Title: Newsvendors Tackle the Newsvendor Problem
1- Newsvendors Tackle the Newsvendor Problem
-
- Martin A.
Koschat -
- AOL Time Warner
- Syracuse University, SOM
- Faculty Seminar
- March 26, 2001
-
2 Time Inc., Retail Sales and Marketing
3 The Business According to AA
4 The Business According to MAK
- The ratio of advertising revenue vs. circulation
revenue varies over time. - Newsstand revenue accounts for 25 of Time Inc.s
circulation revenue. - Newsstand profits account for 33 of Time Inc.s
circulation profits.
5 The (classical) Newsboy Problem
- Optimal draw allocation depends on
- Marginal cost (m)
- Revenue per copy sold (r)
- Distribution of demand (f)
6 The Principle of Marginal Efficiency
- Consider a representative sales pattern for one
store for 12 issues. - Assume that for all 12 issues draw was equal to
10 copies. - In summary, there is a total draw of 120 copies
and total sales of 54 copies for an overall
efficiency of 45.
7 The Principle of Marginal Efficiency
- Rank order issues by sales.
- Shipping 10 copies per issue instead of 9 copies
resulted in an overall draw increase of 12
copies, but resulted in an overall sales increase
of only 1 copy. - The marginal efficiency at a draw of 9 is
therefore ?S/?D
1/12 8. - The marginal efficiency also equals the sell-out
probability.
8 Store-level Allocation
- Assume that a wholesaler has received D copies
that need to be distributed across N stores with
respective demand distributions Fi. - The profit maximizing draw allocation for store i
is - di Fi-1(?)
- where ? has to be chosen such that ?di
D. - In practice, the allocation is made via a formula
file. - If the formula file is chosen properly, this
procedure provides an allocation that is close to
optimal.
9 Store-level Allocation cont.
10 Wholesaler Allocation
- Assume that the national print order for an issue
has been determined to be D. - D is optimally allocated across N wholesalers, if
each wholesaler has the same marginal efficiency. - Since the marginal efficiency is not easily
determined, the gross efficiency is used as a
proxy.
11 Wholesaler Allocation Estimating Marginal
Efficiency
- One can conduct a market experiment.
- If store level draw and sales data are available,
one can conduct a pseudo market experiment.
12 Wholesaler Allocation Marginal vs. Gross
Efficiency
- For 40 wholesalers store level data were
available. - The marginal efficiencies were estimated by a
pseudo-market experiment.
13 National Print Order
- The national print order is determined with
consideration of the marginal efficiency. - In addition, there are other business
considerations that affect the size of the
national print order.
14 Distribution of demand - Parameters
- Search within parametric classes of distribution
- Frequently used parameters -location and
dispersion - Explore relation between location and dispersion
15 Distribution of demand - Parameters
- B-FM B-SV SD-FM SD-SV
- Sports Ill. .11 .06 .96
1.10 - People -.03
-.03 .68 .71 - Time .01 -.14 1.22
1.42 - EW .26 .23
1.06 1.15 - Estimated bias and standard deviation for the
median based estimator (FM) and the sample
variance (SV). - Comparison is on a log-scale for a
pr-period covering the first half of 1997 and a
post-period covering the - second half of 1998. Comparison involves
raw data for 18,000 stores.
16 Distribution of demand - Shape
- Shape of the distribution
- Theoretical and practical research suggests the
use of power transformations - Fourth root transformation transforms data to
near Normality
17 Distribution of demand - Censoring
18 Distribution of demand - Summary
- If the fraction of censored observations is not
too large, the median and the inter-quartile
range of the distribution of sale is also the
median and inter-quartile range of the
distribution of demand.
- Thus, in most instances the median of sale
provides a comprehensive summary of the
distribution of demand.
19 Distribution of demand - Implications
- The optimal draw allocation at the store level is
therefore a function of the median of sale - in
practice, a function of the historical median of
sale for a particular time period. - If the store-level distributions of demand are
known, expected sales under different draw
allocation procedures can be calculated. - Different procedures can be compared.
- In particular, approximate procedures can be
compared with the optimal procedure.
20 Results Sales Lift
Percent Sales Due to Improved Due to
Improved Lift Formula File
Wholesaler Alloc. People 1.38
0.58 0.80 Sports Illustrated
1.26 0.56 0.70 Entertainment
Weekly 1.22 0.62
0.60 Time 1.10 0.41
0.40 Money 1.05 0.79
0.26 Fortune 0.94 0.30
0.64 Parenting 0.82 0.26
0.56 InStyle 0.73 0.44
0.29 Life 0.69 0.39
0.30 Sunset 0.49 0.26
0.23 Teen People 0.47 0.30
0.17
21 The Channel Problem
- A producer and a distributor share the revenue r
of one unit sold that is rp rp r. - For the producer and distributor the respective
marginal cost per unit are cp and cp. The cost
of producing and distributing one unit is cp cp
c.
22 The Channel Problem
- The optimal production level Dp for the producer
is - Dp H-1 (cp /rp)
- The optimal distribution level Dd for the
distributor is - Dd H-1 (cd /rd)
- The actual production/distribution level D
satisfies - min( Dp, Dd ) ? D ? max( Dp, Dd )
23 The Channel Problem
- The channels profits are maximized if
- Dp Dd
- cp /rp cd /rd
- The channels profits are maximized if the
revenues are split in proportion to the marginal
costs. - The channel member with the smaller marginal cost
receives a smaller share of the revenue.
24 The Channel Problem
- An alternative way of maximizing the channels
profits is through a per unit transfer payment t
either to or from the producer to the distributor - - cp lt t lt cd
- The effective net cost for the producer and the
distributor are, respectively, - cp cp t
- cd cd - t
25 The Channel Problem
- The magnitude of the transfer payment determines
the revenue split between producer and
distributor. - The channel member that carries the larger share
of cost reaps corresponding rewards.