Opportunities and Challenges of Interactive
Public Displays as an Advertising Medium
Rui José
University of Minho, Information Systems Department, rui@dsi.uminho.pt
Jorge C. S. Cardoso
Portuguese Catholic University, School of Arts / CITAR, jorgecardoso@ieee.org
Abstract Advertising is often the key element in the business case for public
display networks. However, this is still a limited medium in its support for key ad-
vertising concepts, such as targeting or impact assessment. This chapter analyses
how the evolution of the medium is creating the opportunity for re-shaping adver-
tising models for Digital Signage. In particular, we consider how the increasing
availability of sensing and interaction opportunities may generate the necessary
digital footprints that will help to characterise advertising opportunities in Digital
Signage and measure impact. A generic footprint model would enable advertising
models to be specified at a higher level of abstraction, enabling them to be applied
across global display networks composed by very diverse display types, managed
by many different entities, and serving many different purposes. We propose a de-
sign space that relates the multiple types of digital footprints that can be generated
with multiple modes of campaign targeting and impact assessment, providing the
ground for a fundamental shift from advertising based on measuring attention to
advertising based on active engagement with people.
1 Introduction
Public digital displays are an increasingly ubiquitous element in our visual culture.
We can find them in all sorts of public and semi-public spaces, where they can be
serving multiple types of informative purposes. However, the main driver behind
most deployments has been their use as an advertising medium. Digital Signage
has attracted much interest as a marketing tool because displays can be close to
shopping decision points, they can leverage on situations of strong attention avail-
ability, e.g. queues, and they can reinforce messages from other media. Advertis-

2
ing is thus almost always an integral part of their business case, which revolves
around the revenue obtained from selling screen real estate and time to advertisers.
In many cases, this revenue is indeed the sole purpose for setting up the network,
but even if that is not the case, advertising is still very often an important way to
off-set deployment costs and to guarantee the sustainability of the services.
While the revenue generated from advertising in these networks has been stead-
ily increasing, and the medium continues to attract the interest of marketers, the
evolution of this market has been somewhat restrained by key limitations of the
medium with major implications for advertising. The first is that most public dis-
plays offer very little in ways of interacting with and responding to the people
around them. This absence of systematic interactive or sensing features means that
displays fail to produce meaningful traces of how they are used. The second is that
the advertising process is essentially based on selling undifferentiated screen real
estate within the scope of a single display network. As a result, there is no signifi-
cant impact assessment and targeting is limited to the selection of the display net-
works in which the campaign is to be placed. Without effective solutions to these
key issues, digital agencies will not embrace Digital Signage as a mature and wor-
thy advertising medium with which they can have a clear and sustainable Return
on Investment (ROI). Moreover, web marketing has been generating solid adver-
tising models that are enabling that medium to become increasingly central in
marketing strategies. The sophistication of web advertising provides an inspiring
model for other digital media, but it also means that marketers will have increased
expectations regarding the advertising models for Digital Signage and other digital
media. Failing to live up to those expectations could potentially mean the fallout
of the respective business case. These concerns have been widely acknowledged
by the Digital Signage industry, which has been making an effort to address them,
mainly through more advanced audience metering techniques. Also, a large num-
ber of patents have been filled in recent years on targeted advertising for public
displays, including major companies such as Google, Microsoft, IBM, NEC or
HP, reflecting a significant interest in this topic. With this quickly evolving land-
scape, there are many challenges to be addressed, but there is also a major poten-
tial for a significant evolution of Digital Signage towards more mature advertising
models.
In this chapter, we discuss the opportunities and challenges for the evolution of
Digital Signage towards a mainstream advertising medium for public spaces. We
start by introducing key advertising concepts and how they are addressed across
different advertising media. We then analyse the properties of Digital Signage and
discuss how this medium can evolve to meet the increasingly high expectations of
marketers. We then introduce the concept of digital footprint as a high-level ab-
straction for representing the traces left behind when people implicitly or explicit-
ly interact with Digital Signage. This is the basis for a design space that associates
various sensing and interaction features with the types of advertising functionality
they can enable.

3
2 Pervasive Networks of Public Displays as a New Advertising
Medium
Advertising is a key element of the promotion mix of companies and covers a
wide range of communication tools in order to inform and persuade. As in any
other form of communication, advertising effectiveness depends on understanding
the communication process (encoding, decoding and feedback) and elements
(sender, receiver and media). Developing an advertising process involves at least
four major steps, namely: setting advertising objectives; budget; message; and
media (Kotler and Armstrong 1998). Advertising will thus occur as part of adver-
tising campaigns with specific communication goals that should determine the
structure of the campaign, its intended audience, its duration, and the media used
to support it, very often more than one. A specific budget will delimit the range of
opportunities to be considered and will necessarily lead to optimisations that are
able to produce more results with fewer costs. The economic valuation of advertis-
ing opportunities needs to be established based on the size, characteristics and sit-
uation of the potential audience. Impact assessment will measure the reaction of
the target segments to the message of the campaign. Measuring the effectiveness
of the campaigns is crucial for control and for validating the return on the invest-
ment. When performed in real time, impact assessment may also enable the cam-
paign strategy to be adapted on the fly, according to the reactions it is generating.
The execution of campaigns, especially in digital media, is normally supported by
some type of campaign management service that supports the campaign specifica-
tion (what to present in what circumstances), the actual delivery of the adverts
(generating impressions and tracking actions) and some type of analytics (report-
ing on the results of the campaign).
A key part of the advertising process is thus the identification of the segments
that are the primary targets for the message and the ability to direct the message to
this target audience. This delivery of the right message to the right audience and in
the right context is essential because it allows campaigns to be executed in a much
more cost-effective way. If an advert is only shown to those people or in those sit-
uations for which it might be more relevant, it is much more likely to produce
sales. A targeting process can be very diverse and refined, depending on the avail-
ability of information about the targets and the contexts in which they are consum-
ing the message. With the emergence of digital media, marketers are increasingly
empowered with a very rich set of information about people and their behaviour,
enabling a broad range of new targeting techniques that may include not only the
selection of the targets, but also the dynamic adaptation of the message to increase
its relevance.
Demographic targeting is the classical segmentation approach that considers
properties such as the age group, gender, the residential area or the education lev-
el. Demographic data can be a reasonable predictor for consuming behaviour and
requires less information than other approaches.

4
Behavioural targeting tries to model people based on their past behaviour. It
can be extremely sophisticated, depending on the number and nature of the behav-
ioural data sources, and it has become increasingly important with web advertis-
ing, where it is often possible to collect substantial information about browsing
patterns. For example, with information about previous visits to other web sites,
normally obtained through cookies, advertisers can much more easily select the
most relevant adverts to show to that person when entering another web site. Re-
targeting is another form of behavioural targeting whereby a person is shown ad-
verts from a previously visited web site. The idea is that a person who has already
demonstrated a certain level of interest on the products of that web site is more
likely to generate sales.
Geographic targeting is also an important parameter for segmentation. With a
small business running a local campaign this may involve selecting local media or
global media with support for fine-grained geographical scoping. In global cam-
paigns on digital media, geographical targeting may be used to adapt the message
to different locations, considering for example the respective country, local lan-
guage or time of the day.
Contextual targeting involves selecting places for the adverts that are related in
content to the products. For example, a magazine about video-games would be a
relevant context for an advert about a new video game. A web-site about babies
would be a relevant context for adverts about baby food. Targeting occurs, not on-
ly because the message is being directed to people who seem to have an interest in
those topics, but also because it is presented at a time when they are actively
thinking about them. Contextual targeting may also explore not only the type of
place where the advert is being shown, but also its potential credibility with the
audience or even the affinity that people may have with that place.
2.1 Advertising Models Across Other Media
Advertising models have evolved very significantly in recent years, especially
with the new possibilities emerging from the web, but audiences and impact can
be expressed very differently depending on the supporting medium. Multiple ad-
vertising models have thus emerged to address the specificities of each medium.
This section analyses several advertising media, uncovering their approaches re-
garding targeted advertising, impact assessment and advert valuation. This will
frame our discussion on the advertising models for Digital Signage.
2.1.1 Television
Television is a medium with solid and well-established advertising models that
have sustained and driven that industry for many years. Audience measurement is
a high-profile activity strongly associated with the prestige of the TV shows. The

5
measurement unit is the number of exposures by time unit, with the audience data
being gathered through phone surveys, diaries or even dedicated audience meter-
ing devices that detect and communicate which channels are being watched. Ad-
vertisers want their adverts to be seen by as many people as possible, but they also
want to optimise the process by focusing more on their main target groups. There-
fore, audience data is complemented with demographic data about viewers, ena-
bling audiences to be determined, not only as a global number, but also according
to multiple demographic parameters, such as gender, income, age and residential
area. As a result, the valuation of adverts also considers the type of viewers. As
certain segments may be more valued than others, a larger audience does not nec-
essarily correspond to a higher price. Impact assessment, albeit limited, is typical-
ly achieved through surveys on brand recall.
2.1.2 Static Displays
Static displays differ from the other media being analysed in that they are an
example of a static medium. With static media it does not really matter when peo-
ple are watching because the message always remains the same. Therefore, audi-
ence measurement is mainly based on measuring foot traffic. This can be achieved
manually or using people counters, small sensors that can count the number of
people that walk through doors or virtual gates. Even though this is not effective
in determining how many persons are actually paying any attention to the display,
it is still effective enough to establish a reasonable valuation model for this medi-
um. More elaborate techniques may also include visits to assess display visibility
and obtain refined demographic data about visitors (Philport 2007).
2.1.3 Web
The emergence of the web raised many challenges, but also immense opportu-
nities for advertising. What clearly distinguishes the web from other media in this
analysis is its inherently interactive nature, and the consequent ability to obtain
rich information about user context and behaviour. As a user visits particular web
sites or places search queries for particular keywords, these actions can be used to
infer interest in particular topics and select appropriate adverts or sponsored links.
These characteristics were first explored by Google AdWords and Google Ad-
Sense to address two fundamental issues that had previously undermined web ad-
vertising: establishing a valuation for adverts and assessing their impact. The first
issue involves determining which ads may be closer to the interests of the poten-
tial audience in order to target advertisements. The main alternatives include the
geographic location, the context of the page being visited or some knowledge
about user behaviour. The second issue involves assessing impact by determining
which ads have actually managed to lead people into the desired action. In the web
this is mostly achieved by measuring Click-Through Rates (CTR).

6
Despite being a relatively new medium, the web has now become a thriving
environment for the emergence of new and sophisticated advertising models with
implications for other digital media. The first is that exposure is no longer seen as
the single or even the main element in determining the valuation of adverts. The
focus has shifted from merely determining who receives the message towards
measuring how many people were interested or even how many people actually
bought something. This immediate feedback also means that it is now possible to
try many different campaign strategies and evaluate on-line which ones are more
effective. As a consequence, valuation models have also evolved from the tradi-
tional Pay-per-Impression, based on exposure, to models based on the achieved
impact, such as the Pay-per-Click (PPC), in which payment is determined by the
number of clicks generated by the advert. The economic value associated with
each click as well as the allocation of advertisements to specific advertising oppor-
tunities are typically managed by automated bidding systems. The allocation algo-
rithms try to maximise the value generated to all the parties involved by seeking to
allocate adverts to the opportunities that maximise the generation of clicks, while
subject to the limitation of capital available for the campaign.
Another major characteristic of web advertising is that it can benefit both large
traffic web sites as well as the small blog page. The PPC model and the automated
media planning system have created the ground for open and automated cam-
paigns in which any web site can benefit from advertising revenue and any cam-
paign can reach any web site. A web site does not need to attract a high volume of
traffic to be considered a relevant advertising medium. Through affiliate pro-
grams, even a small website can offer relevant advertising opportunities and gen-
erate advertising revenue, as long as it performs well in attracting some niche
segment.
2.1.4 Social Networking Sites
Social Networking Sites (SNS) have extended even further the already im-
mense possibilities of web advertising by enabling campaigns targeted at very re-
fined user profiles. Campaigns are mainly centred on the demographics of users,
which may include detailed elements such as geography, age, job, company size
or gender. Information about preferences or likes and the overall history of previ-
ous actions are also increasingly part of the targeting process. Additionally, social
networks are particularly strong enablers for Relationship Marketing in which a
brand aims to promote longer term and stronger ties with consumers in a way that
extends far beyond traditional marketing messages.

7
2.2 Specificities of Digital Signage as an Advertising Medium
The evolution of the Digital Signage landscape in terms of technology, scale,
openness and cooperation models provides a promising path towards the evolution
of the medium and its support for more sophisticated advertising models. In par-
ticular, we consider two main evolutions that will provide fundamental contribu-
tions towards that objective: the emergence of open advertising markets for public
displays and the systematic availability of sensing or interaction features.
Currently, the Digital Signage advertising market is very fragmented. There are
many isolated display networks, each providing its own separate set of advertising
opportunities and negotiation procedures. An advertiser who wants to create a
large campaign may need to individually negotiate with multiple network opera-
tors to buy advertising space.
As a result, only large networks can have the necessary scale to be of any inter-
est to advertisers. This limitation is being addressed with the emergence of adver-
tising aggregators, such as SeeSaw, BookingDOOH, Vukunet, Argo Digital Solu-
tion or AdCentricity. These services collect information about the advertising
opportunities across multiple display networks and provide a media planning ser-
vice for matching adverts with advertising opportunities. This is opening the op-
portunity for automated allocation across multiple networks and for a global and
open market for advertising in Digital Signage.
Systematically enriching Digital Signage with sensing and interaction capabili-
ties represents the other major step towards new advertising models for Digital
Signage. Public displays, whether digital or not, can be watched without generat-
ing any information on their usage, but without such information there can be no
effective solution to the key issues of targeted advertising and impact measure-
ment. Displays that react to the presence of people and offer them the possibility
to interact will strongly enhance the user experience by enabling people, not just
to be exposed to adverts, but to engage with content in a much more meaningful
way. This increased engagement will implicitly or explicitly generate information
for characterising social contexts and the advertising opportunity they represent.
Even considering that the resulting information may never be as rich as click-
through patterns on the web, it would still represent a major step forward when
compared with the idea that displays are just to be seen by passive observers.
These evolutions may take the advertising models for Digital Signage one step
closer to the rich advertising models supported by web advertising. However, one
key observation from the analysis of advertising models across other media is also
that the characteristics of the medium are central to the evolution of the respective
advertising models. Therefore, despite the major role played by web advertising in
setting the expectations for Digital Signage, we must also understand the specifici-
ties of Digital Signage and their implications in defining what the most appropri-
ate model might be.

8
2.2.1 Proximity
Public displays are part of a physical and social environment that strongly deter-
mines their communication context. This means that not only geographic location,
the type of venue, or the demographics of visitors should be considered for target-
ed advertising, but also other physical and social proximity elements. For exam-
ple, two places sharing a significant percentage of visitors may be targeted in an
integrated way enabling the mutual exchange of adverts or providing the ground
for Life Pattern Marketing approaches (Zweben 2007) in which the same message
is reinforced by being present, even if slightly adapted, in the multiple locations
that a person may visit throughout the day. Also, nearby displays may act collec-
tively to provide a consistent and strongly situated user experience, such as dy-
namic guidance services (Strohbach et al. 2009) that suggest directions to nearby
businesses.
2.2.2 Place-sensitive
While web browsing is essentially an individual activity, Digital Signage is part
of a social and physical setting in which adverts are seen in a shared environment.
An effective advertising model must acknowledge the role of Digital Signage as
an element with a potentially strong impact in shaping the places where the dis-
plays are located. It should, therefore, give the people responsible for the place in-
creased control over the nature of the adverts displayed. If those adverts are not
aligned with the place’s values, practices or commercial strategies, their public
presentation may become a source of embarrassment.
2.2.3 Generate Engagement
It has been argued in this chapter that a strong user engagement will be crucial
for stronger advertising models in Digital Signage. However, achieving the type
and level of user engagement that will enable audience characterisation and im-
pact assessment is not very natural for this medium. It will require specific ap-
proaches for bringing people into interacting with public displays in ways that are
both meaningful for people and relevant for advertising. Examples may include
specific types of games, quizzes or polls, specifically designed for that purpose or
models of shared control over the display in which increased engagement is re-
warded with increased control. It may also mean that content adaptation should go
beyond targeted advertising and be extended to other forms of content. Adaptive
content will make the display more valuable to people, which will be more willing
to express their preferences to influence displayed content rather than displayed
adverts. For example, the ability to select content may enable targeted adverts to
be injected into the selected content, while being valued by people.

9
2.2.4 Crowd Targeting
Public displays will typically have multiple simultaneous users and the adapta-
tion process must consider the best strategy for dealing with the potentially varied
interests expressed by those people. This generates a trade-off between the selec-
tion of adverts based on a profile combing the multiple interests of the multiple
persons present and the selection based the use of each individual profile, one at
the time (Alt et al. 2009). The first is a balanced approach, but faces the risk of not
really matching anyone’s specific interests. The second approach can be targeted
for each individual, but it raises additional privacy issues and may conflict with
the idea of Digital Signage as a place-making tool.
3 A Design Space for Advertising in Digital Signage
This section will now consider how we can map multiple forms of sensing and
interaction into the characterisation of advertising opportunities and their assess-
ment. The challenge is that Digital Signage is likely to support multiple interaction
modalities, from mobile technologies to touch-sensitive surfaces, and also multi-
ple types of interaction semantics, in many cases associated with specific applica-
tions. Making sense of the myriad sensing and interaction events that can occur
across a display network requires higher-level abstractions that are independent
from the specific affordances or the semantics of the interaction offered by partic-
ular displays. Such abstractions should be able to play for Digital Signage a simi-
lar role to the one that click-through patterns and search phrases play for the web.
We propose a design space created around the concept of digital footprint as an
abstraction for representing the traces left behind when people implicitly or ex-
plicitly interact with Digital Signage. A digital footprint focuses on the nature of
the information generated about the interaction itself and abstracts away from the
particular interaction modality being used or the particular application semantics
in which the interaction occurred. They aim to represent preferences, characteris-
tics and behaviour, regardless of how they are expressed. We have structured digi-
tal footprints around 4 main categories: presence sensing, self-exposure, user-
generated content and actionables, as represented in Table 1.
Table 1: Digital footprints
Presence sensing
(Detection, Characterisation, Identification) Being physically there
Self-exposure Explicitly managing exposed identity
User-generated content Pushing content to the system
Actionables
(Interactive, External)
Responding to the system

10
These categories represent not only different types of interaction, but also in-
creasingly higher levels of engagement with the display system.
The design space that we propose associates the various types of sensing and
interaction features with the type of digital footprint they can generate and maps
these into the types of advertising functionality they can enable. This should in-
form Digital Signage creators about the relationship between sensing and interac-
tion features and their possible role in the advertising process.
We will now describe these footprints in more detail, analysing how they can
be generated and how they can contribute to the advertising process.
3.1 Presence Sensing
The ability to implicitly collect information about the presence of nearby peo-
ple is an important element for characterising situations. Despite its stronger po-
tential for generating rich information about usage, interaction events are neces-
sarily sparser than presence events, as there will normally be many more people
present than those actively engaged with the display at any given moment. Pres-
ence sensing is thus critical to enable targeted advertising to occur even when
there is no one interacting. Presence sensing may involve increasingly complex
levels of presence information, more specifically: the ability to simply detect
presences (detection), the ability to characterise those presences (characterisation)
and the ability to associate presences with unique entities across multiple sessions
(identification).
3.1.1 Presence Detection
Presence detection is the most basic level of presence information in which the
system is simply able to detect whether or not there is someone nearby, and possi-
bly at what distance. Multiple off-the-shelf sensors can be used for this purpose.
Distance to the display can be determined by combining presence sensors with dif-
ferent sensing ranges or using distance sensors that report distance. Computer vi-
sion techniques, such as frame differencing, can also be used to sense movement.
Information about the presence of someone near a display, even without know-
ing who or how many, can mainly be used as a trigger for presenting specific con-
tent, and particularly as part of an attraction loop designed to entice people to in-
teract when passing-by. In Community Wall (Grasso et al. 2003) frame
differencing was used as part of the presence detection system to prevent the dis-
play from changing content while someone was reading from it. Knowing the dis-
tance can be important for the display system because there is normally a strong
correlation between distance and the awareness level that people may have about
the display. If they are in a short distance, it is much more likely that the display is
currently the focus of their attention. In the Range whiteboard (Ju et al. 2008), for

11
example, an infrared distance sensor was used to determine the distance of people
using a whiteboard application.
The digital footprint generated by a presence detection mechanism is essential-
ly a presence/absence pattern that may only provide a very basic characterisation
of the place in terms of people flow.
3.1.2 Presence Characterisation
Presence characterisation is the ability to count and possible determine particu-
lar characteristics about the presences detected near the display, which broadly
corresponds to the concept of audience measurement. A vast range of audience
measurement technologies has been emerging to estimate, not only the number of
people in front of a display, but also to determine their attention span or inferring
some type of characteristic about them, such as age or gender (Quividi 2010;
Trumedia 2010). This normally involves placing a video camera on the display,
typically on the top and facing the audience, and processing the images to generate
reports about the number, attention span, and characteristics of viewers. In an at-
tempt to set industry standards for these reports, the Digital Place-based Advertis-
ing Association (DPAA) (formerly OVAB) has produced a guidelines document
that describes the Average Unit Audience as “the number and type of people ex-
posed to the media vehicle with an opportunity to see a unit of time equal to the
typical advertising unit” (Spaeth et al. 2008). These guidelines have a clear focus
on what is normally called the Opportunity to See (OTS) and on three qualifying
characteristics associated with that opportunity: Presence, Notice and Dwell time.
Audience metering techniques are very important for the Digital Signage indus-
try because they may help to justify and valuate advertisement investment. How-
ever, they assume a model in which the objective is simply to estimate the number
of people potentially exposed to an advert, without considering what their real en-
gagement might be. Presence characterisation can also be used for targeting ad-
verts according to the characteristics of the current audience, as for example in the
case of the Eye Flavour that selects adverts based on age and gender (NEC 2009).
3.1.3 Presence Identification
Presence identification is the ability to identify unique identities within
presences and thus recognise different visits by the same person. A common ap-
proach involves the use of a personal device as a proxy for the person. Bluetooth
has been extensively used for this purpose, as it is widely available and enables
many mobile phones to be discovered and uniquely identified through their Blue-
tooth address. Radio Frequency Identification (RFID) tags can also be used. They
are small and can easily be incorporated into many existing artefacts, as exempli-
fied by the IntelliBadge system (Cox et al. 2003) in which conference participants

12
were given RFID augmented badges and then tracked through the conference
rooms. A display would then generate several visualisations of the resulting data.
Presence identification raises considerable privacy issues and therefore it
should only occur as part of a process in which people are informed and value the
extra functionality that the existence of this information may be providing. The
ability to enable or disable presence identification according to the circumstances
would be equally important. This is a point in which Bluetooth, which allows dis-
covery to be easily disabled, takes considerable advantage over RFID tags.
The digital footprints resulting from presence identification correspond to indi-
vidual presence information and open many new opportunities for the advertising
process. They may serve to build individual profiles that characterise people ac-
cording to their presence patterns, e.g. regular visitors vs. first-time visitors. They
can be used to optimise advertising by selecting adverts that have not yet been
shown to the people that are currently present, as exemplified by the BluScreen
system (Sharifi et al. 2006) that uses Bluetooth detection to avoid showing adver-
tisements to users more than once. If a network of displays is involved, presence
identification may also serve to show some variant of the same advert to the same
person at different locations, thus reinforcing the message, or to select adverts
based on where the person has been before. The existence of multiple sensing
points would also enable new forms of impact assessment. For example, it may be
possible to measure how many people who have been shown an advert about a
nearby store will then go to that store, or simply how the number of first-time visi-
tors has increased as the result of a campaign.
3.2 Self-exposure
Self-exposure is the ability of the display system to collect information that
people have explicitly created to represent their characteristics, preferences or per-
sonal tags. This personal information can take many forms, such as a reference to
a user’s personal webpage, the user id in a Social Networking Site or a purposely
created profile. Self-exposure should also be seen as way for people to take con-
trol over how their identity is projected in the public space by the Digital Signage.
It should be part of a process whereby people expose something about them be-
cause they understand and value the implicit response of the display to that self-
exposure. Moreover, the intended level of self-exposure may be strongly influ-
enced by a broad range of circumstances. Therefore, a key feature for self-
exposure mechanisms is some type of control on how that exposure should occur
at any given moment. This means not just the ability to enable or disable presence
detection, but also the ability to dynamically manage self-exposure in a way that is
easily controlled and understandable. Self-exposure differs from presence in the
sense that it is an explicit act of identity management, while with presence we as-
sume that all the information was implicitly generated. It differs from interaction
in the sense that there is no pre-established connection between self-exposure and

13
a specific type of reactive behaviour by the display system. Potentially, different
display systems may react very differently to the same form of self-exposure, or
even the same system may also react differently under different circumstances.
Self-exposure may be achieved by combining presence identification with the a
priori definition of a user profile that is associated with the identity. In Proactive
Displays (McDonald et al. 2008), users attending a conference registered their af-
filiation, interests and personal webpage before the conference day and were given
RFID augmented conference badges at the conference site. Throughout the con-
ference, several displays reacted to the nearby participants showing and creating
associations between their profiles and creating opportunities to socialise. A limi-
tation of this approach is that people could not easily change their profile, and had
no control over what information to expose at any given moment. Another alterna-
tive is the use of a personal information device running a custom application to
manage and expose a profile. This application can connect automatically, or on
demand, to the display system and communicate users’ preferences. This can be
exemplified by the Hello.Wall (Prante et al. 2003), a system in which people
would carry a ViewPort device (a modified PDA) to communicate with an ambi-
ent display. Similarly, in the Camera-Phone (Toye et al. 2004) system, a custom
mobile application can be configured with personal information that is made au-
tomatically available to the display system when a user interacts with it. One ad-
vantage of these approaches is that the information is always available to be up-
dated by its owner and it may be possible to have greater control on how to
manage self-exposure, but they have the disadvantage of requiring a dedicated ap-
plication. The use of Bluetooth names to express self-exposure, as explored by Jo-
se (Jose et al. 2008) and Davies (Davies et al. 2009), is an opportunistic alternative
that is easily available on almost any device. It allows people to enter predefined
commands in the Bluetooth name of their mobile phone. When that person ap-
proaches a display, these commands can be obtained and interpreted as part of the
person’s preferences. For example, Ribeiro (Ribeiro and Jose 2010) has shown
how tags exposed this way by multiple people could be used to select from the In-
ternet content feeds for presentation on a local display.
In all the previous examples, there was the assumption that information would
be specifically created for the purpose of influencing a display system. However, a
very promising alternative is to explore connections with the many Social Net-
working Sites where people already have extensive descriptions about themselves
and their preferences. WhozThat (Beach et al. 2008) uses the SNSs profiles of
people nearby to create context information that can then be used to support spon-
taneous interactions or drive the music selection. People are expected to use a mo-
bile phone running an identity sharing protocol that will advertise their on-line
identities to the other nearby devices. This system does not consider the use of
Digital Signage or any explicit selection of which information to share, but it is an
example of using SNSs profiles as a sort of personal data aura that can be used to
mediate digital self-exposure. (Böhmer and Müller 2010) conducted a study on the
exhibition of SNSs profiles in public settings. Using mockup images they asked
people about their willingness to expose profile information in two types of what

14
they called social signs. The first was a personal social sign projected around the
person and showing parts of the respective profile. The second was an interper-
sonal sign, projected in such a way to link two people and representing some type
of connection between them, such as having a mutual friend or sharing an interest.
The study provides an interesting example of the type of identity projections that
can be generated by these connections with SNSs.
Information associated with self-exposure footprints can be very varied. In its
simplest form, the result may be just a list of keywords generated from multiple
forms of preferences expression. However, with access to profiles, much more
structured information can be obtained. The potential of self-exposure in the ad-
vertising process is particularly relevant for targeted advertising, where depending
on the type of information collected, it may support multiple forms of targeting,
such as demographic, contextual or behavioural.
3.3 User-generated Content
User-generated content is the ability of the system to accept content originating
from the users of the display. This is achieved by allowing people to post their
own content for publishing at the display, either directly or indirectly through a
reference to the content (e.g. a URL). Many displays have been created that sup-
port some variant of this feature. WebWall (Ferscha and Vogl 2002), for example,
allows people to submit content using SMS, email or a web interface. The Plasma
Poster (Churchill et al. 2004) system supports content (photos, text, web pages)
submission through two interfaces: email and a web form. MMS has also been ex-
tensively used as an input interface for display systems. For example, the Joe
Blogg project (Martin et al. 2006) includes a display designed in the form of an in-
teractive artwork where people can send pictures and text messages through MMS
or SMS. Bluetooth can also be used to push content to a display system using ei-
ther standard OBEX exchanges or custom mobile applications. Both Hermes
(Cheverst et al. 2005) and Snap and Grab (Maunder et al. 2007) use the OBEX
feature to enable users to send pictures and other media to a display. In both cases,
the user just selects the content on his mobile phone, selects the “send via Blue-
tooth” command and selects a particularly named device. An example of a custom
Bluetooth application for sending content is Touch & Interact (Hardy and Rukzio
2008), which allows users to choose a picture in their mobile phone and touch the
display in the position they want to place that picture. The main advantage of a
custom application is that it can be built to interact specifically with a given dis-
play, thus allowing a richer and more convenient interaction. For example, the ap-
plication may be able to automatically determine the Bluetooth address of the dis-
play system, alleviating the user from the task of manually searching for nearby
devices and selecting the correct device name. The application may also be able to
determine beforehand what type of content is acceptable by the display system,
impeding the user from sending content that will be rejected. OBEX on the other

15
hand, has the obvious advantage of not requiring the installation of any additional
software in the mobile device, and thus enable more opportunistic interactions.
Also, from a developer’s perspective, OBEX is more attractive because, given the
multitude of different devices, it is often the case that multiple versions of the ap-
plication have to be created to deal with the hardware and software variants in de-
vices. Also, these applications must be somehow distributed to users which may
be yet another barrier to usage.
When generating content to a display, users are implicitly associating that con-
tent with that particular place. The nature of the digital footprints generated by
such process depends on the type of content. However, the analysis of the content
and its meta-data should produce a relevant characterisation of that place and its
local practices. This is already common on the web, where a vast and sophisticat-
ed range of techniques has been emerging to analyse user-generated content. For
example, tags in Flickr images have been used to extract place semantics (Ratten-
bury and Naaman 2009), and tweets have been used to determine mood and emo-
tions (Bollen et al. 2009). If displays were able to generate enough content from
users, similar techniques could also be applied to generate footprints that would be
very relevant for targeted advertising processes.
3.4 Actionables
Content on Digital Signage is often some type of actionable, a message or an
interactive feature intended to cause people to act (Müller et al. 2007). Actionables
provide a specific demand for action, such as downloading a content item, select-
ing one of several available options or reading a 2D code from the display. An ac-
tionable footprint represents a reaction to one of those actionables.
Generating this type of footprint requires some mechanism for tracking the ac-
tions taken and also some type of system-wide reference for the actionables, e.g. a
URL or some unique Id. The existence of a unique reference that can be tracked
system-wide separates actionables from the myriad interactive features that may
exist in a display system and which in many cases will not be meaningful outside
the specific application in which they are available.
Actionables may provide one of the most promising paths for advertising in
Digital Signage. They enable several forms of automated impact assessment and
advert valuation, ultimately leading to something similar to the pay-per-click con-
cept. When considering how to generate actionable footprints, we need to take into
account two major types of actionable: interactive actionables in which the reac-
tion occurs within the system and can thus be automatically detected by the sys-
tem; and external actionables in which the action that the actionable is promoting
occurs outside the context of the display system and cannot be tracked directly by
the system itself.

16
3.4.1 Interactive Actionables
Interactive actionables are invitations to interaction that can be interpreted by
the system. They involve giving users some type of control and being able to track
their options. Allowing people to pull content from the display system is a com-
mon example of an interactive actionable. Knowing who has downloaded which
items enables the system to infer interest on the content items that are being of-
fered. Giving people some type of control over the system behaviour may also be
used to generate interactive actionables. For example, Jukola (O'Hara et al. 2004),
which allows people in a bar to vote on the next music to be played by selecting
music from a list, is an example of a system that could generated this type of foot-
print. Rating content items allows people to explicitly say they like or dislike an
item. Ratings give the display system information about the popularity of content
items. It is also possible to conceive a Games With A Purpose approach (Ahn and
Dabbish 2008) in which all sorts of polls, quizzes, questionnaires, or games are
designed to provide engaging experiences, while allowing the display to collect
users’ preferences and interests. These actionable footprints may be anonymous or
identifiable depending on the interaction mechanism provided, but the key point is
that the meaning of the reactions can be interpreted system-wide and not just with-
in the scope of a particular application.
3.4.2 External Actionables
With external actionables, the intended action is to occur outside the scope of
the display system, and therefore, the system is unable to automatically assess
their efficiency. At best, their efficiency can be assessed through some type of off-
line mechanisms, such as the collection of redeemed coupons.
MobiDiC (Müller and Krüger 2009) is an example of the use of digital coupons
in public display advertisement. The public display shows adverts with coupons
that people can photograph with their mobile phone and redeem at the associated
shop to receive the announced item. The time and location where the ad was dis-
played is encoded in the coupon. When the coupon is redeemed and the shop
owner enters the code back to the system, the specific advert and display are iden-
tified and the aggregated results can tell how successful the advert was.
3.5 Analysis
Our goal with this design space was to inform the process of mapping multiple
types of sensing or interactive features in Digital Signage into specific advertising
models. The digital footprints that have been presented support the first part of
that mapping. By focusing on the essence of the information generated from sens-
ing and interaction features, they provide a framework for reflecting on the mean-

17
ings of interaction without being caught by the specificities of any particular inter-
action mechanisms. In this final analysis, we discuss the second part of that map-
ping, which is to describe what types of digital footprints may be needed to sup-
port specific pervasive advertising goals. We hope that, by making the relationship
between advertising goals and footprints more explicit, this framework may con-
tribute to focus the design of public displays on the data generation objectives,
while avoiding the pitfalls of focusing too much on a particular application or in-
teraction technique.
Table 2 presents a summary of these mappings. Several types of advertising goals
are listed and associated with the specific types of digital footprint that could sup-
port them.
Table 2: Mapping between advertising goals and digital footprints
Advertising
goals
Digital footprints
Attraction
loops
Presence detection can trigger specific content to get peo-
ple’s attention
Audience
measurement
Presence characterisation may provide sophisticated reports
about viewers and their attention
Demographic
Targeting
Presence characterisation may infer basic demographic in-
formation from presences, but self-exposure and specific ac-
tionables may generate even richer information, given that
they involve people explicitly saying something about
themselves.
Contextual
Targeting
Self-exposure and specific actionables may generate key-
word sets and other types of aggregate descriptions that will
help to dynamically characterise the context of a public dis-
plays in ways that are relevant for contextual targeting, such
as hot keywords, social situations or on-going local activi-
ties.
Behavioural
Targeting
Presence identification and identifiable actionables generate
rich information about people’s behaviour and enable mul-
tiple types of behavioural targeting. Presence identification
will mainly revolve around presence patterns, possibly at
multiple locations, whereas actionables will revolve around
expressed preferences inferred from actions.
Optimise
individual
exposure
Presence identification supports several types of optimisa-
tions related with specific individuals, from avoiding repeti-
tions to purposely generating periodic repetitions for rein-
forcement.
Impact
Assessment
Actionable footprints provide the most appropriate measure
for engagement and therefore for really understanding the
reaction of people to the message

18
Based on this framework, Pervasive Advertising solutions may start by setting
their specific goals, consider the digital footprints that the display systems may
need to generate, and finally conceive sensing or interactive features that are ca-
pable of generating the appropriate information while providing an adequate user
experience. This last step should also consider that these different digital foot-
prints are not completely independent from each other. They are more like a stack
of increasingly richer information about the audience of a public display. For ex-
ample, availability of presence characterisation also generates presence detection.
Presence identification is implicit in self-exposure and identifiable interactive ac-
tionables. The correct choice of the interactive features may optimise the genera-
tion of the most appropriate set of digital footprints.
4 Conclusions
Advertising models define key properties of the advertising process within a
particular medium. More specifically, they should address key issues such as how
to target the message to the most relevant opportunities and how to measure the
impact of a campaign. Advertising models for Digital Signage are still very lim-
ited regarding these two key points, especially when compared with the rich and
sophisticated web advertising models that are increasingly part of the mainstream
advertising industry. Interactive displays and the emergence of open display net-
works are two trends that will necessarily re-shape the medium and will also open
entirely new possibilities for advertising.
We have presented a design space that maps between interaction mechanisms
and their contribution to the generation of digital footprints that are relevant for
targeted advertising. While this design space has been conceived mainly for Digi-
tal Signage, many of these same concepts would also apply to other situated forms
of situated marketing, such as Bluetooth Marketing. Overall, the emergence of a
generic footprint model would represent a fundamental shift from advertising
based on measuring attention to advertising based on active engagement with peo-
ple. This would in turn provide a major contribution to overcome some of the
most fundamental limitations in current models for advertising in Digital Signage.
Acknowledgments
Jorge Cardoso has been supported by “Fundação para a Ciência e Tecnologia”
and “Programa Operacional Ciência e Inovação 2010” (POCI 2010), co-funded by
the Portuguese Government and European Union by FEDER Program and by
“Fundação para a Ciência e Tecnologia” training grant SFRH/BD/47354/2008.
The research leading to these results has also received funding from the Euro-
pean Union Seventh Framework Programme (FP7/2007-2013) under grant agree-
ment no. 244011.(PD-Net).

19
References
1. Ahn L, Dabbish L (2008) Designing games with a purpose. Commun ACM, vol 51, no 8, pp
58-67
2. Alt F et al (2009) Adaptive User Profiles in Pervasive Advertising Environments. In Proceed-
ings of the European Conference on Ambient Intelligence, pp. 276-286
3. Beach A et al (2008) WhozThat? Evolving an ecosystem for context-aware mobile social net-
works. Network, IEEE, vol 22, no 4, pp 55, 50, Jul. 2008
4. Böhmer M, Müller J (2010) Users' Opinions on Public Displays that Aim to Increase Social
Cohesion. In Proceedings of The 6th International Conference on Intelligent Environments.
Kuala Lumpur 2010, Malaysia; to appear.
5. Bollen J, Pepe P, Mao H (2009) Modeling public mood and emotion: Twitter sentiment and
socio-economic phenomena. CoRR, abs/0911.1583, 2009
6. Cheverst K et al (2005) Exploring bluetooth based mobile phone interaction with the hermes
photo display. In Proceedings of the 7th international conference on Human computer inter-
action with mobile devices & services, vol 111, Salzburg, Austria: ACM Press, pp 47 - 54
7. Churchill EF, Nelson L, Denoue L, Helfman J, Murphy P (2004) Sharing multimedia content
with interactive public displays: a case study. In Conference on Designing interactive sys-
tems: processes, practices, methods, and techniques, pp 7 - 16
8. Cox D, Kindratenko V, Pointer D (2003) IntelliBadge. In First International workshop on
Ubiquitous Systems for Supporting Social Interaction and Face-to-Face Communication in
Public Places
9. Davies N, Friday A, Newman P, Rutlidge S, Storz O (2009) Using bluetooth device names to
support interaction in smart environments. In Proceedings of the 7th international conference
on Mobile systems, applications, and services, pp 151-164
10. Ferscha A, Vogl S (2002) The WebWall. In Workshop on Collaboration with Interactive
Walls and Tables
11. Grasso A, Muehlenbrock M, Roulland F, Snowdon D (2003) Supporting communities of
practice with large screen displays. In Public and Situated Displays - Social and Interactional
Aspects of Shared Display Technologies, Kluwer, pp. 261-282
12. Hardy R, Rukzio R (2008) Touch & interact: touch-based interaction of mobile phones with
displays. In MobileHCI '08: Proceedings of the 10th international conference on Human
computer interaction with mobile devices and services, pp 245-254
13. Jose R, Otero N, Izadi S, Harper S (2008) Instant Places: Using Bluetooth for Situated Inter-
action in Public Displays. Pervasive Computing, IEEE, vol 7, no 4, pp 52-57
14. Ju W, Lee BA, Klemmer SR (2008) Range: exploring implicit interaction through electronic
whiteboard design. In CSCW '08: Proceedings of the ACM 2008 conference on Computer
supported cooperative work, pp. 17-26
15. Kotler P, Armstrong G (1998) Principles of Marketing. 8th ed. Prentice Hall College Div
16. Martin K, Penn A, Gavin L (2006) Engaging with a situated display via picture messaging. In
CHI '06 extended abstracts on Human factors in computing systems, pp 1079-1084
17. Maunder A, Marsden G, Harper R (2007) Creating and sharing multi-media packages using
large situated public displays and mobile phones. In Proceedings of the 9th international con-
ference on Human computer interaction with mobile devices and services, pp 222-225
18. McDonald DW, McCarthy JF, Soroczak S, Nguyen DH, Rashid AM (2008) Proactive dis-
plays: Supporting awareness in fluid social environments. ACM Trans Comput-Hum Interact,
vol. 14, no. 4, pp 1-31
19. Müller J, Krüger A (2009) MobiDiC: Context Adaptive Digital Signage with Coupons. In
Ambient Intelligence, vol 5859, M. Tscheligi et al, Eds. Springer Berlin / Heidelberg, pp 24-
33
20. Müller J, Paczkowski O, Krüger A (2007) : Situated Public News & Reminder Displays.
Proceedings of European Conference on Ambient Intelligence 2007, Darmstadt

20
21. NEC (2009) NEC Launches Eye Flavor, Japan’s First All-in-One Digital Signage Board with
Face Recognition Technology. http://www.pr-inside.com/nec-launches-eye-flavor-japan-s-
first-r1030399.htm . Accessed 9 November 2010
22. O'Hara K, Lipson M, Jansen M, Unger A, Jeffries H, Macer P (2004) Jukola: democratic mu-
sic choice in a public space. In Proceedings of the 5th conference on Designing interactive
systems: processes, practices, methods, and techniques, pp 145-154
23. Philport JC (2007) TAB Eyes On: Out Of Home Media Measurement,
http://www.eyesonratings.com/images/pdf/recent-publications/TAB-Eyes-On-OOH-Media-
Measurement-by-Philport.pdf
24. Prante T et al (2003) Hello.Wall - Beyond Ambient Displays. In Adjunct Proceedings of
UbiComp 2003, pp. 277--278
25. Quividi (2010), Quividi: Automated Audience Measurement of Billboards and Out Of Home
Digital Media.http://www.quividi.com/. Accessed 9 November 2010
26. Rattenbury T, Naaman M (2009) Methods for extracting place semantics from Flickr tags.
ACM Trans Web, vol 3, no 1, pp 1-30
27. Ribeiro FR, Jose R (2010) Autonomous and Context-Aware Scheduling for Public Displays
Using Place-Based Tag Clouds. In Ambient Intelligence and Future Trends -: International
Symposium on Ambient Intelligence (Isami 2010)
28. Sharifi M, Payne T, David E (2006) Public Display Advertising Based on Bluetooth Device
Presence. In Mobile Interaction with the Real World (MIRW 2006) in conjunction with the
8th International Conference on Human Computer Interaction with Mobile Devices and Ser-
vices
29. Spaeth J, Singer S, and Hordeychuk M (2008) Audience Metrics Guidelines. Digital Place-
Based Advertising Association. http://www.dp-aa.org/guidelines.php. Accessed 9 November
2010
30. Strohbach M, Kovacs E, Martin M (2009) Towards Pervasively Adapting Display Networks.
In 1st International Workshop on Pervasive Advertising, held in conjunction with Pervasive
2009
31. Toye E, Madhavapeddy A, Sharp R, Scott D, Blackwell A, Upton E (2004) Using camera-
phones to interact with context-aware mobile services. University of Cambridge, Computer
Laboratory
32. TrueMedia (2010), TrueMedia: iTALLY: Opportunity to See (OTS) People Counter.
http://www.trumedia.co.il. Accessed 9 November 2010
33. Zweben M (2007) Life Pattern Marketing: A new way to reach people Out-of-Home. SeeSaw
Networks. http://www.seesawnetworks.com/services/whitepapers/life-pattern-marketing/.
Accessed 9 November 2010