Vesa
Kämäräinen, Juha Saranen, Jan Holmström
|
The Authors |
Vesa
Kämäräinen, Helsinki University of Technology, Finland
Juha
Saranen, Helsinki University of Technology, Finland
Jan
Holmström, Helsinki University of Technology, Finland
International
Journal of Physical Distribution & Logistics Management
Volume 31 Number 6 2001 pp. 414-426
Copyright © MCB University Press ISSN 0960-0035
Home delivery is not a new service in the grocery business.
In fact, the Internet has just helped to reintroduce an old service. The
Internet has had an enormous effect on the ordering process, making it cheaper
and faster. However, the delivery process has not experienced a similar
efficiency improvement.
In spite of current poor operational performance, it is
forecasted that the e-grocery business will grow quickly in the future (Andersen
Consulting, 1998; Mullaney and Leonhardt, 1999; Cuglielmo, 2000). However, to
grow quickly, the e-grocers have to lower the high consumer prices. At the
moment, prices are often higher than in traditional supermarkets. Cutting prices
is only possible if the e-grocers manage to cut operational costs. Currently,
the worst problems are slow picking and inefficient home delivery.
First, it is possible to reach considerably lower operational
costs by cutting picking costs. Better picking efficiency can be reached by
picking the items from distribution centers instead the existing supermarket.
Second, inefficient home delivery increases operational costs
in the e-grocery business. Inefficient home delivery limits growth because of
the long delivery time per customer. However, we can reduce delivery time by
using different solutions for goods receipt at the households. Some e-grocers,
such as Streamline[1] in the USA and SOK[2] in Finland, use reception boxes to
cut delivery costs. The groceries are delivered to locked refrigerated boxes,
which are located in the households. By using reception boxes, customers become
independent of the delivery timetable, i.e. they do not need to be at home to
receive the goods. The reception boxes also reduce home delivery costs for the
e-grocer.
In this paper, we examine different solutions for goods
receipt. We examine the service levels they offer to consumers. In addition, we
compare two different receiving alternatives - a reception box in the household
and attended reception - by simulating delivery times and home delivery
efficiency. As a result, we will observe a significant cost difference between
these two receiving alternatives.
From the consumer's point of view, most e-grocers seem to
operate according to the same business model; ordering is done using the
Internet and the items are delivered to customers' houses. However, if we
examine electronic grocery shopping more closely we can see two totally
different business models (Dagher, 1998; Heikkilä et al., 1998;
Holmström et al., 1999; Kämäräinen, 2000).
The first alternative is to operate as an intermediary in the
supply chain by picking groceries from a supermarket or "cash and
carry" and delivering these to the households (Figure
1). This is how, for example, Peapod[3] in the USA and Y-halli[4] in Finland
operate. It is the most common e-grocery business model today.
This model is based on occasional deliveries. Customers do
not usually plan ahead what they want to have delivered, but instead order when
they notice the need for groceries. Therefore, they want their groceries
immediately and demand quick home delivery. This requires short response times
from the e-grocer in the form of short picking times and tight delivery windows
that can make the whole supply chain inefficient.
Occasional purchases and short response times cause capacity
problems, especially in the home delivery function, because it is impossible to
forecast customers' demand. However, this concept is easy to implement when
sales volumes are low and electronic grocery shopping is only a value-added
service for the supermarkets. It is not very cost effective, but it is a fast
way to increase sales volumes by moving quickly into new areas (Holmström et
al., 1999). On the other hand, when volumes increase, a more effective way
to operate is needed.
When sales volumes are sufficiently large, a totally new
direct channel between the producers and the consumers can be created. In this
business model, the e-grocer purchases items straight from the producers or
importers, stocks products in a local distribution center and delivers directly
to the consumers (Figure
2). This is how, for example, Streamline[1] and Webvan[5] in the USA and
Matomera[6] in Sweden operate.
In this model, picking and packing is more efficient than
when operating from a conventional store, because picking speed increases when
the operations are especially designed to serve home delivery (Holmström et
al., 1999). However, investments are higher in this model. For example,
Webvan's highly automated distribution centers cost between $25 million and $35
million apiece (Cuglielmo, 2000). Competitors who use less automation have costs
of between $4 million and $6 million per distribution center, while supermarkets
can offer home delivery service without large additional investments.
Until lately, the e-grocers have operated as intermediaries
in the supply chain (Holmström et al., 1999). This service concept has
worked well as a value-added service for supermarkets. However, the increased
amount of e-grocery customers has forced retailers to look for more efficient
solutions. When an increasing number of customers require better services and
lower prices, picking needs to be done from distribution centers, not from
supermarkets. In addition, more flexible ways of receiving the goods at the
consumer end are needed for lowering the delivery costs.
Usually, the reception has to be attended when the e-grocer
delivers the products. This means that the customer has to be at home to receive
the delivery. This restricts the number of potential customers. Therefore, more
alternatives for receiving the groceries are needed. We have examined different
alternatives and identified four potential receiving concepts for electronic
grocery shopping:
(1) Pick-up by the customer: customers pick up groceries from
the local supermarket or a warehouse.
(2) Shared reception boxes: there is a shared pick-up point
near the consumer.
(3) Own reception box in the household; orders are delivered
to the household's reception box.
(4) Attended reception: the shopkeeper or a third party
delivers groceries to a given place where the customer accepts the delivery.
In the first approach, the e-grocer only offers picking
services. Shopping is easier with this approach but pick-up from the store does
not eliminate the customers' troubles completely, because they still have to
travel to and from the store. This means that the service level is far below
that of the other reception concepts. However, in some cases this service level
is good enough. This is the easiest and cheapest way for the e-grocer to start.
In the second concept, shared reception boxes are used. The
customers are independent of the delivery timetable, i.e. they do not have to be
at home to receive the goods. The groceries are delivered to locked reception
boxes thatare allocated to a specific customer with every delivery, which makes
it possible for many customers to use the same reception box. The customer
receives the number of the box and the code needed to unlock the box, for
example to his mobile phone by text message.
The reception box operates in many temperatures and keeps the
groceries in good condition for as long as a normal refrigerator would. By using
shared reception boxes, it is possible to drop off many orders at one stop and
reduce the delivery time per customer. This is the most cost-effective solution
for the e-grocer, when delivery is required. However, the service level
perceived by the customer only improves significantly if the reception boxes are
located near the household. Furthermore, before the delivery savings are
achieved, the e-grocer, the customer or a third party has to invest in the
reception boxes. Currently, there are no e-grocers who use this concept, but its
future potential is huge.
In concept three, every household has its own reception box.
This is convenient for the customers because they get the groceries straight to
the home. The customers are also independent of the delivery timetable. In
addition, the reception box has a significant impact on home delivery efficiency
(Feare, 1999). The delivery time is not as short as with shared boxes because it
is possible to drop only one order per stop. However, the delivery time is
considerably shorter than if the reception were attended, as we will show later
in this paper.
Also in this alternative, someone has to invest in the
reception boxes. Currently, there are many different kinds of reception boxes
entering the market (Bennett, 2000). The costs vary significantly (estimated up
to e2000 per box), depending on the type of the box, i.e. the location, size,
different temperature zones, electronics, etc. of the box. The owner of the
boxes can be the e-grocer who buys the boxes and rents them to the customers.
This is how, for example, Streamline[1] in the USA and SOK[2] in Finland
operate. The customer could also buy his/her own box or there could be third
party financing the boxes.
In the last alternative, no investment in reception
technology is required. This is why attended reception is the most common
receiving concept today. However, tight delivery windows and personal contact
with the customer increase delivery time. The long delivery time also increases
the costs to the e-grocer. On the other hand, attended reception offers an
opportunity to offer face-to-face service to demanding customers.
All four alternatives are potential service models in the
e-grocery supply chain. However, only in two of them - a reception box in the
household and attended reception - are the items delivered to the final
consuming point. These two alternatives are the focus of this paper. Next we
will examine the service levels of these reception concepts. Delivery time and
cost simulations for home delivery with and without reception boxes (concepts
three and four) are also presented.
Both attended reception and delivery to the household's own
reception box offer a good service level, but in different ways. Personal
contact with the delivery personnel is important for some customers, while
others appreciate being able to pick up their groceries from their own reception
box at a time of their choice.
The reception box model also offers new service opportunities
(Småros andHolmström, 2000). For example, Streamline[1] offers its customers a
replenishment service where regularly needed products are automatically
delivered, e.g. once a week. In addition to groceries, Streamline offers, among
other things, dry cleaning, videos and film processing.
In Table
I we compare service levels of attended reception and the reception box
concept. We can see how both concepts provide good service depending on the
delivery targets and the customer's personal preferences.
From the customer's point of view, the reception box is the
easiest way of receiving, since this relieves the customer from having to be at
home when the goods arrive. Furthermore, by using a reception box, it is
possible to provide a fixed delivery frequency for the customers and even apply
vendor-managed inventory (VMI) principles to the households (Feare, 1999; Småros
and Holmström, 2000). This creates better opportunities to arrange picking and
transportation operations more efficiently (Waller et al., 1999). Fixed
delivery frequencies also level demand. However, from the customer's point of
view a fixed delivery frequency is not enough if additional groceries are needed
quickly. Furthermore, some customers require personal service.
Attended reception is a puzzle for the e-grocer, because it
requires offering a short delivery window and response time. This leads to a
situation where both delivery and picking timetables become more inflexible.
However, if personal service is not required, a reception box in the household
offers the possibility of creating a more flexible delivery timetable. For
example, Streamline drops about ten orders per hour compared to about three for
a typical e-grocer using attended receiving (Lardner, 1998).
Table
II presents the service concept of some existing home delivery suppliers.
We built a computer model to study the home delivery cost of
different service concepts. Figure
3 presents the framework of this model. In addition to the cost estimate,
each scenario results in an associated service estimate. Due to the nature of
the business, we have chosen to compute the cost with a fleet required to meet
100 percent on-time delivery performance within the model. The delivery cost is
computed directly from vehicle usage hours.
The customer base is limited to the geographic area where the
service is offered. Furthermore, it is reduced by the customer penetration of
the e-grocer in the area. The typical customer spends a certain amount of money
on the product categories offered by the e-grocer. After that the shopping
frequency defines the number and size of the daily orders.
The service concept is defined by the characteristics already
presented in TableII:
In the computer model, each scenario is constructed in two
steps. First orders are generated and then they are routed. The first step
results in an order file. In the second step, the orders are routed using
RoutePro, a routing software from CAPS Logistics. The routes are constructed
using the tree algorithm included in the software. The algorithm builds routes
one at a time and tries to build a full truckload with one vehicle before using
another vehicle. The order with the most distant destination is chosen first.
The algorithm grows the route by adding the unrouted candidate order whose
summed distance to the seed location and the depot is the smallest.
The order file format presented in Table
III is determined by the requirements of RoutePro. In addition to the volume
of the order, the routing is limited by the two time windows included in the
order file. The delivery, specified by drop-off start and drop-off end, depends
on the drop-off type. If there is a reception box in the household, the delivery
window equals the operating hours. Otherwise the operating hours are divided
into delivery windows as defined by the service concept. The delivery window of
each order is defined independently using a distribution. The shape of this
delivery distribution describes when the customers want their deliveries to
arrive. The pick-up window, which models when the order is available for pick-up
from order assembly, is determined in the order file by using pick-up start and
pick-up end. The delivery concepts presented later deal with next-day deliveries
and the order is assumed to be ready for pick-up when the operations begin in
the morning. If the service concept allows same-day deliveries, the delivery
will be available for pick-up at the last possible ordering time for each
delivery window.
Figure
4 presents an example of the relationship between the time windows in the
order file. As noted above, the delivery window specification is based on the
delivery time distribution. The service concept specifies a last possible order
arrival time for each delivery. This is used as the starting point of the
pick-up window. The pick-up window ends as the delivery window ends.
In the case study, which is based on an ongoing pilot in the
suburban area of Helsinki, we studied the attended reception and reception box
concepts by using the home delivery framework presented above. The service area
in the model is a circle with a 5km radius. The size of the customer base in the
case is 300 households. This number equals 2 percent of the families living in
one-family houses or semi-detached houses in the selected area, and can be
regarded as a potential penetration for the service.
The concepts studied are based on next-day delivery. As can
be seen in Table II, companies using attended reception typically offer a
shorter order lead-time. The modification, however, is made to make the service
level of the concepts more comparable. Two deliveries a week per customer are
assumed, which results in 120 daily orders, with no daily demand variation and
no deliveries on weekends. The fleet consists of one vehicle, which can operate
12 hours a day.
Several scenarios were constructed to explore the effect of
different characteristics of the system on the efficiency of the delivery. One
factor examined was the capacity of the vehicle: the vehicle can carry 40 or 60
orders at a time. The other factor studied was the drop-off time, i.e. the
duration of the stop at the customers. Despite the fact that delivery to a
reception box can be considered quicker, the factor was given the values of one
and two minutes in both concepts. Furthermore, the concepts were compared using
the same vehicle capacity and drop-off time respectively. Each scenario was
replicated five times by sampling from the 300 customers.
The main result of the study is presented in Table
IV. On the average, with the reception box concept an over 40 percent cost
reduction can be achieved compared to the attended reception concept. The
results presented in Table IV are for vehicles with a capacity of 40 orders and
a two-minute drop-off time.
The characteristics of the system using the box are explored
in Table
V and Table
VI. Increasing the capacity from 40 to 60 orders per route does not affect
theresults significantly. In fact, the results in Table V indicate that the
larger vehicle performs even worse in some cases. This is due to the fact that
the routing algorithm uses the full capacity of the vehicle when constructing
each route. The monthly cost of a suitable delivery vehicle with a driver has
been estimated to be around e3,000. As there are 21.5 working days a month,
thedaily cost will be e140. The pure delivery cost per order is therefore
approximately e1.20 in the reception box concept.
In the attended reception concept, one vehicle becomes
overbooked even with a one-minute drop-off time, regardless of vehicle capacity.
It can deliver 96 of the 120 orders. Delivering these orders it travels 173km,
which equals 1.8km per order. In the busiest delivery window the vehicle travels
1.5km per stop. During this period it can deliver up to 26 orders in 108
minutes, which equals roughly 14 orders an hour.
The average figures per order above cannot be used directly
to present the delivery efficiency of the attended reception concept. This is
because the orders actually delivered are not selected randomly but using the
routing algorithm. To get a comparable mileage, an additional vehicle is needed.
The results with two vehicles with the capacity of 40 orders can be found in Table
VII. The additional vehicle is employed for five hours. On top of the
delivery cost of the first concept comes the additional cost of five vehicle
hours. According to Road Freight Finland, the hourly rate for a vehicle with
partial utilization is e22.5. The additional cost per order is therefore e0.94,
giving a total delivery cost of e2.10 per order.
The results show a delivery cost difference of over 40
percent when using different receiving concepts. The difference in costs is due
to the wider delivery window enabled by the reception box. This wider delivery
window can be used to minimize the visiting of different geographical areas but
also to level out the delivery workload during the day. Because of these effects
alone, the orders can be delivered with a more than 50 percent shorter distance
driven compared to attended reception.
As always, the model used includes simplifications that may
affect the results. These simplifications, however, favor the more inefficient
alternative and thus do not reduce the observed comparative advantage in
practice. The model studies only a very narrow part of the e-grocery supply
chain. Accordingly, no direct conclusions about the overall efficiency of the
concepts can be made. Because of the strong link between delivery and order
picking, however, it is fairly safe to say that the balancing of the workload in
delivery affects order picking similarly. The model does not take into account
that the reception box concept with fixed delivery dates would allow even more
compact routes when the service is offered to different areas on different
weekdays. Fixed delivery dates also enable the leveling out of within-week
fluctuations experienced in traditional grocery shopping. Furthermore, the
concepts differ in aspects that go somewhat beyond routing. An order generated
by a customer acquiring a reception box is probably bigger on average, since
such households are probably larger and customer loyalty is stronger.
The major conclusion of this paper is that the way e-grocery
consumers receive the goods has a major impact on the grocery supply chain. In
addition to inefficient home delivery, picking causes problems in the e-grocery
supply chain. Actually, inefficient home delivery links very closely to the
picking problem, because an efficient supply chain cannot be created if these
two operations do not level demand together.
The results give support to the claim that an e-grocery
business operated from a local distribution center and using reception boxes
could have a cost advantage even over traditional grocery stores (Macht, 1999).
When using reception boxes, there are also better opportunities to create
lasting, regular customer relationships by providing good prices and customized
services. This means that a better service can be offered while lowering the
delivery cost at the same time.
The distribution center, combined with the reception boxes,
opens up opportunities to level out demand and create a more efficient supply
chain. By using reception boxes and variable pricing depending on time and day,
the e-grocer can level out demand peaks which are currently concentrated on late
afternoons and weekends. This means level demand in home delivery and picking.
In addition, this can level out demand in production as well.
Further research will focus on two topics. First, how does
the system behave, when there are both attended reception and reception boxes?
Initial analyses show that neither one of the readily available routing
algorithms included in the RoutePro v. 6.0 fully utilizes reception box
deliveries to level out the workload. As the value of the average order is
relatively low, it is crucial that the manual work associated with routing be
minimized. Second, future models will explore larger scale operations. This
larger scale will have effects deeper in the supply chain. This opens new,
exciting research questions. With next-day deliveries, the exact demand of the
local distribution center is known one day in advance. There is no need for
storing products with a one-day or less order lead-time. Furthermore, this
information could be used in scheduling production of such products.
The development of e-grocery operations is very fast at the
moment. Players work hard by developing different kinds of solutions for
reaching better home delivery and picking efficiency. In the future, there will
probably be many different kinds of alternatives for receiving the goods
depending on customer preferences and willingness to pay for the service.
However, looking at the operational costs of e-grocers, we believe that the
reception boxes combined with efficient distribution centers will have a clear
benefit compared to other home delivery solutions.
1. Streamline, www.streamline.com
2. SOK, www.s-kanava.fi/s-box
3. Peapod, www.peapod.com
4. Y-halli, www.yahilli.fi
5. Webvan, www.webvan.com
6. Matomera, www.matomera.se
Figure 1.
The e-grocer as an intermediary in the supply chain
Figure 2.
An e-grocer using a distribution center in the supply chain
Table I.
Most significant differences between the attended reception and reception box
concepts
Table II.
Classification of some existing e-grocers (Streamline, 1999; Matomera, 1999;
S-market, 1999)
Figure 3.
Home delivery framework
Table III.
Example order file (transpose)
Figure 4.
Order file windows
Table IV.
Delivery cost per order
Table V.
Busy hours of the vehicle delivering to reception boxes
Table VI.
Distance travelled by the vehicle delivering to reception boxes
Table VII.
Results in the attended reception concept
Andersen Consulting, 1998, "On-line grocery shopping
on track for rapid growth",
http://www.ac.com:80/services/food/food_news.html.
Bennett, N, 2000, "Homeport addresses the delivery
dilemma", The Daily Telegraph.
Cuglielmo, C, 2000, "Can Webvan deliver the goods?",
Inter@ctive Week, http://www.zdnet.com.
Dagher, N, 1998, "Online grocery shopping",
1998 INSEAD, Fontaineblau.
Feare, T, 1999, "Building a new kind of online
business", Modern Material Handling,
http://www.manufacturing.net/magazine/mmh/archives/1999/mmh0801.99/08
pioneer.htm.
Heikkilä, J, Kallio, J, Laine, J, Saarinen, L., Saarinen, T,
Tinnilä, M., Tuunainen, V., Vepsäläinen, A.P.V, 1998, "Ensi askeleet
elektronisessa kaupassa", Tekes digitaalisen median raportti, 3, 98.
Holmström, J, Hoower, W., Eloranta, E., Vasara, A, 2000,
"Discovering your value offering points: how to become a shaper of
demand-supply chains", The International Journal of Logistics
Management.
Holmström, J., Tanskanen, K., Kämäräinen, V, 1999, "Redesigning
the supply chain for Internet shopping - bringing ECR to the households",
Working paper, Logistics Research Network Conference, Newcastle.
Kämäräinen, V, 2000, "Internet muuttaa
jakelurakenteita - Elektronisen päivittäistavarakaupan liiketoimintamallit ja
loppujakelupaikan valinta", Pro Gradu, Helsinki School of Economics.
Lardner, J, 1998, "Please don't squeeze the tomatoes
online", US News and World Report.
Macht, J, 1999, "Errand boy", Inc.
Magazine.
Mullaney, T.J, Leonhardt, D, 1999, "A hard sell
online? Guess again", Business Week.