Comparative analysis of the
carbon footprints of conventional
and online retailing
A “last mile” perspective
Julia B. Edwards, Alan C. McKinnon and Sharon L. Cullinane
Logistics Research Centre, Heriot-Watt University, Edinburgh, UK
Purpose – The purpose of this paper is to focus on the carbon intensity of “last mile” deliveries (i.e.
deliveries of goods from local depots to the home) and personal shopping trips.
Design/methodology/approach – Several last mile scenarios are constructed for the purchase of
small, non-food items, such as books, CDs, clothing, cameras and household items. Ofﬁcial government
data, operational data from a large logistics service provider, face-to-face and telephone interviews
with company managers and realistic assumptions derived from the literature form the basis of the
calculations. Allowance has been made for home delivery failures, “browsing” trips to the shops and
the return of unwanted goods.
Findings – Overall, the research suggests that, while neither home delivery nor conventional
shopping has an absolute CO
advantage, on average, the home delivery operation is likely to generate
than the typical shopping trip. Nevertheless, CO
emissions per item for intensive/infrequent
shopping trips by bus could match online shopping/home delivery.
Research limitations/implications – The number of items purchased per shopping trip, the
choice of travel mode and the willingness to combine shopping with other activities and to group
purchases into as few shopping trips or online transactions as possible are shown to be critical factors.
Online retailers and home delivery companies could also apply measures (e.g. maximising drop
densities and increasing the use of electric vehicles) to enhance the CO
efﬁciency of their logistical
operations and gain a clearer environmental advantage.
Practical implications – Both consumers and suppliers need to be made more aware of the
environmental implications of their respective purchasing behaviour and distribution methods so that
savings can be made.
Originality/value – The paper offers insights into the carbon footprints of conventional and online
retailing from a “last mile” perspective.
Keywords Carbon, Delivery services, Internet shopping, Distribution channels and markets,
Paper type Research paper
Some online retailers have been actively claiming that internet shopping yields
environmental beneﬁts (Smithers, 2007). Equally, consumers seem to have a widely held
view that online purchases and home delivery are beneﬁcial to the environment because
they reduce personal travel demand (Royal Mail, 2007). Such opinions are also prevalent
among researchers. For instance, Rotem-Mindali and Salomon (2007, p. 178) point out that:
[...] studies of the impacts of teleshopping on transport usually assume that the delivery trip,
by the retailer or a third party, to multiple customers is more efﬁcient than individual trips.
The current issue and full text archive of this journal is available at
Analysis of the
Received March 2009
Revised August 2009
Accepted October 2009
International Journal of Physical
Distribution & Logistics Management
Vol. 40 No. 1/2, 2010
q Emerald Group Publishing Limited
To date, however, little research has tested the claims that online retailing is
environmentally superior. An early paper by Matthews et al. (2001), which compared
the environmental impact of online and conventional book retailing in the USA, offered
some support for this view. A review of the academic literature revealed little relevant
research on this issue since the early 2000s.
This paper sheds new light on this subject. It focuses on the so-called “last mile” (i.e.
the last link in the supply chain to the home) and compares the level of carbon
emissions from a conventional non-food shopping trip with those of delivering
non-food items to the home. It is based on research in the UK, where online retailing
has now captured around 17 per cent of total retail sales (IMRG/Capgemini, 2008).
Transport at the local level is not just the most visible; it can also be the most
energy-intensive. Browne et al. (2008) note for conventional shopping that personal
shopping trips can use more energy than the entire upstream supply chain, even when
production is included. Several past studies have examined the last mile delivery
(European Information Technology Observatory, 2002; Abukhader and Jo
Sarkis et al., 2004; Farag et al., 2006), although none have systematically compared
consumer travel with freight delivery in terms of energy expenditure and CO
emissions per delivery drop/item bought.
This paper is structured as follows: ﬁrst, we brieﬂy examine previous work on the
relative environmental effects of online and conventional shopping. This includes a
comparison between the last mile and upstream supply chain activities to determine
the importance of last mile operations relative to emissions from end-to-end supply
chain activities. This is followed by a review of the data sources used and discussion of
the assumptions underpinning our comparative model. Various home delivery and
shopping trip scenarios are presented in the main results section. In addition to
summarising the main ﬁndings, the conclusions explore some of the wider issues
raised by the research.
2. The environmental effects of online versus conventional shopping
Probably, the greatest difference between online and conventional shopping can be
seen in the fulﬁlment and distribution processes required to meet customer
expectations (de Koster, 2002). In the traditional shopping model, customers do most
of the labour-intensive work (such as order-picking and transporting the goods home),
whereas in e-fulﬁlment, retailers must deliver personalised orders to highly dispersed
locations within relatively narrow time windows.
Concerns have been expressed about the steep increase in home deliveries, some of
them relatively inefﬁcient, which reduce the net beneﬁt of online retailing (Romm, 1999;
ger et al., 2003). Early research suggested that, although car-based shopping trips
could be reduced by as much as 10 per cent as a result of internet shopping, more
research was needed to assess possible environmental disbeneﬁts (DTLR, 2002). These
might result from:
the fragmentation of the fulﬁlment process with consumers making more
frequent purchases of relatively small quantities of goods, often from several
different web-based companies;
other car-based out-of-home activities being undertaken by either the car owner
or other household member in the time saved by online shopping;
the net increase in the total amount of material consumption; and
additional transport created by failed deliveries, when no one is at home, and the
return of unwanted goods.
Much of the previous research comparing online and conventional shopping has
concentrated on the grocery retail sector (Cairns et al., 2004; Cairns, 2005; Foley et al.,2003;
Gould and Golob, 1997). In the traditional grocery supply chain, goods are delivered to
store where the customer picks the items before taking them home. For e-grocery,
however, there are three scenarios: picking and distribution from existing stores; direct
home delivery from a dedicated fulﬁlment (or “pick”) centre and home delivery from a
central warehouse via a satellite depot (van der Laan, 2000; Agatz et al., 2006). The most
successful UK model to date has been order-picking and distribution from existing stores
(Hackney et al., 2006). Punakivi (2003) noted a considerable trafﬁc reduction (of between 54
and 93 per cent depending on delivery method) when e-grocery replaced car-borne
shopping trips to supermarkets in Finland. Further savings can be achieved when grocery
reception boxes are used instead of attended delivery (where the consumer must be at
home to receive the goods). Ka
inen et al. (2001) recorded home delivery distance
savings of over 50 per cent from the use of grocery reception boxes in Finland.
While most of the research to date has concentrated on the “last mile” stage, a few
studies have compared the energy consumption of consumer travel and home delivery
with energy use further upstream in the supply chain. Jespersen (2004) conducted
telephone interviews to establish consumer travel behaviour when purchasing rye
bread from shops. Assumptions were made about trip chaining (50 per cent of an
average 5 km trip was for shopping) and the weight of goods purchased (20 kg). His
ﬁndings revealed that the amount of energy consumed by the customer’s trips to and
from the shop was greater than the energy used in all the other transport associated
with the production and distribution of the bread. Browne et al. (2006), in investigating
the various stages of the production and distribution of jeans, observed that the energy
used for a dedicated consumer shopping trip (of 11 km) was approximately the same as
that used in transporting the product from the jeans factory (based in the USA or
Turkey) to the UK port, despite the huge differences in journey lengths. Similarly,
Weber et al. (2008), when comparing the energy use and CO
emissions generated by
both the online and conventional distribution of an electronic ﬂash drive, found that
approximately 65 per cent of total emissions for traditional retailing came from the
customer trip to and from the retail store.
These studies not only highlight the differences in carbon intensity across the
“end-to-end” supply chain, but they also show how transport energy and emission
calculations are dominated by last mile operations. They, therefore, provide
justiﬁcation for our decision to focus on this last link in the retail supply chain.
3. Research approach
This paper presents the results of a comparative study of the CO
emissions from home
deliveries and conventional shopping trips in the non-food retail sector. The focus is on
small non-food products such as books, CDs, clothing items and electronic devices,
which, because of their physical characteristics, are responsible for very similar
amounts of energy use and emissions when transported in freight vehicles, cars or
public transport. In this analysis, they are considered to be identical in terms of their
Analysis of the
transport-related carbon footprint. The paper examines the carbon emissions from
transporting them solely at the “last mile” stage in the supply chain (from store to
home or local depot to home). Using published UK Government statistics and primary
data from one of the UK’s largest home delivery companies, it has been possible to
model the amounts of CO
emitted by conventional and online purchases of small
non-food items. An Excel spreadsheet was constructed for this purpose.
The emission factors for home delivery operations by diesel- and petrol-fuelled vans
were obtained from four statistical sources and averaged:
(1) Defra’s (2008)emissions factors for vans.
(2) National Atmospheric Emissions Inventory (2008) emissions factors for vans: data
for Euro II vehicles, and speeds of 40 kph (default speed), 20 kph (representative of
average urban speeds) and 10 kph (worst-case scenario) are applied.
(3) RHA Cost Tables, 2008: emissions factors are calculated from Defra values,
based on average fuel consumption of 9.6 km per litre (27 miles per gallon) for a
van (Road Haulage Association, 2008).
(4) Freight Transport Association (2007) distribution costs 2008: emissions factors
are calculated from Defra values, based on average fuel consumption of 8.9 km
per litre (25 miles per gallon) for a van.
This averaging ensured both consistency and reliability in the calculations.
Several last mile scenarios are proposed based on publicly available data,
face-to-face and telephone interviews with practitioners in industry and the results of
previous studies. Sensitivity analyses have been performed to assess the impact of
varying key parameters. Acquiring qualitative insight from practitioners helped to
verify the robustness of data obtained from the home delivery company.
Average emission factors for car and bus journeys (expressed as CO
travelled) to the shops were found in Defra (2007). In the case of cars, additional
calculations are made for speciﬁc vehicle exercise duty (VED) bands, particularly for
low-emissions vehicle (Band A), a hybrid vehicle (Band B) and a high-emissions vehicle
(Band G). Band-speciﬁc emissions have been sourced from the Vehicle Certiﬁcation
Agency’s records (www.vcacarfueldata.org.uk).
4. Modelling assumptions
4.1 Online shopping
Methods of delivery. The vast majority of online purchases result in the physical
movement of a small package (or single item) to an individual address (typically a
consumer’s home) by parcel carrier (RAC Foundation, 2006; Retail Logistics Task
Force, 2001). In general, these deliveries are distributed from local parcel carrier depots
and consist of mixed loads in the back of vans. Volumes delivered are high: the leading
parcel delivery carrier in the UK delivers some 300,000 parcels daily. Concern has been
expressed about the environmental repercussions of this expanding home delivery
market (Webster, 2007). Total mileage travelled by vans has risen by 40 per cent over
the past ten years in the UK, partly reﬂecting the growth in online retailing
(Department for Transport DfT, 2009c). Vans also have relatively high
carbon-intensity, expressed as g of CO
per tonne km (McKinnon, 2007), particularly
as much of their mileage is run on urban roads. For this reason, there has been
increased interest in the use of electric vehicles for home delivery, especially in the
online grocery sector. Sainsbury’s (2007), for example, plans to convert its entire online
grocery delivery ﬂeet to electric vans by 2010.
Vans are not the only delivery vehicles employed by parcel delivery companies. The
use of self-employed couriers has been on the increase recently (Beveridge, 2007).
Several leading online retailers now use third-party courier networks for deliveries.
These deliveries are generally made by private cars and are much shorter than typical
van-based delivery rounds. As courier rounds have different delivery and vehicle
characteristics, they have been excluded from the analysis.
Traditionally, vehicle load factors have been measured with respect to weight. For
vans, in the home delivery sector, the number of drops per round is more representative
of vehicle utilisation than the total weight of the consignments. Rather than
considering vehicle ﬁll as a percentage of maximum permissible weight, parcel
delivery companies are concerned with achieving high-drop density rates per round by
maximising the number of deliveries, a key productivity measure in the home delivery
sector. The parameters of vehicle ﬁll and empty running are not therefore included in
this analysis. All delivery drops are treated equally, regardless of when in the round
they are actually delivered. This approach may be criticised, as those deliveries
dropped ﬁrst, it could be argued, should be apportioned less CO
than those items
delivered later in the round. While correct in theory, assigning emissions based on the
sequencing of delivery drops would be an almost impossible task.
Home delivery companies do not normally adopt a strategy of dropping-off the
heaviest (or bulkiest) loads ﬁrst. It is clear from observations of loading practices by
van drivers and discussions with depot managers that customer location is the main
determinant of the loading/unloading sequence. Therefore, across the range of small,
non-food consumer products typically bought online, the physical nature of the
products has little effect on the energy intensity and carbon intensity of the delivery,
i.e. weight/density are not signiﬁcant. The main variable is the number of drops per
round. Given the granularity of this analysis it is not necessary to distinguish between
speciﬁc product types within the general category of small non-food items, which can
be collected by consumers from high-street shops.
Two other factors are likely to have a greater inﬂuence on the level of CO
emissions: the chances of making a successful deliver ﬁrst-time and the nature of the
returns process for unwanted/damaged goods. These will be considered next.
Incidence of ﬁrst-time failed delivery. It has become more common for people not to
be at home during the working day when most home deliveries are made. Prologis
(2008) reported that the number of working households increased by 22 per cent
between 1992 and 2006. As a result, parcel carriers must cope with increasing incidence
of failed delivery. Actual failed delivery rates among carriers vary considerably.
Beveridge (2007), a leading consultant in the home delivery market with wide
experience of managing last-mile delivery networks, indicated a range between 2 and
30 per cent, depending on the carriers’ policies for dealing with “no-one-at-home”. Some
parcel delivery companies achieve very high ﬁrst-time delivery rates as they are
prepared to leave deliveries in alternative locations, such as with neighbours or in the
garden shed (McKinnon and Tallam, 2003). While these places are often insecure, the
use of dustbins is now generally avoided owing to earlier reported mishaps! Other
carriers require proof-of-delivery signatures, and consequently have a much higher
Analysis of the
delivery failure rate. As a result of different delivery arrangements, estimates of
ﬁrst-time delivery failure rates vary widely from six out of every ten small-package
deliveries (Retail Logistics Task Force, 2001) to a more conservative one in eight
(IMRG, 2006). This study uses three failed delivery ratios. First, a ﬁrst-time failure rate
of 25 per cent of deliveries, in line with ﬁndings by McLeod and Cherrett (2006) and
Song et al. (2009); second, a 12 per cent failure rate (assumed by Weltevreden and
Rotem-Mindali (2008), and based on IMRG (2006) ﬁndings), and ﬁnally, a very
successful ﬁrst-time failure rate of 2 per cent, achieved by parcel companies whose
delivery drivers seek alternative locations at which to leave items.
The return of unwanted goods. Customers return items for a number of different
reasons. They may, for instance, be the wrong product, because of errors in order
picking, unsuited to the consumers needs, or damaged in transit. Online retailers also
have widely varying returns policies from unconditional money back guarantees to
store credit only to no refund whatsoever (Mukhopadhyay and Setoputro, 2004).
Typically, between 25 and 30 per cent of all non-food goods bought online are returned
(de Koster, 2002) compared with just 6-10 per cent of goods purchased by traditional
shopping methods, though this varies widely among product groups and probably
geographically (Nairn, 2003; Fernie and McKinnon, 2009).
The environmental implications of these online returns are strongly inﬂuenced by
both parcel carriers’ returns policies and consumers’ preferred habits. Parcel carriers
who collect returned items as part of their usual delivery round generate very little
additional mileage. In these cases, an allowance is made for collections within planned
delivery drop-rates, and any additional energy use is subsumed within the overall
delivery round. On the other hand, some delivery companies send vans on separate
pick-up runs dedicated solely to collecting returned items. Consideration of this
dedicated collection process is treated separately.
The situation is complicated further by customers often having a choice of returns
channels. For retailers with a high-street presence, customers may choose to return
items to a physical store. The popularity of this method depends on the number of
high-street stores operating such a returns policy. For instance, a high percentage of
online supermarket clothing returns are handled through supermarkets, whereas some
multi-channel retailers have very little returned to stores owing to their relatively
sparse high-street presence.
Alternatively, customers can send items back through the standard postal service.
Where there is a choice between parcel carrier or postal services, approximately half of
returns are via carrier collection and half by post (Beveridge, 2007). Some high-street
retailers ﬁnd that half their returns are to stores, and the remaining half-split between
carrier collection and the post. The model takes account of these different returns options.
4.2 Conventional shopping
There is no such thing as a “typical” high-street shopper. In creating characteristic shopper
proﬁles consideration needs to be given to several key questions: how people travel to the
shops, how frequently they shop, what they buy and in what quantities they purchase
goods. Finding general answers to these questions is difﬁcult owing to a lack of
behavioural data at the consumer level (Rotem-Mindali and Salomon, 2007). Some retailers
undertake their own customer surveys, but are usually reluctant to release the results.
Our analysis has relied mainly on government statistics available at the national level.
Dedicated shopping-only trips. The National Travel Survey, undertaken by the DfT,
collects data on personal travel behaviour over time, which allow comparison between
food and non-food shopping trips. Table I lists the average distances travelled for
shopping by different transport modes. The National Travel Survey deﬁnes a trip as a
one-way journey with a single main purpose, with outward and return halves of a
return trip treated as two separate movements (DfT, 2009b). Therefore, an average
dedicated shopping trip would require a doubling of the distances shown in Table I.
Average distances travelled for non-food purchases are longer than for food shopping
trips, at 6.4 miles for car travel (car driver) and 4.4 miles for bus travel (DfT, 2009a).
These distances are used to represent average shopping trips.
Car and bus travel are the two motorised transport modes most used by
conventional shoppers, accounting for 72 per cent of all shopping trips (DfT, 2009a),
and as such, are the only modes considered in this paper. Rail is omitted, as it is not a
regular mode for shoppers (less than 1 per cent of shopping trips are by rail) (DfT,
2009a). Walking and cycling have been also been excluded from the calculations, as
both modes involve human effort (a category excluded from typical life cycle
assessments), and neither emits easily attributable CO
emissions. The environmental
and social beneﬁts of both are acknowledged, however.
Combined and/or browsing-only shopping trips. Trip chaining is a widely used term
to describe a combined trip. Although having no agreed deﬁnition, it can be described
as a household’s tendency to combine different activities during a single trip (Popowski
Leszczyc and Timmermans, 2001), with a trip segment representing the travel between
a particular pair of activities (Primerano et al., 2008). Often, minor detours to a store are
incorporated into a trip made primarily for some other purpose, adding only
marginally to the total distance travelled.
As Brooks et al. (2008, p. 29) state: “the high incidence of multi-stop trips in
empirically observed behaviour makes the single-stop assumption unrealistic”; most
trips for shopping involve multi-stop activities either between different stores or
different activities, including shopping (i.e. from work to home, calling at shops on the
way). In such cases, the allocation of energy consumption related to the purchasing
activity needs to be reduced accordingly (Browne et al., 2008).
Usually, consumers visit more than one shop per trip especially when shopping for
non-food products (Brooks et al., 2008). Establishing the number of items consumers
buy on each trip is far more problematic as individual retailers only have information
Mode Average distance (miles)
Car/van (driver) 6.4
Car/van (passenger) 8.3
Other private 4.3
Local bus 4.4
Other public 12.5
All modes 5.4
Note: 2005-2006 (one-way)
Source: Derived from DfT (2009c) Personal communications: National Travel Survey
Average trip length for
non-food shopping by
Analysis of the
about the number of products bought in their own stores, and not as part of the
shopping trip as a whole. It seems that no information is collected about the overall
quantities of goods bought per shopping trip. Therefore, in the analysis reported here,
we have had to estimate a range of values for this critical variable. It would clearly be
preferable to have empirical data on the number and types of item bought on the
shopping trips. In the absence of this information, however, calculations based on
theoretical values still allow cross-channel comparisons of a “what if [...]” nature.
It must also be remembered that some shopping trips do not result in a purchase.
Some trips to the high street may be for information-gathering purposes only. This
“browsing” category has been largely ignored by researchers owing to a lack of data
(Moe and Fader, 2001), yet frequently a fact-ﬁnding visit results in a later purchase
(often online) (Skinner et al., 2004).
4.3 Speciﬁcation of the model
Online shopping: delivery rounds and drop characteristics for non-food. From
face-to-face and telephone interviews with logistics managers, local depot supervisors
and delivery van drivers from four different leading parcel delivery companies, we
established that a:
highly efﬁcient home delivery operation would have a drop density of
approximately 150 drops on a 60 mile delivery round; and
city centre-focused round would usually cover about 25 miles and comprise
approximately 110 drops on average.
Given these operational characteristics and for ease of comparison, this study
examines an average delivery round by a van, which we assume consists of 120 drops
on a 50 mile round.
It is assumed that each package delivered as part of this representative home
delivery round weights less than 25 kg (the maximum permissible weight for a
one-man delivery). Equally, no distinction is made between the different types of
products delivered; as all items are treated equally in the delivery process.
Calculations of the number of items per drop have been performed. Initial results are
shown for a single item per drop. However, a more realistic assumption, based on
discussions with a leading book wholesaler, is for each drop to contain either 1.4 items in
the case of deliveries containing books/DVDs/CDs or 2.5 items for other non-food goods
(e.g. clothes and household items) (Beveridge, 2009). Therefore, additional calculations
for multiple items per drop are also included. Some online retailers have a dispatch
policy where they delay distribution until all items purchased are available for delivery,
while others prefer to send one item per package regardless of the number of goods
ordered at the time of transaction. For direct comparison with conventional shoppers’
emissions have been calculated on an item basis. The assumptions made
about “last mile” delivery are listed in Table II and represent the expert knowledge of
those working in the industry or are derived from previous work in this area.
Conventional shopping: personal travel. The average car driver makes a round trip
of 12.8 miles for non-food shopping purposes (DfT, 2009a). For bus passengers, the
average return journey to the shops for non-food items is slightly less at 8.8 miles (DfT,
2009a). The consumer travel behaviour characteristics assumed in the model are listed
in Table III.
When focusing exclusively on the last link in the retail supply chain (from depot or
shop to the home), home delivery by parcel carrier is often presumed to be more
efﬁcient than an individual travelling to the shops to buy the item in person. The
results in Table IV appear to support this supposition. Typically, one drop of 120 such
drops on a 50 mile delivery round is apportioned 181 gCO
. This ﬁgure has been
derived from the four freight emissions factors outlined in Section 4 and is a drop’s
“share” of the average emissions produced by the overall delivery trip (21,665 gCO
Assumptions Type of delivery round
(deliveries per round)
Van (, 3.5-t) Average 50 120
Efﬁcient 60 150
Failed ﬁrst-time deliveries (per cent) 25
Returns (percentage of orders) 25 (40 for clothing)
Method of return Collection
Freight “last mile”
Mode Round trip (miles)
Browsing (as percentage of all shopping trips) 10 (average)
Trip chaining (percentage of mileage attributed to
10 (only applies to trips by car)
Returns (percentage of all purchases) 8
Consumer travel and
per item delivered/
Standard delivery van (, 3.5-t) (120 deliveries 21,665 g 181 g (drop)
per 50 mile round trip) 137 g (1.4 items)
72 g (2.5 items)
Car (dedicated shopping trip of 12.8 miles) 4,274 g 4,274 g (single item)
Bus (dedicated shopping trip of 8.8 miles,
assuming average patronage
) 11,641 g 1,265 g
Note: Parcel carrier/car/bus
per average trip and
Analysis of the
Assuming that a shopper, using a standard car, makes a round trip of 12.8 miles to the
shops solely for the purpose of buying one item, the trip would generate 4,274 gCO
of which could be assigned to that one item). In this example, the CO
car-based travel is 24 times greater than the CO
produced by a single drop within the
average home delivery round.
An alternative way of interpreting these results is to say that a person would need
to buy 24 non-food items in one standard car-based trip for this method of shopping to
be less CO
intensive than having one non-food item delivered (on the ﬁrst attempt) to
their home by a parcel carrier. For a VED Band A vehicle (99 gCO
/km), 12 non-food
items would need to be purchased and for a mid-range Band G vehicle 31 items
km). A bus passenger, assuming average bus occupancy levels of 9.2
passengers for an 8.8 mile round bus trip, would need to purchase seven or more
non-food items to compete favourably with a home delivery in terms of carbon
The above calculations assume one item per drop for home delivery and only one item
per shopping trip. Although some deliveries to the home do only contain one item (some
online retailers only send items out individually regardless of order size), it would be more
realistic to increase the “items per drop” variable. With an average content of 1.4 items per
drop (e.g. a typical book order) the CO
per item is reduced to 137 g for home delivery.
When a home delivery (e.g. for clothing and household goods) consists of 2.5 items, the CO
per item is 72 g. These assumptions further increase the number of goods a conventional
car-based shopper would have to buy in one trip to 32 or 59 non-food items, respectively, to
contend with home delivery in terms of CO
efﬁciency. For bus travel, a shopper would
have to carry ten or 18 non-food items, respectively.
Although home delivery appears to have a strong environmental advantage over
consumers’ personal travel to the shops, this result requires several qualiﬁcations.
The investigations only compare theoretical delivery trips based on average values.
The last mile delivery is much more complex than these initial ﬁndings suggest
(Figure 1). For instance, a standard home delivery will vary by failed delivery rates (the
number of failed ﬁrst-time deliveries); distances covered (including type of road
network), and the method by which unwanted items are returned. These variants will
be examined in the Section 5.1.
5.1 Effects of varying home delivery parameters
Failed ﬁrst-time delivery rates. Failed delivery is both uneconomic for the carrier and
inconvenient for the shopper. Various failed delivery scenarios are considered, based
on the following:
(1) A 2 per cent ﬁrst-time failure rate, achieved by van-based parcel delivery
carriers who accept alternative drop-off arrangements when no-one is at home
for ﬁrst-time delivery.
(2) A 12 per cent ﬁrst-time failure rate, quoted by IMRG (2008) and considered to be
an average to good failure rate.
(3) A 25 per cent ﬁrst-time failure rate, often experienced by those carriers
requiring proof of delivery signatures. It was also the proportion of failed
ﬁrst-time deliveries noted by McLeod and Cherrett (2006) and Song et al. (2009)
The online retail channel:
Low emissions Electric
Items per package
Store Post Carrier
100% successful ﬁrst-time
Van-based deliveries: gCO
per item 181 g 185 g 203 g 226 g
item including failed
Analysis of the
Emissions of CO
per average drop increases from 181 g for a successful, ﬁrst-time
delivery to the worse-case scenario of an average 226 g per drop when one-in-four
Most delivery companies schedule the repeat delivery for the next working day after
the ﬁrst-failed attempt, and as a result a high percentage of second attempts also fail,
compounding the effects of the initial failed delivery. After a second failed attempt
non-delivered goods are held at the local depot, and “carded” customers (those
receiving a failed delivery card through the letterbox) have to visit the depot in person
to collect the item. Around 3 per cent of home delivery recipients make a trip to collect
an item left at a post-ofﬁce, depot or outlet DfT (2009b).
Returns. The returns process for unwanted goods can take a number of forms.
When a parcel carrier schedules collections into an outbound delivery round the gCO
per collection/item is effectively the same as per delivery. However, when alternative
arrangements are made (either on the part of the consumer or the carrier) more
complicated calculations are necessary. These are examined in Section 5.2.
5.2 Effects of varying shopping trip parameters
Consumer travel and shopping behaviour. The model also captures much of the
variability in consumer shopping behaviour. Some shoppers make dedicated trips to
shops when shopping is their only intention, while others may choose to combine
shopping with other activities as part of a trip chain. Additionally, both online and
conventional shoppers frequently choose to inspect items in stores (prior to buying
either in-store or online), and may make several trips to do so. When shoppers wish to
return unwanted items, they often have a choice of returns methods. Figure 2 shows an
indication of some of the choices available to the conventional shopper.
A certain number of shopping trips will end in no purchase, owing to the:
consumer failing to decide which item to buy;
particular good sought being unavailable; or
consumer having no intention to purchase anything, using the trip for
information gathering purposes only.
In these cases, the unsuccessful trip needs to be factored into the calculations. On the
“realistic” assumption that one in ten shopping trips for a particular product results in
no immediate purchase, the gCO
in each of the above-dedicated shopping trips would
increase by a factor of 1.1. Nevertheless, at a personal level, a shopper’s CO
footprint would be twice the amount to take account of a second (later) journey to the
shops. So, while total emissions for a “browsing plus purchase” average car trip would
be 4,701 gCO
£ 1.1), for the individual undertaking the second journey it
would be 8,548 gCO
Furthermore, a consumer may choose to acquire the item as part of a larger shopping
expedition when many items are bought, and/or to combine the shopping trip with other
activities (Table VI). The combined trip, in this instance, realistically assumes that
shopping-related mileage is a quarter of the overall trip mileage (25 per cent).
In Table VI, it can be seen that the most efﬁcient ways to purchase and collect a product
would be either as part of a much larger shopping trip when many items are bought at the
same time or where shopping is incorporated into trips made principally for another
purpose. Any consolidation of shopping activities clearly reduces their carbon intensity.
Bus travel can compete with home delivery in terms of CO
efﬁciency. During peak
leisure times (e.g. on a Saturday afternoon), when occupancy levels are high and most
non-food shopping occurs, from an environmental point of view, bus travel is an
effective method of collecting shopping. For example, assuming a shopper travels the
average distance (8.8 miles) by bus, in the company of 29 other passengers, and buys
ﬁve items, each purchase would be allocated a share of just 78 gCO
, less than half that
The conventional retail
channel: consumer choices
Low emissions Electric
Analysis of the
for a typical home delivery (181 gCO
). Encouragingly, from an environmental point of
view, most shoppers (63 per cent) state that they would have no difﬁculty getting to the
shops by public transport (DfT, 2005).
Returns. The actual gCO
per online order is very sensitive to the proportion of
products returned and the method of return. Two scenarios are considered:
(1) Where the unwanted item is collected on a subsequent delivery round. In this
case, the integrated returns collection is allocated 362 gCO
(twice the CO
outbound drop), as the unwanted item has the combined emissions of an
outbound and return trip (in effect two outbound deliveries).
(2) Where the consumer returns the item to a high-street store. In the case of an
online shopper making a separate car trip to return the item, the CO
þ 4,341 gCO
), calculated on an average car-based round
trip (13 miles). This is clearly the worst-case scenario. The marginal CO
could be greatly reduced by returning the item as part of another shopping trip
or by “trip chaining”.
emissions: last mile versus upstream activities
It is not only on the last link that the online and conventional retail channels vary, but
the structure of their upstream supply chains also differ and this too will affect their
relative carbon footprints. Ideally, one should compare the carbon intensity of the two
channels as far back as the point in the supply chain at which they diverge because up
to this point the amount of CO
emitted will be common to both channels (Figure 3).
This would allow us to put differences at the carbon intensity of last mile operations
Trip type Items bought
Dedicated Single item (one item) Car 4,274
Electric car 1,586
Multiple purchase Car 855
(ﬁve items) Electric car 317
Browsing (two trips to shops: one for Single item (one item) Car 8,548
browsing, one for purchase) Electric car 3,172
Combined (shopping 25 per cent of trip Single item (one item) Car 1,069
mileage) Electric car 397
Combined then dedicated (25 per cent of Car 5,343
mileage: initial browsing followed by Electric car 1,983
dedicated trip to buy an item) Bus 1,581
Combined (grocery shopping: distance 7.12
) Multiple (50 items) Car 48
Average round trip distance to a supermarket (Future Foundation, 2007)
Implications of shopping
trip type on CO
The calculations in this paper and available published evidence suggest that emissions
from car-based shopping trips can far exceed those from distribution operations back
along the supply chain. It is likely, therefore, that the environmental comparison of online
and conventional shopping channels will be dominated by what happens at the local level.
Differences in CO
emissions between car-borne shopping trips and home deliveries are
likely to be much more important determinants of the respective carbon footprints of
online and conventional shopping than differences in upstream logistical operations as far
back as the point at which the two distribution channels diverge.
This study summarises the results of a comparative study of CO
emissions for the
home delivery and conventional shopping trips. While this so-called “last mile” has
received considerable attention from researchers, none of the previous studies have
attempted such a comparison on per trip, drop or item basis. Several scenarios were
investigated, and wherever possible representative values, derived from national
statistics, previous research or industry practice, were applied to different freight and
Numerous factors inﬂuence emissions from home deliveries. They include: drop
densities; the distance and nature of the delivery round; the type of vehicle used;
and the treatment of failed deliveries and returns. On average, when a customer buys
fewer than 24 items per shopping trip (or fewer than seven items for bus users) it is
likely that the home delivery will emit less CO
per item purchased. These ﬁndings
require several qualiﬁcations, however. They assume:
the car-based trip was solely for the purpose of shopping (no other activity was
undertaken during the course of the trip);
the product ordered online was delivered successfully ﬁrst time;
the shopper was satisﬁed with the purchase and did not return the item;
home deliveries and shopping trips were made over average distances; no
allowance was made for different types of road network or trafﬁc conditions; and
only the last mile and not the upstream supply chain has been considered in the
analysis (although reference has been made to previous studies of the relative
environmental impact of upstream activities).
Stages of book production
% figures indicate relative portion of greenhouse gas emissions
Source: Derived from green press initiative (2008)
Point of divergence
Analysis of the
The environmental implications of consumer behaviour have been illustrated by a
series of different shopping scenarios. Having already established that a standard
home delivery for a non-food item would be allocated 181 gCO
combined and browsing-only shopping trips were then compared. From the modelling
evidence presented here and from the results of previous research, it seems that
emissions from the average shopping trip, particularly by private car, can be greater
than emissions from all upstream logistical activity irrespective of the distribution
channel. Further work is underway to examine this issue in greater detail. The mode of
personal travel is particularly important. When a shopper travels by bus at busy times
and makes several purchases, the emissions per item are lower than when a home
delivery van delivers just one item to a consumer’s home.
It is acknowledged that people appear to regard shopping as a social, recreational or
even hedonistic activity to be enjoyed in a physical store. Given increasing concern for
climate change, however, it is important that they are made aware of the CO
consequences of their chosen shopping behaviour. With a little planning and thought
on both the part of consumers and carriers/retailers, emissions related to the transport
element of any shopping activity could be minimised through a few simple actions.
Carriers should aim to maximise drop densities (something that is likely to happen
anyway as a consequence of the growth of online retail sales), avoid dedicated
collection trips when picking-up returned items and where possible use low emissions
vehicles, e.g. electric vehicles. The use of reception boxes at people’s homes and
separate collection points would eliminate failed deliveries, the consolidation of orders
to a particular address in a single delivery would cut vehicle-kms and wider adoption
of variable delivery pricing would promote off-peak/out-of-hours deliveries, allowing
delivery vans to run more of their mileage at fuel-efﬁcient speeds.
Conventional shoppers meanwhile should ensure that when they go shopping
wherever possible they should combine their shopping trip with other activities and
thus avoid making a dedicated journey to buy a single item.
The relative carbon intensity of the different forms of retail distribution depends on
their particular circumstances. Neither has an absolute environmental advantage.
Some forms of conventional shopping behaviour emit less CO
than some home
delivery operations. On average, however, in the case of non-food purchases, the home
delivery operation is likely to generate less CO
. This environmental advantage can be
reinforced in various ways if online retailers and their carriers alter some of their
current operating practices.
Analysis of the emissions from the distribution of products with very different
characteristics, such as refrigerated food or bulky items (. 25 kg) that require a two-man
delivery, may yield very different results. The methodological approach outlined in this
paper could be applied to a comparative carbon analysis of these other sectors of the
retail market. Further research is also required to reﬁne the analysis for small non-food
items. This could explore the impact on carbon emissions of other behavioural responses
to the growth of online retailing not considered by the present study. Some online
customers, for example, may continue to shop as much by conventional means, but
merely buy less on each trip, effectively increasing the carbon intensity of each item
purchased by this means. Others may use the internet not just for purchasing goods but
to inform their conventional shopping decisions, allowing them to select products and
shops in advance and thereby rationalise their shopping-related travel. This future
research, like the present study, will require an extension of logistics’ traditional focus on
the transport of goods in dedicated freight vehicles to include the various forms of
personal travel associated with the movement of goods on the “last mile” to the home.
1. The carbon footprint of a product is the sum of all the carbon emissions for that product from
raw materials through manufacturing, distribution, use and disposal, taking account of all
related activities and materials. The calculations in this paper are for CO
emissions for the
last stage in the distribution only.
2. Trip-chaining occurs when a person visits several locations for different purposes in the
course of a single trip.
3. A van denotes a light goods vehicle up to 3.5 tonnes maximum permissible gross vehicle
weight of van-type construction on a car chassis that operates on diesel fuel unless speciﬁed
Abukhader, S.M. and Jo
nson, G. (2003), “The environmental implications of electronic commerce:
a critical review and framework for future investigation”, Management of Environmental
Quality: An International Journal, Vol. 14 No. 4, pp. 460-76.
Agatz, N., Fleischmann, M. and van Nunen, J. (2006), “e-Fulﬁlment and multi-channel
distribution: a review”, Report Series Research in Management Working Paper No.
ERS-2006-042-LIS, Erasmus Research Institute of Management, Rotterdam.
Beveridge, I. (2007), “Personal communications”, meeting with Beveridge Associates, 19 July.
Beveridge, I. (2009), “Personal communications”, 3 March.
Brooks, C.M., Kaufmann, P.J. and Lichtenstein, D.R. (2008), “Trip chaining behaviour in
multi-destination shopping trips: a ﬁeld experiment and laboratory replication”, Journal of
Retailing, Vol. 84 No. 1, pp. 29-38.
Brown, M., Allen, J. and Rizet, C. (2006), “Assessing transport energy consumption in two
product supply chains”, International Journal of Logistics: Research and Applications,
Vol. 9 No. 3, pp. 237-52.
Browne, M., Rizet, C., Leonardi, J. and Allen, J. (2008), “Analysing energy use in supply chains:
the case of fruits and vegetables and furniture”, Proceedings of the Logistics Research
Network Conference, Liverpool, September.
Cairns, S. (2005), “Delivering supermarket shopping: more or less trafﬁc?”, Transport Reviews,
Vol. 25 No. 1, pp. 51-84.
Cairns, S., Sloman, L., Newson, C., Anable, J., Kirkbridge, A. and Goodwin, P. (2004),
“Home shopping”, Smarter Choices – Changing the Way We Travel, Ch. 12, Department for
Defra (2007), “Passenger transport emissions factors”, Methodology Paper, Defra, London, June.
Defra (2008), Guidelines for Company Reporting on Greenhouse Gas Emissions, Defra, London,
Annexes, updated April.
de Koster, M.B.M. (2002), “The logistics behind the enter click”, in Klose, A., Speranza, M.G. and
van Wassenhove, L.N. (Eds), Quantitative Approaches to Distribution Logistics & Supply
Chain Management, Ch. 3, Springer, Berlin.
DfT (2005), Focus on Personal Travel, Department for Transport, London.
Analysis of the
DfT (2009a), “Personal communications”, data supplied from National Travel Survey,
DfT (2009b), Public Experiences of Home Working and Internet Shopping, Department for
DfT (2009c), Van Activity Baseline Survey 2008: Provisional Results, Department for Transport,
DTLR (2002), The Impact of Information and Communications Technologies on Travel and
Freight Distribution Patterns: Review and Assessment of Literature, Hopp Association,
European Information Technology Observatory (2002), “The impact of ICT on sustainable
development”, Part 2 in Annual Report, EITO, Paris.
Farag, S., Krizek, K.J. and Dijst, M. (2006), “e-Shopping and its relationship with in-store
shopping: empirical evidence from the Netherlands and the USA”, Transport Reviews,
Vol. 26 No. 1, pp. 43-61.
Fernie, J. and McKinnon, A.C. (2009), “The development of e-tail logistics”, in Fernie, J. and
Sparks, L. (Eds), Logistics and Retail Management, Ch. 10, Kogan Page, London.
Foley, P., Alfonso, X., Brown, K., Palmer, A., Lynch, D. and Jackson, M. (2003), The Home
Delivery Sector in the UK 1995 to 2010, Report to the Department for Transport,
De Monfort University, FTA & Associates, Leicester.
Freight Transport Association (2007), The Manager’s Guide to Distribution Costs, FTA, Sutton.
Future Foundation (2007), “Shopping miles”, Future Foundation Report, Future Foundation,
Gould, J. and Golob, T.F. (1997), “Shopping without travel or travel without shopping?
An investigation of electronic home shopping”, Transport Reviews, Vol. 1 No. 4, pp. 355-76.
Hackney, R., Grant, K. and Birtwistle, G. (2006), “The UK grocery business: towards a
sustainable model for virtual markets”, International Journal of Physical Distribution &
Logistics Management, Vol. 35 Nos 4/5, pp. 354-68.
IMRG (2006), Valuing Home Delivery: A Cost-beneﬁt Analysis, IMRG, London.
IMRG (2008), Valuing Home Delivery: e-Retail Industry Review, IMRG, London.
IMRG/Capgemini (2008), e-Retail Sales Index, IMRG, London.
Jespersen, P.H. (2004), “The transport content of products”, World Transport Policy & Practice,
Vol. 10 No. 3, pp. 28-35.
inen, V., Saranen, J. and Holmstro
m, J. (2001), “The reception box impact on home
delivery efﬁciency in the e-grocery business”, International Journal of Physical Distribution
& Logistics Management, Vol. 31 No. 6, pp. 414-26.
ger, K., Fergusson, M. and Skinner, I. (2003), “Critical issues in decarbonising transport:
the role of technologies”, Tyndall Centre Working Paper No. 36, Tyndall Centre, Norwich.
McKinnon, A.C. (2007), CO
Emissions from Freight Transport in the UK, Commission for
Integrated Transport, London.
McKinnon, A.C. and Tallam, D. (2003), “Unattended delivery to the home: assessment of the
security implications”, International Journal of Retail & Distribution Management, Vol. 31
No. 1, pp. 30-41.
McLeod, F. and Cherrett, T.J. (2006), “Optimising vehicles undertaking waste collection”, ﬁnal
report for the Department for Transport, Department for Transport, London, unpublished,
Matthews, H., Hendrickson, C. and Soh, D.L. (2001), “Environmental and economic effects of
e-commerce: a study of book publishing and retail logistics”, Transportation Research
Record, Vol. 1763, pp. 6-12.
Moe, W.W. and Fader, P.S. (2001), “Uncovering patterns in cybershopping”, California
Management Review, Vol. 43 No. 4, pp. 106-17.
Mukhopadhyay, S.A. and Setoputro, R. (2004), “Reverse logistics in e-business: optimal price and
return policy”, International Journal of Physical Distribution & Logistics Management,
Vol. 34 No. 1, pp. 70-88.
Nairn, G. (2003), “Not many happy returns”, The Financial Times, 5 February.
National Atmospheric Emissions Inventory (2008), Vehicle Emissions, Version 8, NAEI,
Popowski Leszczyc, P.T.L. and Timmermans, H. (2001), “Experimental choice analysis of
shopping strategies”, Journal of Retailing, Vol. 77 No. 4, pp. 493-509.
Primerano, F., Taylor, M.A., Pitaksringkarn, L. and Tisato, P. (2008), “Deﬁning and
understanding trip chaining behaviour”, Transportation, Vol. 35, pp. 55-72.
Prologis (2008), “Internet retailing: opportunities and challenges for the UK’S distribution
property markets”, Prologis Supply Chain Research Reports, Winter.
Punakivi, M. (2003), “Comparing alternative home delivery models for e-grocery business”,
unpublished PhD, Helsinki University of Technology, Helsinki.
RAC Foundation (2006), Motoring Towards 2050: Shopping and Transport Policy , RAC, London.
Retail Logistics Task Force (2001), @ Your Home: New Markets for Customer Service and
Delivery, Foresight Programme, Department of Trade and Industry, London.
Road Haulage Association (2008), RHA Cost Tables 2008, RHA, Weybridge.
Romm, J. (1999), The Internet Economy and Global Warming: A Scenario of the Impact of
e-Commerce on Energy and the Environment, Centre for Energy and Climate Solutions,
Rotem-Mindali, O. and Salomon, I. (2007), “The impacts of e-tail on the choice of shopping trips
and delivery: some preliminary ﬁndings”, Transportation Research Part A, Vol. 41,
Royal Mail (2007), Home Shopper Tracker 2007, RAPID Marketing Services, London.
Sainsbury (2007), “First electric vans to hit road with green shopping”, Company News,
available at: www.j-sainsbury.co.uk/index.asp?PageID¼418&subsection¼&Year¼2007&
NewsID¼893 (accessed 9 October 2008).
Sarkis, J., Meade, L.M. and Talluri, S. (2004), “E-logistics and the natural environment”, Supply
Chain Management: An International Journal , Vol. 9 No. 4, pp. 303-12.
Skinner, I., Fergusson, M., Kro
ger, K., Kelly, C. and Bristow, A. (2004), “Critical issues in
decarbonising transport”, Final Report of the Theme 2 Project T2.22, Tyndall Centre for
Climate Change Research, Norwich.
Smithers, R. (2007), “Supermarket home delivery service promotes its green credentials”,
The Guardian, 12 September, available at: www.guardian.co.uk/environment/2007/sep/12/
plasticbags.supermarkets (accessed 6 December 2007).
Song, L., Cherrett, T., McLeod, F. and Guan, W. (2009), “Addressing the last mile problem – the
transport impacts of collection/delivery points”, paper presented at the 88th Annual
Meeting of the Transportation Research Board, Washington, DC.
van der Laan, J.W. (2000), “The future of online food retailing”, Retail Economics, available at:
www.retaileconomics.com/aboutpub1.htm (accessed 17 January 2008).
Analysis of the
Weber, C., Hendrickson, C., Jaramillo, P., Matthews, S., Nagengast, A. and Nealer, R. (2008), Life
Cycle comparison of Traditional Retail and E-commerce Logistics for Electronic Products:
A Case Study of buy.com, Green Design Institute, Carnegie Mellon University,
Webster, B. (2007), “Boom in internet shopping may be adding to carbon dioxide emissions”,
The Times, 7 July.
Weltevreden, J.W. and Rotem-Mindali, T.O. (2008), Mobility Effects of b2c and c2c E-commerce:
A Literature Review and Assessment, Working First Version, Netherlands Institute of
Spatial Research, Netherlands.
Borealis Centre for Environment & Trade Research (2007), Environmental Trends and Climate
Impacts: Findings from the US Book Industry, Research Report Commissioned by the Book
Industry Study Group and Green Press Initiative.
Brown, M., Rizet, C., Anderson, S., Allen, J. and Keı
ta, B. (2005), “Life cycle assessment in the
supply chain: a review and case study”, Transport Reviews, Vol. 26 No. 6, pp. 761-82.
Carbon Trust (2006), Carbon Footprints in the Supply Chain, Carbon Trust, London.
Carbon Trust (2008), Product Carbon Footprinting: The New Business Opportunity, Experience
from Leading Companies, Carbon Trust, London.
Corpus, G. and Peachman, J. (2003), “Measuring the impacts of internet usage on travel behaviour
in the Sydney Household Travel Survey”, paper presented at the Australasia Transport
Research Forum Conference, Wellington, October.
European Commission (2007), “Results of the review of the community strategy to reduce CO
emissions from passenger cars and light-commercial vehicles”, Communication from the
Commission to the Council and the European Parliament, COM(2007) 19 ﬁnal.
Future Foundation (2008), In-home Eating Habits and Food Shopping, Future Foundation Report,
Future Foundation, Bellevue, WA.
Green Press Initiative (2008), Reducing Climate Impacts: A Guide for the Book and Newspaper
Industries, Green Press Initiative.
Vehicle Agency Certiﬁcation (2008), Emissions, available at: www.vcacarfueldata.org.uk (accessed
12 October 2008.
About the authors
Julia B. Edwards is a Research Associate in the Logistics Research Centre at Heriot-Watt
University, Edinburgh. She joined the Group in 2006 as part of the multi-university
“Green Logistics” project. Prior to that, she was a Senior Lecturer of Environmental Management
at the University of Wales, Newport. She has been researching and teaching in the areas of
transport and environmental issues for the last ﬁfteen years. Currently, her research interests
include carbon auditing of supply chains, e-commerce and the environment and consumer
travel and shopping behaviour. Julia B. Edwards is the corresponding author and can be
contacted at: firstname.lastname@example.org
Alan C. McKinnon is a Professor of Logistics and a Director of the Logistics Research Centre
at Heriot-Watt University, Edinburgh. A graduate of the universities of Aberdeen, British
Columbia and London, he has been researching and teaching logistics for 30 years and has
published extensively on many different aspects of the subject. He has conducted studies for
numerous public and private sector organisations and been an adviser to several UK
Government departments, parliamentary committees and international agencies. Much of his
current research relates to the decarbonisation of logistical activity.
Sharon L. Cullinane has continued to lecture, research and publish in the ﬁeld of transport
policy and the environment around the world, since gaining her PhD in logistics 20 years ago
from Plymouth University. Previous to that she has been employed at the University of
Hong Kong, Oxford University, the Egyptian National Institute of Transport, the Ecole
Superieur de Rennes and Plymouth University. She is now an independent consultant. She is
widely published internationally. She has lectured, researched and published in logistics and
transport around the world. Her most recent post was as a Senior Lecturer at Heriot-Watt
University in Edinburgh, UK.
Analysis of the
To purchase reprints of this article please e-mail: email@example.com
Or visit our web site for further details: www.emeraldinsight.com/reprints