This was a really great team to
work with and they also like

the first two teams had a couple
unique challenges before

they go out and do any of the design and
then the modeling and the analysis.

First they had to go off and
read a bunch of research papers and

learn some stuff that we usually
only teach at the graduate level and

then after that they
had to turn around and

teach all that stuff to clear and
understandable.

U.P.S. kid could think about it and
say yes the way this team is proposing to

go about the project really is the right
way and they should go ahead with it.

So essentially in addition to doing their
own job of working on the project they had

to do my job of teaching stuff and

they they really did a great job
with both enjoy their presentation.

All right.

And also we would like to think
Dr so-called was the semester for

his guidance and support from the very
beginning because it's very rewarding for

us some hope for him as well.

And so just thank you for everything.

This semester.

Good evening everyone.

Welcome.

My name is Kevin kitchens I'm
here tonight with Richard Ward.

Jane intern Katie buckler Patrick O'Neill
and Carl and I kind of ham and

we've been working on a network auction
an assignment for U.P.S. direction.

Direction as a service that's offered
to retailers U.P.S. to deliver goods

from retail distribution centers to
consumers and delivery is guaranteed

in two three business days because
direction utilizes a zone skate model.

Now some of the companies that
used to wreck ship are the gap

footlocker and William.

Sonoma illustrate this service.

Let's say someone here living in
Austin Texas decides to buy a toy online

from Toys R Us the retail distribution
center for Toys R Us in growth per hour.

Heigho.
Takes that package and

puts it on a truckload shipment that
serviced by a third party carrier and

shipped directly to the Texas skipping the
intermediate U.P.S. Habs along the way and

then once arriving in my ski packages
the delivery to the consumer ground and

this full truckload lane here is
an example of the direction network.

There are over fifty seven hundred five
truckloads shipments that are sent

per month over the network from
fifty retail distribution centers to

thirty six over two hundred eighty
lanes in that network is still growing.

Each of these lanes is a road to a carrier
through a U.P.S. auction process.

First you request to the carriers
the carriers then respond back

with the bid price that they would charge
for each track road on each lane and

then U.P.S. takes the racks and notifies
the carriers of their winning assignment.

The last time that you POS auction to
the direction network was in March of two

thousand and seven but by November of last
year they were already dissatisfied with

the network set up so
our project this semester focused on

a certain design strategies for
Steps one and three the auction process.

For that March two thousand and seven
auction assignment there were over fifty

six hundred received from one hundred
forty four carriers target rate

was created based on a function
of the average bids proclaim.

And then that target rate was
offered to the incumbent carriers.

Now as a result of placing a high
emphasis on incumbency and also low cost

thirty six carriers were contracts and
twenty six Of these only serviced one or

two right and there have been several
observed problems since March two.

Thousand and seven one of which is
an satisfactory service level for

several of the large retailers and
also that it was difficult to manage such

a large number of carriers for
just one network.

Here we can see eleven of the top
twenty five retailers for direction.

That did not receive ninety eight
percent on time service from

April to December of two thousand and
seven.

It's also important to
note here that U.P.S.

pays a penalty cost for each package.

Even though the carriers
are Spears responsible for that.

So with these observed problems U.P.S.
is reassigning the network

here in two thousand and eight with a new
and different strategy including using

a fewer number of carriers and assigning
each of those carriers as an equal portion

of the network as possible based on
revenue and mileage and then from there.

U.P.S. is also looking to provide a ninety
eight percent on time service level for

the largest twenty five
retailers in direct ship.

Our project this semester to help
implement the strategy that first

with recommending a carbon Tauriel
auction format from there.

Creating an energy program to solve the
winner determination problem while also

meeting with desired network
properties of U.P.S. and then.

Building a user interface to allow us to
use this for both the current auction and

also for future auctions.

Our first recommendation was for

you to use a comment or your auction and
for comments where your auction carriers

can bid both on combinations of lines
as well as individual lanes and

that can be illustrated here let's
say a carrier service from L.A.

to Salt Lake City and so
it would end up in Salt Lake City but

to be able to get that truck back there's
a possibility that they would have to pay

for Deadhead miles to get
the truck back to Los Angeles.

Now if we bring Dallas into the picture.

If they Kiri or was to only when
the lane from Salt Lake City to.

Dallas.

It wouldn't be beneficial
because they would still

have to pay Deadhead mileage
from Dallas back to Los Angeles.

However if they did comment internally
both the Salt Lake City to Dallas and

the Dallas to Los Angeles lanes they would
be able to increase their business while

also decreasing the amount of Deadhead
miles that they would have to pay and

from research.

Dr Saka mentioned that we had
to do in the very beginning and

several auctions conducted by
the Home Depot and also Sears there is

an expected three to fifteen percent cost
savings by using a comment or action.

However I also like you said again when
we first presented this idea to you.

P.S. They were a little skeptical and

illustrate some of the commentary option
properties we created two different games.

The first of which U.P.S.
employees acted as

carriers and they were able to
bid on the smart network here for

both individual lanes and combinations
of lanes and this showed both some of

the Common to auction properties as far
as how bidding would work and also.

How a commentary auction would yield.

Lower prices for the second game as
employees were given some market data.

There were seven lanes in the network and
twenty two bids in packages and

only three of the eleven U.P.S.

employees were able to identify the lowest
cost assignment for the seven lanes and

this also showed the complexity of
the winner determination problem.

And also the need for an energy program to
be able to choose the winning carriers.

So why

it raw and enter a program is necessary
to solve the winner determination problem

of the Common to toil auction of energy a
program becomes a little more complicated

when you're meeting the.

Need to Zire network properties that
U.P.S. is defined as you might imagine

the objective of our entire program is to
minimize the expected network cost but

some of the constraints are toward
each line to a carrier to me

an expected on time service performance
level based on past performance and

also to assign each carrier
an equal portion of the network.

As was mentioned with
respect to Miles in revenue.

We took the March two thousand and
seven bit data.

And did an analysis with our
energy program to show you.

P.S. What type of problems can be solved
and what kind of questions you can ask.

And what you should look at to
determine which carriers which lanes.

Some of the things to consider how many
carriers what service level you desire

the level of balance rather
carriers receive very close closely

size assignments or they can vary quite
a bit and then finally where to negotiate.

In the auction process before you P.S.
signed contracts

with the carrier there's an intense
negotiation process that takes place.

So this auction kind of different from
other auctions where the best price is

always the winner that's not the case
here taking the March two thousand and

seven bit data we asked a couple of these
questions we want to see how cost is

affected by the number of carriers used
as you can see you can see the trend of

how cost decreases as the number of
carriers increases and also you can see

a trend of how when you influence more
tightly balanced assignments cost also

goes up this is for the blue in the the
yellow show the balanced constraints and

cost also goes up as you and force or
better as expected on time service.

Using that March two thousand and
seven data and

going through lots of different scenarios
we made one recommendation the U.P.S.

used eight carriers with
balanced assignments and

the network had a ninety eight
percent expected on time perform.

The price of this network was one
point three million dollars cheaper

than the assignment that U.P.S.
chose back in March which place and

insists on the incumbent carriers and that
that sign as you can see up top used for

thirty six carriers and
the assignments were not balanced.

Thank you.

And another thing because of that
negotiation process another thing our tool

can be used for is to identify
lines that are good to negotiate.

If you look this first as an example
a bit by carrier US excise

did not eight dollars
behind the winning bid.

But because of the network constraints.

It would enter the solution if that bid
were reduced by sixty two cents and

that's what we call the critical value.

Now what's something interesting to
know is that Rob That's an eight dollar

difference the sixty two cents
is all it would require but

also as you can see in the bottom row.

Sometimes the little lower
price bids did not win for

example with carrier TECO or they had
a poor performance so by accepting this

bid another more expensive bid somewhere
else and that would have to be accepted.

So you can see that their bid actually
beat the winning bid by twelve dollars but

it would still need to be reduced a dollar
twenty nine to enter the solution.

This type of analysis can be done to
identify where to negotiate lanes

while still reassigning
the network in a balanced way.

Since P.S.
is not familiar with energy or programming

rights familiar with initial programming
or express which we implemented IP and

we decided to create a Java program which
would automate the process and allow

the user not to have to even touch express
the way I do it works is it takes excel.

It takes big data from Excel and

is loaded into a Java program
Java program then formulates and

into a program and calls express to
solve it expressed sense there is.

It's back to Java which
I displayed graphically.

Now the first thing a user needs
to do when doing this type

of analysis is to set a constraint
set up the input parameters and

the upper left hand corner you can see
the user will set the number of cares.

He or she wants to use
around with the mileage and

balance constraints an expected
service level next user can add more

specific constraints such as to first
step to care to be used to force or

accept a bid to be accepted or
rejected bid and

also those two may be more subtle or
more interesting.

Constraints of the user can add the first
one is they can exempt a carrier

from having to satisfy
the balance constraints.

One thing U.P.S.
said is that maybe in the future but

want to give a very large portion of
the network to run superior carrier

in that case for example Swift may receive
say up to fifty percent of the network

which would throw off
the balance constraints and

setting in this constraint the image
of program will allow such a scenario.

And finally this accept what
we mean by that is if you P.S.

want to accept a specialty bid they
can accept a bit on the one or

two lanes and still keep
the rest of the network balanced.

This might be useful for example if
there's a strong regional carrier or

if there's a carrier who excels in kind of
a specialty service such as refrigerated

delivery on the left side a user can
define the name of the specific scenario.

And then would hit the run button
to display the results and

example the results page at the top the
blue dot represents our specific scenario

with eight carriers and

twenty percent mileage constrained ninety
eight percent expect the service you can

compare that to the series of red dots
which represents the low cost solution.

That's the price you pay if you have
no other constraint than the number.

Care is when the pie charts show how
balanced the assignments are with returns

in terms of mileage and revenue and

down below list the specific
constraints what defined that scenario.

At any time the user can hit the rain
assignments and display a spreadsheet

listing of the assignment on each line for
which carrier has run that lame.

Now the real value in this in this
program is that it allows you P.S.

to host an auction and solve that when
a determination problem meeting their

desired network property it's just to
review our project to run the first

things we did was to recommend a common to
toil auction and we formulate an injured

program and wrote the computer application
to solve the winner determination problem.

We went through the additional
constraints that the user can add and

how can they use in
the negotiation process and

the expected benefits as was mentioned
the solution that we came up with

was one point three million dollars per
year better than the solution U.P.S.

chose in March two thousand and seven and
it met the balance constraints and

ninety eight percent expected
on time service level and

it is just a carriers which is a number
that U.P.S. in the case it was desirable

finally these common authorial bidding
can be used in future auctions both for

the direction service and
another freight shipping services.

Thank you.
Are there any questions.

Sure.

Actually that was.

I guess one of the more difficult task
that required most of the work you.

Only had individual Excel
data sheets that listed here

your server here your failures
in the your successes but

it was sort of based on
the individual retailers.

So we had to compile all of that data and
we actually created a relational database.

To be able to sort that by carrier and
come for all on time performance level

by carrier and
it's also important to note for the for

the model that that on time performance
level by carrier is included.

But if we didn't have any data for
a carrier because

they hadn't been utilizing the direction
network before U.P.S. actually gave us

a list of carriers that they would want to
work with and so it is included in the in

the program that those curious can
be involved in the assignments and

without just the ship the past performance
was used to say if a carrier in the past

serviced ninety percent and that was the
value that was used for what they do for

future service or

one of the first things they said
early on when they said fewer

carriers was that perhaps eight
to ten would be desirable.

So as you saw with that trend how price
actually goes down as use more carriers.

We used eight to kind of be on the high
side of what they initially said

a beta ten carriers were

more.

Yeah it's a good question and
the other question.

Thank you.