New market mechanisms must at a minimum improve
  some or all the basic processes.

27
Achieve critical mass quickly

• Subsidize early user adoption
• Increase the cost of alternative transaction
  mechanisms
• One step at the time.
• Reduce transition risk and effort

28
A Framework for Action

• Buyers
  Market Maker Sellers
• 
  or Auctioneer
• Processes
• Search
• Pricing
• Logistics
• Payment
• Settlement
• Authentication
• Product representation
• Regulation
• Risk management
• Influence
• Dispute resolution
• Communications
• Computing

29
For each process, conduct the five step analysis

• Map the current structure of market processes
• Identify how new technologies may be used to
  reengineer major market processes
• Consider how required process changes will affect
  each stakeholder
• Develop strategies for attracting important
  stakeholders
• Develop an action plan for introducing new
  trading processes

30
Second study Screen Auctioning

• What are the implications of electronic product
  representation?
• Field study at a large Dutch flower auction
  (Koppius, van Heck, and Wolters, forthcoming in
  Decision Support Systems)

31
Screen Auctioning why?

• High logistical complexity of transporting
  flowers through the auction block.
• Logistical and trade processes are tightly
  coupled.
• Breakdown of logistics causes immediate halt of
  trading.
• How to decouple the logistical processes from the
  trade processes?

32
Screen Auctioning Implementation

• Replace the physical product representation with
  electronic product representation.
• Flowers remain in cold storage warehouse and go
  directly to the shipping area after the sale
• Buyers are still in the auction hall and see a
  (generic) picture of the flower instead, plus the
  regular product characteristics of the old
  situation.
• Not a fully electronic market, but a step towards.

33
Screen Auctioning Implementation
34
Screen Auctioning Implementation

• Screen auctioning introduced in February 1996
• for Anthuriums, later also for Gerbera

35
Screen Auctioning Theory

• Electronic product representation lacked certain
  information cues for bidders
• Color
• Possible diseases or imperfections
• Stiffness of the stem (important freshness
  indicator!)
• Lemons problem! (Akerlof, 1970)

36
Screen Auctioning Main Hypotheses

• Overall less product quality information
  available, so we have
• Hypothesis 1 Screen auctioning will lead to
  lower prices
• Hypothesis 2 The screen auctioning effect will
  be stronger for more expensive flowers

37
Screen Auctioning Data

• Transaction database available, containing data
  on the transaction (price, quantity, date), as
  well as the flower (diameter, stemlength, quality
  code) and the identity of buyer and grower.
• Additional control variable VBN-price, average
  Anthurium price at all other Dutch flower auction
  for that month
• All Anthurium transactions from 1995-1997
• (N 372,856)

38
Screen Auctioning Analysis

• OLS Regression model
• PRICE ? ?1DIAM ?2WKDAY ?3VBN
  ?4QUANT ?5,IFLWTYPEi ?6 SCRAUC ?.
• R2 0.588
• ?6 is negative overall, as well as for 8 of the
  9 flower-subtypes separately.
• Conclusion hypothesis 1 accepted

39
Screen Auctioning Analysis

• Hypothesis 2 R2 -.735 (sig. lt 0.05)

40
Screen Auctioning Discussion

• Two alternative explanations for lower prices
• Earlier auctioning time for screen auctioning,
  but this would have led to higher prices.
• Introduction of third auction clock, but the
  increased cognitive complexity would be likely to
  lead to higher prices, given risk-averse buyers.

41
Buying behavior under quality uncertainty

• Behavioral decision theory in the absence of
  salient cues, people rely more on the available
  cues (compensatory decision-making)
• Corollary diameter should become a more
  important factor after screen auctioning
• Pre ?(Diam) 14.094
• Post ?(Diam) 16.214

42
Screen Auctioning Conclusion

• Effects of electronic product representation
  separated from effects of lower search costs.
• Lower prices in electronic markets can partially
  be explained by deficiencies in product
  representation (not just lower search costs) and
  expensive products suffer more.
• Aucnets product representation and quality
  rating system increased prices, so a good product
  representation is essential for success.

43
Third study Buying-At-A-Distance (KOA)

• The first study dealt with difference in product
  representation, but another category of
  differences is relevant
• Market State Information public, non-transaction
  signals that influence trader behavior (adapted
  from CovalShumway, 2001)
• Buzz

44
The KOA initiative

• Electronic bidding at a large Dutch flower
  auction
• Online/KOA-bidders bid on the same clocks as
  offline bidders
• Detailed comparison possible!
• Two categories of KOA-bidders internal (in the
  same building) and external (off-site)

45
(No Transcript)
46
(No Transcript)
47
KOA Bidder differences

• Internal KOA-buyers vs. auction hall-buyers
  lower search costs and lower switching costs.
• External KOA-buyers vs. auction hall-buyers
  lower search costs and lower switching costs,
  less information about product quality and also
  less market state information.
• Internal KOA-buyers vs. external KOA-buyers more
  information about product quality and market
  state.

48
KOA Hypotheses

• H3 Because of lower search costs and lower
  switching costs, KOA-buyers will bid less than
  hall-buyers
• H4a Because of lower search costs and lower
  switching costs, both internal and external KOA
  buyers will bid less than auction hall buyers
• H4b Because of more product quality information
  being available to them, internal KOA buyers will
  bid more than external KOA buyers

49
KOA Model

• Regression model
• PRICE ? ?1DIAM ?2WKDAY ?3VBN
  ?4QUANT ?5,IFLWTYPEi ?6KOAINT ? 7
  KOAEXT ?.
• 81,803 transactions for flower Anthurium
• Sequential regression first the controls, then
  the KOA variable

50
KOA Results

• R2 0.713 after the first step, after addition
  of KOA only marginal, but significant increase.
• KOA-coefficient ?6lt0, in accordance with H3
• H4 KOAINT negative as expected, but KOAEXT
  slightly positive and not significant

51
KOA Discussion

• Two surprises
• External KOA-bidders pay more than internal
  KOA-bidders
• External KOA-bidders pay the same as bidders in
  the auction hall
• Possible explanations
• Bidder heterogeneity is present, but no really
  logical explanation
• Market state information is important,
  particularly regarding number of bidders

52
KOA Limitations

• Explanatory power of KOA for flower buying model
  negligible (but the goal was establishing a
  theoretical effect)
• Causality of market state information is
  inferred, not rigorously controlled for ex ante
  (but laboratory experiments are in preparation)
• Results only for one flower type (but replication
  data is being analyzed currently)

53
Fourth study KOA Bidder Analysis

• Are the differences due to bidder heterogeneity?
• Use screen auctioning dataset to estimate bidder
  differences
• Compare KOAINT and KOAEXT for 1995 (pre-screen
  auctioning) and 1998 (post-KOA)

54
Results KOA Bidder Analysis

• 1995
• ?(KOAINT) -1.65 ?lt0.01
• ?(KOAEXT) 1.109 (but not significant)
• 1998
• ?(KOAINT) -3.608 ?lt0.01
• ?(KOAEXT) -2.767 ?lt0.05
• Future external bidders indistinguishable from
  auction hall bidders, but future internal bidders
  already bid lower than average
• Strong KOA-effect for both types of bidders, even
  more so for the external bidders.
• Lower search and switching costs more salient
  than product quality information and market state
  information

55
Interpretation KOA Bidder Analysis

• Internal KOA bidders were the early adopters and
  they still have the best of both worlds
• But the external KOA bidders (fast followers) are
  catching up
• More KOA-adopters implies more market
  transparency, further lowering prices
• Corroborating evidence influence of VBN prices
• KOAINT, KOAEXT ?(VBN)lt1
• Hall ?(VBN)gt1

56
What about quality information?

• Similar argument as in the screen auctioning
  case the less quality information, the more
  important diameter
• KOAINT ?(Diam)16.803
• KOAEXT ?(Diam)18.749
• Slight spanner in the works ?(Diam)17.954 for
  the auction hall buyers, even though they should
  be closer to the internals than the externals

57
Discussion what about market state information?

• How many people and who exactly are bidding, is
  salient information to bidders, but what if this
  is missing?
• Option 1 Make conservative estimates, which
  would lead to earlier (and higher?) bidding
• Option 2 Wait in the wings, which would lead to
  later (and lower?) bidding
• Option 3 ???

58
Conclusions

• Study 1 Markets are the meeting point for
  multiple stakeholders with conflicting
  incentives. No new IT-based initiative is likely
  to succeed if any powerful stakeholder is worse
  off after the IT-enabled innovation.
• Study 2 Lower prices of electronic markets are
  partly due to lower quality of product
  representation
• Study 234 Different types of information cues
  (product information, market state information)
  in electronic markets lead to subtle changes in
  buying behavior
• Study 34 Lower search and switching costs lead
  to higher market transparency and therefore lower
  prices
• Information architecture of the electronic market
  is important.

59
Look at www.makingmarkets.org
60
And more info

• Otto Koppius,
• Information Architecture
• and Electronic Market Performance,
• PhD thesis, ERIM nr.13, May 2002.
  (www.erim.eur.nl)
• Best PhD Dissertation ICIS 2002 Barcelona

61
Theory of Electronic Markets (Koppius, 2002)