E-commerce and Last Mile Distribution – Mini Lecture by Professor Miguel Jaller

E-commerce and Last Mile Distribution – Mini Lecture by Professor Miguel Jaller


[Intro music] Hello, my name’s Miguel Jaller, I’m a faculty
here at the UC Davis Institute of Transportation Studies, and today I’m going to tell you a
bit about the work we’re doing in terms of the impact of E-commerce in last mile distribution.
So as you know last mile distribution is part of the freight system and it’s involved in the
last stages of the supply chain, the last stages of the distribution part of any good
or service that goes into commercial establishments or residential areas. So you can see in the
slide we have kind of the traditional typography or topology of the supply chain where you
have shippers, where you have distribution centers, sortation centers, and then they
go from the sortation centers, distributions centers, all the packages and goods go to
the consumers or the commercial establishments. What happens is that in many cases there are
some returns that we call reverse logistics that then people or consumers or the establishments
then send back to the shipper or the company. This can be done directly from the person
going to a drop off facility, or having the vehicle come to the location and pick up the
product. So all of this is kind of a traditional supply chain, and now in the age of the E-commerce
there are more and more of these products and packages going to residential locations.
So the focus of this talk is about this last part. When products leave a warehouse or distribution
center and they go to their residential location or a commercial location for that matter.
what are the main important aspects of this? Well, E-commerce has been growing since 2008,
2009, at a rate of a bout i don’t know, 10 double digits every year. Still today, E-commerce
sales in the retail industry are only represent about 1% of retail sales in the U.S., so even
though it’s still a small proportion of all the sales that are seeing in the retail sector,
all of the issues associated with the large growth of these residential deliveries we
are seeing in every single city. So what are those problems? Well, we’ll discuss later,
but just to give you a sense, today, about 55% of the U.S. population shops online, however,
about 80% of all the shopping is affected in some way or another by E-commerce. In one
particular day about 2-3% of individuals shop online whereas 40% will shop in a store or
not online. So again, it has been growing a lot, about half of the population shops
online, most of the shopping is affected by E-commerce either in the searching, the return,
or in the comparison aspect of the shopping behavior, and this is affecting the way people
travel, this affects the way companies have been locating their facilities, the warehouses,
the distribution centers, how they manage inventories, how much products they carry
in a store versus how much they carry in their website or on their online platform. All of
this is what is now known as commissional distribution, before there was almost single
channel, either you were only online, or only in a store, now companies are offering all
these multiple channels, so that’s affecting the choices consumers have and therefore it’s
affecting what people choose to do when they are, ah, when they want to buy something.
And buying is not only the act of purchasing products, it involves searching, it involves
the comparison, it involves then, what happens after you buy the products, sometimes you
keep it, sometimes you return it. So all these phases of the purchasing process are affecting,
also, travel, and it’s affecting the last mile distribution. So, what is, why is this
a problem? Well, because it has been rising SO fast that then there are many more trucks
and many more commercial vehicles, coming into locations that they didn’t used to come
before. So this is creating more congestion, more emissions, more energy is being consumed,
there are more conflicts with pedestrians and other vehicles, there is a fight for curb
space, there is fight for these limited resources inside the cities. So there are many, many
more problems. So what do we need to do? Well, we need to start studying it, or we need to
go deeper into our analysis on what are the actual impacts of E-Commerce in last mile
distribution? Or in the urban settings? However, this is a very complex, complex, problem.
There are many aspects that affect the sustainability of E-Commerce, the sustainability of shopping
patterns and shopping behaviors, and that’s what I’m going to be talking about for the
rest of this lecture. So, the impacts of online shopping depend on first the consumer side,
the demand side, and then the supply side. Why am I saying that there are these two components?
Well first on the demand side, it depends what consumers are doing in the light of E-commerce.
People can substitute what they were doing before, and now shop everything online, or
they can do both, they may have some purchases in a store, and some purchases online, or
because it’s so convenient now and you have so many other products that were not available
to your market, now there many be some induced demand in that sense. In addition to that,
because it’s so convenient, because it’s so cheap, almost free for you to get any product
online, then now, you, some people are ordering many products that they don’t need, and then
they return the one that didn’t fit, that they didn’t like. Now on the supply side,
you have all the things the supply companies are doing to try to tackle this challenge.
As I mentioned before, some of these companies have been relocating their facilities or opening
more facilities near the the consumer to be able to offer faster and more reliable deliveries,
but also, they are using different types of products, different types of modes, different
types of strategies to distribute. Sp these have two different aspects that are making
the system either sustainable or not sustainable, and those are the things that we need to study.
So what have we been doing? Well we have been analyzing these two things. So in this work
that I’m going to be showing you in the next few slides, we have been following, kind of
the methodology. First we estimate the demand side by estimating behavioral models of shopping
behavior using the American Time use Travel Survey. Then we analyze how people travel
for shopping, are they going from their home or work to the store, and then coming back,
or are they doing the shopping activities as part of a tour in a daily, doing their
daily activities. So we’re trying to analyze that, and we use data from the American Time
Use Travel Survey and also from the National Household Travel Survey. Then we need to analyze
how companies are delivering. So we got some data, that we will explain later, to analyze
those delivery patterns from vehicles, and what is the efficiency? Then we compared the
two, we compared what is better, going shopping, or having the product come to you? Then we
are analyzing two different things, kind of going deeper into more analysis. The first
one is, what are the impacts of the really fast deliveries? Companies like Amazon are
offering two day, one day, same day, one hour, two hour deliveries. What is the impact of
those service levels into sustainability of online commerce. So in terms of shopping demand,
as I mentioned, we estimated behavioral models, using the American Time Use Travel Survey.
So the ATUS is a survey, it’s a one-day diary, there are about 10,500 individuals, that’s
the counted sample size, and they look at all of the activities, the location, and the
timing of those activities during a 24 hour period. So we can use the data to analyze
what they do in a day. So as you can see in the slide, there are different types of activity
names, and the locations of those activities. So we then first cluster analyze which activities
related to shopping. Can we say they were done in a store as opposed to being done online?
And then we work through the data to have the online, the in store, and people who may
have done both. Then we develop behavioral models the estimated and multinomial logit
choice model where we have basically four choices in a day, either you don’t shop in
a day, you shop in a store exclusively, you shop online exclusively, or you may be doing
both things during the typical day. And these are the results of the model. We tested different
characteristics of an individual, so the socioeconomic and demographic characteristics. We also tested
variables related to where they are, we also tested variables related to the season of
when the shopping is done, and as you can see there are differences between the characteristics
of the people that shop online versus in store and people that buy both. n general, income,
location, people in the western large cities shop more online than people in the rest of
the country. People in the southern parts are more in store, people in cities that are
larger, are more than 1 million, shop overall more than other individuals. Females also
more online than males, and there are other factors that kind of hinder the probability
of somebody shopping in a day, in either one of the channels. So we are using these models
now to estimate what is the propensity of shopping for any individual in many city throughout
the U.S. The next part is, ok we know if somebody is shopping, but we want to explain now, how
they are shopping. Either it’s in store, we have to now identify, are they going just
to the store? Or are they going to different places and one of the places that they are
going is to do the shopping? So we got data from ATUS and from the National Household
Travel Survey to come up with some distribution functions. So what we see int he slide is
that first, on the top lef, we have these distribution of the number of shopping tours
that somebody will do in a day, and when I’m talking about shopping tours I’m actually
predicting any tour that has at least one activity or a stop within the tour. The next
one, part B , is the number of stops people do in a shopping tour. Mostly one, two, or
three are the common, are the majority of all the tours. However, there may be people
that do up to ten stops in a tour, in a day. However the probability of this is quite low.
Now we also have, what is the approximate distance, or travel distance of these tours
depending on the number of stop that people are going to do in one single tour. And then
finally, what’s important for analysis is, if you have a tour, then what is the percentage
of all the activity that you are doing in a tour, that are related to shopping? Remember,
shopping’s not only the shopping but is also the searching or comparison as disclosed in
the ATUS. Now we move to the demand side, as I mentioned we got data from some of the
delivery GPS, from data from NREL. And we analyzed the data for different last mile
delivery locations. As you can see in the slide we have beverage delivery, warehouse
delivery, parcel, linen, food, and local delivery. We are going to concentrate on parcel because
that’s the majority of the E-commerce that goes through residential locations. As you
can see, most of the routes are about 50 miles, around the average, but what we found is that
about 95% of all the routes that we got in the sample were less than 100 miles. So we
create product distribution functions for this distribution route that we’re going to
be using, and this is an example. We fit a Weibell distribution, and then we also have
another important aspect of last mile distribution is, how many stops, or how many deliveries
can you do along a route? And deliveries and stops are different because the driver may
stop, but also deliver to one, two, three, or four customers for every single stop, but
eh stops can give you a kind of rough idea of the consolidation that they can do along
a route. To give you perspective the National Average is about 75 stops in a tour, however
when you go to companies like DHL, FedEx, or UPS, they can do 100-102, and if you look
into packages, one example on the higher side could be the Amazon vans, the blue vans that
you see in some of your cities, they can handle about 200 packages in one delivery route.
So we now have the behavioral models for estimating if somebody is shopping, if it’s shopping
online, in a store, now we can estimate how they are traveling for shopping, and then
we can also estimate how, and how much milage is being traveled by the delivery vehicle
that is bring around the products to your house or your choice location. Then we can
do some of the analysis, now we v=can go in and compare the two. What is the efficiency
of going to the store versus the efficiency of having products delivered to you? So we
are, because we are, we need to make some additional assumptions, we are assuming that
everybody that is going to a store is using their regular car, everybody that’s getting
a delivery, that delivery is coming into a regular delivery truck, kind of a class 5,
and then you have in this slide is kind of the emission factors for CO, NOx, CO2, particulate
matter, and so on, for the two types based on the impact database from ARB. And this
kind of simulation analysis, again what we do is, we go to the study area, in this case
let’s say San Francisco or the City of Dallas, or the County of San Francisco and the City
of Dallas, and we estimate, we generate the synthetic population based on the census data
about each individual, and then we apply the models directly to each of those individuals
and using Monte Carlo, we can estimate the probabilities if they shop or not. After we
know who is shopping, then we estimate how much they are traveling, what is the VMT,
and then based on the VMT, we can use the rate, emissions rate, and estimate all the
environmental impacts from their travel. Similarly, for the people that are shopping online, now
we can estimate how much travel is generated by the delivery trucks, and all the emission
factors. So what you can see in this table is that there is not much difference between
Dallas and San Francisco, there is about a 10% difference in most of the different assumptions.
However, what is important I’ll show you in the next few slides, is that today when comparing
Omni-Channel that means what we are doing today, some people are doing both, some people
are only doing one of the channels, if we compare that to the case where everybody is
just shopping in a store, there is not much of a difference. When we get 5% reduction
here, 5% reduction there, we get some increases especially in NOx because commercial delivery
vehicles generate more NOx than the passenger vehicles, but in general we are not seeing
that much of a difference, from this static analysis. As I’ll show you later, this is
hanging because of the decisions that either individuals are doing or the companies are
doing for the delivery patterns. However, when we were, if we are able to substitute
all the travel related to shopping, so nobody else, so nobody is going to stores anymore,
and everything that they buy is online, that is what we are comparing in these slides.
It’s kind of the case of everything online versus the case of everything in a store.
As you can see, now there are potential savings is we are able to consolidate. We don’t think
that’s going to happen, because people are going to still going to go to locations, they
are still going to generate some travel, but anyway there is going to be substantial reductions
if we are able to consolidate, in the trucks, and where we are shopping. However, we have
to be careful of the NOx because, still we get about a 25% increase in Nox from these
kind of wall to wall scenario. However, there is a caveat. When I was showing the results
for the Omni-channel to all in a store, that’s assuming that we are able to consolidate shipments
in the truck. What is happening today? Well, we have two things. One is what we call the
basket size, and one is what we call how much consolidation you can do because if you only
have one hour, two hours, to distribute, there is not that many products you can put in a
truck to be able to distribute in a city. So what we did her is, well, let’s see what
will be the break even point between everything online versus everything in-store under some
assumptions. The first type of assumptions are, what is the basket size, and when I talk
about basket size, it’s how many products are you buying in every single purchase? On
average, people buy 2-3 products every time they go to a store. However, we only buy about
1.1 products every time we shop online. So there is kind of a replacement rate, if you
are buying in a store 3 products, that will mean like 3 online purchases, are we doing
those 3 online purchases one after the other? Or are we waiting a little bit so they become
completely different purchases that will come in three different trucks to you house? So
what we find in this slide is, for the case of the basket size where there is one to one
relationship in terms of basket size and when there is a 2.5 to 1 relationship in basket
size we see that in the worst case related to NOx we need about 40 delivered to have
to be made in every single delivery route in order to be compatible to people going
to the stores, however when you go to the 2.5 to 1, that actually jumps to about 90,
and for the large companies that are doing consolidated deliveries it may be a good number,
but for the average int he U.S. this is highly, this is really high. Now the other thing we
wanted to try is, ok what happens when companies want to offer the one hour, two hours I was
mentioning? In that time frame there is not time to consolidate. So for the same amount
of products that are gonna be in a city, now a company will have to send more and more
vehicles with less products to be able to distribute in the one hour, two hour. And
as you can see here is on the X axis you have the number of deliveries or stops in a delivery
tour, and on the Y axis is the average distance between stops. And s you reduce the number
of stops per delivery, then the average distance of these or the average distance impact per
delivery increases exponentially, so we need to be careful on all these services. So finally,
what have we found about these works? First, there are differences in terms of demographics,
and regional differences about shopping behavior. Females, higher propensity of online shopping,
people in large cities, especially in the west, shop online more, and general there
is overall complementarity, and that’s just kind of a limitation of the data we are using,
because we are not able to analyze kind of at a commodity or per commodity base in terms
of what the shopping choices are. Why is this important? Well let’s say a lot of people
have been able to substituteI don’t, know, electronics, we buy most of our electronics
online, and we are not buying them in-store. However we still go to stores to buy another
type of product. On average we are doing both, but maybe we are able to substitute part of
out deliveries or our shopping. Substitution effect, as I I was mentioning, have to be
done at the commodity level, shopping externalities, well we have not analyzed upstream effects.
What is the impact of a facility location? What is the impact of the packaging, what
is the impact of for online shopping, all the energy that has to be used for all the
warehouses for the servers and all the information and so on. And then some things that is important
that I mentioned before, basket size, rush hour deliveries, and levels of consolidation.
They can make or break the sustainability of online shopping as opposed to everybody
going to stores and so on. So, thank you, I hope you enjoyed the talk and we can have
questions later.

Leave a Reply

Your email address will not be published. Required fields are marked *