2013 karlin ford squiers taxonomy-libre

1 23

Energy Efficiency ISSN 1570-646X Energy EfficiencyDOI 10.1007/s12053-013-9227-5

Energy feedback technology: a review andtaxonomy of products and platforms

Beth Karlin, Rebecca Ford & CassandraSquiers

1 23

Your article is protected by copyright and all

rights are held exclusively by Springer Science

+Business Media Dordrecht. This e-offprint

is for personal use only and shall not be self-

archived in electronic repositories. If you wish

to self-archive your article, please use the

accepted manuscript version for posting on

your own website. You may further deposit

the accepted manuscript version in any

repository, provided it is only made publicly

available 12 months after official publication

or later and provided acknowledgement is

given to the original source of publication

and a link is inserted to the published article

on Springer's website. The link must be

accompanied by the following text: "The final

publication is available at link.springer.com”.

ORIGINAL ARTICLE

Energy feedback technology: a review and taxonomyof products and platforms

Beth Karlin & Rebecca Ford & Cassandra Squiers

Received: 17 December 2012 /Accepted: 10 September 2013# Springer Science+Business Media Dordrecht 2013

Abstract Feedback is a promising strategy for energyconservation, and many energy feedback products (i.e.technologies with hardware) and platforms (i.e. technol-ogies without hardware) have emerged on the market inrecent years. Past research suggests that the effective-ness of feedback varies based on distinct characteristics,and proposes categories to better understand and distin-guish between these characteristics. A review of existingcategories, however, identified the following issues: (1)current structures group feedback technologies into four(or fewer) categories, making device distinction andselection onerous; (2) current categories often ignoretechnical and psychological distinctions of interest toresearchers; and (3) none provide a systematic descrip-tion of the specific characteristics that vary by category.This paper presents a classification structure of feedbacktechnology, derived theoretically from a review of rele-vant literature and empirically via content analysis of196 devices. A taxonomy structure of feedback technol-ogy was derived based on the characteristics of hard-ware, communications, control, display, and data collec-tion. The resulting taxonomy included the followingnine categories: (1) information platform, (2) manage-ment platform, (3) appliance monitor, (4) load monitor,(5) grid display, (6) sensor display, (7) networked sen-sor, (8) closed management network, and (9) open

management network. These categories are mutual ex-clusive and exhaustive of the identified technologiescollected and are based on characteristics, which areboth stable and important to feedback provision. It ishoped that this feedback classification will be of use toboth researchers and practitioners trying to leverage thepotential of these technologies for energy conservation.

Keywords Feedback . Conservation . Electricity .

Technology

Feedback refers to the process of giving people infor-mation about their behaviour that can be used to rein-force behaviour and/or suggest behaviour change. It isconsidered an important dimension of behaviour change(Skinner 1938; Bandura 1969) and has been used toinfluence behaviour in a wide variety of fields, includingeducation, public health and consumer research.

This emphasis has received increasing attention inrecent years due to the rapidly changing energy infra-structure in the developed and developing world.Advances in data sensing, storage and disseminationhave made it possible for information about behaviourto be collected, stored, and presented to consumers atspeeds and on scales that were previously impossible.“Adding sensors to the feedback equation helps solveproblems of friction and scale. They automate thecapture of behavioural data, digitising it so it can bereadily crunched and transformed as necessary. Andthey allow passive measurement, eliminating the needfor tedious active monitoring” (Goetz 2011). Such

Energy Efficiency

DOI 10.1007/s12053-013-9227-5

B. Karlin (*) :C. SquiersUniversity of California, Irvine, CA, USAe-mail: [email protected]

R. FordUniversity of Otago, Otago, New Zealand

Author's personal copy

changes in data collection also require changes in datastorage—Austin Energy, for example, increased yearlydata storage from 20TB to 200TB for just 500,000 m(Danahy 2009). Finally, changes in data presentationare being seen in the form of ambient displays,gamificaiton strategies and innovative dashboard de-signs for both mobile and web platforms.

Within the realm of energy efficiency, changes ininfrastructure are enabling this transition on a widescale. Countries are spending billions of dollarsupgrading the current electrical grid with what is re-ferred to as the “smart grid”; a network of controls,computers, automation and new technologies that en-able sensing of and response to conditions on the trans-mission lines, as well as two-way communication be-tween utilities and customers. One important compo-nent of this is the replacement of traditional electricitymeters with advanced metering infrastructure, or “smartmeters”, which are defined as “a metering system thatrecords customer consumption (and possibly other pa-rameters) hourly or more frequently and provides fordaily or more frequent transmittal of measurements overa communication network to a central collection point”(Federal Energy Regulatory Commission 2008, p. 5).Currently, <10 % of the world’s meters are considered“smart”, but this number is expected to change rapidly.In the USA alone, smart meters have already beeninstalled in over 25 million homes and an estimated 65million will be installed by 2020, serving over 50 % ofUS Households (Institute for Electric Efficiency 2011).Likewise, the European Union Directives aim for 80 %coverage by 2020 (Faruqui et al. 2010; Sánchez 2012).

Both the public and private sectors have recognisedthe potential for increased information provision and arecreating and supporting new technologies to providefeedback about energy use to consumers. The USWhite House recently launched the Green ButtonInitiative to encourage utilities to provide consumerswith real-time access to their energy information andpromote private sector development of technologies thatintegrate with this initiative (Chopra 2011). Additionally,a variety of companies, ranging from major players suchas General Electric and Panasonic to start-ups such asOPOWER and Navetas are creating new technologies toprovide energy feedback to consumers, both directlythrough hardware as well as through integration withsmart meter technology.

Programs like the Green Button Initiative, as well asthe 200+ feedback products (i.e. technologies with

hardware) and platforms (i.e. technologies withouthardware) that have emerged, are based on the ideathat receiving information about energy use leads tobetter decisions about energy use. As the use of elec-tricity in the home is “abstract, invisible and untouch-able” (Fischer 2008, p. 80), feedback has beenhypothesised to serve a vital function in making thisenergy use visible and interpretable to the consumer.Reviews of energy feedback research have found thatthe provision of feedback leads to average savings of10 % (r=.1179, p<.001), with effects ranging fromnegative effects (i.e. increase in energy consumption)to up to 20 % in energy reduction (Darby 2006;Ehrhardt-Martinez et al. 2010; EPRI 2009; Fischer2008; Karlin and Zinger 2013); others argue that arealistic effect may be closer to 3–5 % (McKerracherand Torriti 2013). These studies suggest that the effec-tiveness of feedback varies based on type, and proposecategories to better understand and distinguish be-tween these types. However, current classifications offeedback lack the technological sophistication to ac-count for the diversity in available products and plat-forms. The goal of this paper is to analyse currentenergy feedback technologies in the marketplace andpresent a comprehensive taxonomy of feedback tech-nology based on product characteristics.

We start by introducing past psychological and tech-nical approaches to understanding feedback. Next, wereview previous literature on energy feedback, focus-ing on past attempts to define, describe and categorisefeedback. Then, we introduce and describe our studymethodology, which is a content analysis and classifi-cation of currently available feedback technology.Using data collected from 196 feedback technologies(both products and platforms), we present a list ofenergy feedback characteristics and identify key char-acteristics for categorization. Finally, we present ataxonomy structure of energy feedback that incorpo-rates these key characteristics.

As feedback technologies are becoming increasinglyubiquitous in our society, with a growing capacity toleverage personalised energy information, there is anurgency to ensure that they are utilised to their fullpotential. Whilst there is a growing body of researchand evaluation on the potential effectiveness of feedbackin consumer trials, there has been little research into theactual products and platforms available in the market-place. This paper extends previous literature in that thetaxonomy presented is derived both theoretically and

Energy Efficiency


empirically and all categories are designed to be mutu-ally exhaustive and mutually exclusive, given currenttechnological capabilities. It is hoped that this report willassist both researchers and practitioners in the fields ofenergy efficiency and conservation and that it may serveas the basis for publicly available product informationon feedback technology as this market grows in futureyears.

Literature review

Early feedback—cybernetics

Feedback has been studied in both the physical andsocial sciences for decades (e.g. Skinner 1938; Wiener1948). The basic premise is simple: Feedback enablesthe output of a dynamic system or process (i.e. onewhose behaviour varies over time) to be compared to agoal or reference point, in order to enable improvedcontrol over that system or process (Goyal and Bakshi2008). Figure 1 illustrates the difference in the struc-ture and control of a process when no feedback isprovided (a), and when feedback is provided (b).

First applied to steam engines and other mechanicalsystems in the eighteenth century, feedback systems arebased on control theory; which has three key aspects: (1)

a goal or reference point to which the system is con-trolled with respect to, (2) a means to compare actualperformance to the goal or purpose, and (3) a process toenable information about the output of the system to becommunicated back to the input to enable modificationof the process (Duffy 1984). Improved control overdynamic systems is thus enabled by the presence offeedback loops and the communication of information(Åström and Murray 2009). A simple example of thistype of system is a home heating, ventilating, and airconditioning system. The system is set up with a desiredtemperature. The temperature sensor on the thermostatenables the temperature of the building to be measuredand compared to this desired state. The difference anddirection between the desired and actual temperature isthen communicated back to the controller, which in turnactivates the system to minimise any differences be-tween the actual and desired state.

In the 1940s, Norbert Wiener explored how thesetypes of communication and control theories might beextended to human systems. This, he argued, called for anew science of feedback, human behaviour, and infor-mation, which was coined “cybernetics” (Wiener 1948).Cybernetics is fundamentally concerned with the studyof how information may be manipulated and communi-cated around dynamic systems for the purpose of con-trol, with a particular focus on behaviour, processes and

Fig. 1 Control of dynamic systems with and without feedback, adapted from Goyal and Bakshi (2008)

Energy Efficiency


circular communication (Duffy 1984). In cyberneticsystems that integrate humans into the control process,the resulting system is often non-mechanical and moreflexible than machine-only control counterparts, butalthough it is more complex it operates in much thesame way (Klein 1989). The movement towards goalsor reference points requires individuals to identify theircurrent behaviour with respect to the established refer-ence, which may require more than just a simple me-chanical sensing of the existing environment. The dis-crepancy between their current behaviour and referencepoint needs to be evaluated, and a mechanism needs tobe employed to reduce this discrepancy (Klein 1989;Lawrence et al. 2002).

In the past 40 years, the use of cybernetic systemsfor feedback and control have been implemented andvalidated across many disciplines, including psychol-ogy, epidemiology, environmental studies, engineer-ing, economics and more (Goetz 2011). Although theyhave proven to be very popular methodologies, thereare concerns with their use in social systems, particu-larly in cases where goals or references points may notexist, where accomplishments may not be measurable,or where the information provided cannot be used(Hofstede 1978). Psychological theory may, therefore,serve to integrate system principles of cybernetics withthe complex systems related to human behaviour. Thenext section will discuss this integration.

Psychological theories of feedback

The earliest cited psychological theory of feedback isThorndike’s Law of Effect (1927), which asserts thatthose behaviours that produce a positive effect are morelikely to be repeated and behaviours that produce anegative effect are less likely to be repeated. Skinner’smodel of operant conditioning (1938) expanded on this,stating that knowledge of positive results could be seenas a positive reinforcement and knowledge of negativeresults as a punishment, thus serving to encourage ordiscourage subsequent behaviour. Albert Bandura(1969) is also noted for seminal theoretical work onfeedback; he found that providing a goal and informationabout progress towards that goal could serve as a form ofbehaviour modification, much like a reward or punish-ment. Subsequent theories based on this early workinclude control theory (Carver and Scheier 1981), goalsetting theory (Locke and Latham 1990), and actionidentification theory (Vallacher and Wegner 1987).

Kluger and DeNisi (1996) conducted a meta-analysis of feedback interventions across a wide vari-ety of behaviours and introduced the FeedbackIntervention Theory (FIT) based on their analysis ofpast empirical and theoretical contributions. FIT arguesthat behaviour is regulated by comparisons to pre-existing or intervention-provided goals or standards.These standards can be personal goals or comparisonsto past behaviour or others in a social group. This is acommon concept in previous feedback theories, in-cluding goal setting theory (Latham and Locke 1991)and control theory (Carver and Scheier 1981), whichboth assert that behaviour is generally goal directedand that people use feedback to evaluate their behav-iour in relation to their goals. When behaviour differsfrom the standard, this creates what they call afeedback-standard gap, and it is the desire to decreasethis feedback-standard gap that mediates the effective-ness of feedback.

Four options are available to individuals when pro-vided with a feedback-standard gap. They can respondby changing behaviour to match the standard, changingthe standard to match behaviour, rejecting the feedbackor leaving the situation altogether. Both the strength ofthe goal and the size and direction of the feedback-standard gap can impact this choice. As control theorywould suggest, negative feedback is more likely to leadto behaviour change than positive feedback (Campionand Lord 1982; Podsakoff and Farh 1989).

However, it is also possible for positive feedback tolead to behaviour change. Action identification theory(Vallacher and Wegner 1987) asserts that people iden-tify actions with specific meaning related to their iden-tity. Thus, the action of taking a test in class can beconstrued as answering exam questions or as workingtowards future career success. As people learn a taskand develop mastery, performance becomes automaticand they are able to begin thinking about it in terms ofhigher levels of meaning. This is presented in contrastto a strictly control theory, or cybernetic, model offeedback, in which the goal of reducing the feedback-standard gap remains constant. This explains why pos-itive feedback can also impact behaviour even thoughit does not serve to reduce a feedback-standard gap. Insuch a case, a higher-level goal can be created andtherefore create a new standard.

Finally, FIT suggests that feedback is effective inso-far as it changes the locus of attention of the individualto the feedback-standard gap. Feedback may direct

Energy Efficiency


attention to a specific goal or behaviour that was notpreviously the focus of attention. Thus, it can serve notonly to provide information about a behaviour-standardgap but also to draw attention to a specific behaviour inthe first place. The visibility and availability of feedbackis also an essential element and serves as a key factor inits effectiveness.

This framework provides an introduction to thedifferent components involved in feedback that haveclear implications for feedback interventions across awide variety of behavioural domains. Applying theseprinciples to energy feedback can increase our under-standing of the most important variables upon which tocategorise such technologies. The next section reviewsprevious literature on feedback in an attempt to linkbasic feedback theory to the specific case of energyfeedback.

Past research on energy feedback

Over a hundred empirical studies of energy feedbackhave been conducted over the past 40 years and over200 articles have been published about energy feedbackduring that time. Reviews of this research have foundthat feedback is effective, on average, with effects rang-ing from increases in energy use to savings of over 20%(Darby 2006; Ehrhardt-Martinez et al. 2010; EPRI2009; Fischer 2008). However, definitions, descriptionsand categorisation of feedback vary from report to re-port. Operational definitions for energy feedback, aswell as key characteristics and categories, are lacking.Before presenting our own analysis, we review pastattempts to define, describe and categorise energy feed-back. For the purposes of our discussion, we define acharacteristic as a single variable with two or morelevels, and a category or type as a group of products orplatforms that share one or more characteristics.

Definitions of feedback

Whilst research on energy feedback is abundant, thereseems to be gap in the literature regarding a specificoperational definition of energy feedback. Both Darby(2006) and EPRI (2009) rely on dictionary definitionsof feedback; although technically accurate, they are notspecific either to energy or to consumer-facing infor-mation (e.g. that which involves people in the process).EPRI (2009) and Abrahamse et al. (2005) further char-acterise energy feedback as household-specific

electricity consumption information and Ehrhardt-Martinez et al. (2010) define feedback in the contextof consequence strategies for behaviour change, which“attempt to change behaviour by influencing the deter-minants of a behaviour after the behaviour in questionhas been performed” (p. 38). These definitions focusthe definition on home energy use and incorporatemotivational element of feedback, but are still vaguewith regards to what kind of information constitutesfeedback.

Without a clear operational definition, it is difficult todetermine what exactly distinguishes feedback fromenergy information or control. Areas of ambiguity in-clude (but are not limited to) estimated feedback (e.g.carbon calculators, based on user input) and automated

systems (e.g. appliances that receive and respond tofeedback directly, removing the user from the loop).Within the literature, the feedback system relates to theenergy consumption of a dwelling; therefore, an energyfeedback technology is one that receives informationabout the actual energy consumption of the dwelling(or part of the dwelling). Likewise, definitions focus onthe role of feedback in informing consumers and affect-ing behaviour; therefore, a definition should includeprovision of this energy data back to the consumer.Therefore, we define energy feedback as informationabout actual energy use that is collected in some wayand provided back to the energy consumer.

Reviewing this definition helps to provide some clar-ity to ambiguous areas in the past literature. Estimatedfeedback, which collects approximated energy usageinformation from the user (therefore not actual energyuse), and automated systems that completely remove theuser from the feedback loop (i.e. not providing data backto the energy consumer) are, consequently, not classifiedas a feedback technology.

Characteristics of feedback

Several authors over the past decades have discussedspecific characteristics of feedback that may be mosteffective in promoting energy conservation or distingui-shing between types of feedback technologies (seeTable 1). The most commonly cited characteristic wasimmediacy (Darby 2001, 2006; Donnelly 2010;Ehrhardt-Martinez et al. 2010; EPRI 2009; LaMarcheet al. 2011; Stein and Enbar 2006), which breaks downfeedback into the two categories of direct (immediate)and indirect (not immediate). Additional characteristics

Energy Efficiency


Table 1 Characteristics identified during literature review

Characteristic name Characteristic levels Characteristic definition References

Immediacya Direct, indirect How soon after an action feedback is provided Darby 2006; Donnelly 2010; Ehrhardt-Martinez et al. 2010; EPRI 2009;LaMarche et al. 2011; Stein and Enbar2006

Data collectiona Estimated Feedback, Sensor How feedback information is collected EPRI 2009; Hochwallner and Lang 2009;LaMarche et al. 2011

Frequency Continuous, daily, weekly, monthly, bimonthly How often feedback is given Fischer 2008; Fitzpatrick and Smith 2009;Froehlich 2009

Duration Weeks, months, years How long feedback is provide Fischer 2008

Content (Measurement unit) Electricity, cost, environmental impact,temperature, utility messages

The units of measurement the feedback is given in Fischer 2008; Fitzpatrick and Smith 2009;Froehlich 2009; Herter and Wayland2009; Stein and Enbar 2006

Breakdown (Data granularity) Room, appliance/device level, time of day,building, indoor/outdoor, rate period

The resolution of the feedback data Fischer 2008; Fitzpatrick and Smith 2009;Froehlich 2009; Herter and Wayland2009; Hochwallner and Lang 2009

Presentation mode (Visual design) Numeric, graphic, ambient, artistic The format feedback is presented in Fischer 2008; Fitzpatrick and Smith 2009;Froehlich 2009;Wood and Newborough2007b

Presentation medium Electronic media, written material, in homedisplay, mobile apps, web portals and socialmedia

Themedium throughwhich feedback is presented Fischer 2008; Froehlich 2009; Hochwallnerand Lang 2009; LaMarche et al. 2011

Comparisons Historical, normative, forecast, other buildings,other appliances, other rates, other rateperiods, personal goals

Whether feedback is measured against somestandard

Wood and Newborough 2007b; Fischer2008; Fitzpatrick and Smith 2009;Froehlich 2009; Herter and Wayland2009

Additional information and other instruments(Recommending action)

Incentives, goals, personal commitments, advice Whether information other than usage Fischer 2008; Froehlich 2009; Shutlz2010

Location Activity-based displays, embedded displays,central displays, localised, independent

Where the feedback display is found Wood and Newborough (2007b); Fitzpatrickand Smith 2009; Froehlich 2009

Push/Pull Push, pull Whether the feedback is sent to the user or theuser navigates to it

Froehlich 2009

Control device (Automation) Central, device level, on-board, low automa-tion, high automation, no automation

Whether the feedback system enables control Donnelly 2010; Ehrhardt-Martinez et al.2010; LaMarche et al. 2011

Feedback level Low-level feedback, high-level feedback Whether the feedback is specific to an action orsummative

Froehlich et al. 2010

Communications Fixed, wireless, gateways, range extenders,home area networks

Devices used to enable data transformation Hochwallner and Lang 2009; LaMarcheet al. 2011

Communication protocol X10, UPB, Insteon, Z-Wave, Zigbee Standards used to enable data transmission LaMarche et al. 2011

aCharacteristic names not used explicitly by authors

Energy

Efficiency

Au

tho

r's perso

na

l cop

y

relate to the frequency and duration of feedback collec-tion and provision, the type of information provided andthe messages used, and variables related to both thevisual display and hardware components of feedback.

It is important to note that non-technology factors canalso impact the effectiveness of feedback. Literaturereviews have identified several such factors, includingstudy duration, sample size, and recruitment method(Darby 2006; Ehrhardt-Martinez et al. 2010; EPRI2009; Fischer 2008). Statistical meta-analysis of 42feedback studies support many of these claims, identi-fying several non-technological variables that moderatethe effectiveness of feedback, including study duration,frequency of provision, combination with goal-settingand incentives and the population from which the studysample was drawn (Karlin and Zinger 2013). As most ofthese analyses are conducted from between-study (vs.within-study) comparisons, further research is needed toclarify the role of these non-technological variables,both apart from and in conjunction with any identifieddifference in treatment effect associated with the type offeedback, as discussed in this paper.

The current paper focuses on the type of feedbackproduct or platform used, which has been identified asone of the key variables moderating the effects of feed-back. Several authors have proposed specific categories,or types, of feedback to help distinguish among themany available technologies available. Although specif-ic terms are used to describe different types of energyfeedback systems, they are not always clearly definedand authors may use different terms to describe similarfunctions, or similar terms to describe different func-tions. Before developing and presenting our taxonomystructure of feedback technologies, we discuss currenttypes and typologies of energy feedback.

Types of feedback

The most commonly cited types of feedback are directand indirect feedback. Darby (2001, 2006) uses the termindirect feedback to refer to frequent utility bills, basedon accurate usage data. EPRI (2009) uses it to categoriseboth standard and enhanced billing (billing with addi-tional information and advice) as well as estimatedappliance specific feedback (e.g. through the use ofhome audit software). The American Council for anEnergy Efficient Economy (ACEEE) (Ehrhardt-Martinez et al. 2010) distinguishes between indirectfeedback provided by the utility (offering improved

customer service, better outage, power quality, morefrequent meter readings and feedback to customers),and indirect feedback provided by vendors (offeringimproved feedback information, advice, estimated dis-aggregation, goal setting capabilities and social andhistoric comparisons).

In contrast to indirect feedback, Darby (2001, 2006)defines direct feedback as feedback that is “immediate,from the meter or an associated display monitor” and“available on demand”. EPRI (2009) defines directfeedback as “feedback that is provided real-time ornear-real-time”. The ACEEE build on this to furtherstate that direct feedback systems “provide energy useinformation at the time of consumption (or shortly afterconsumption)” (Ehrhardt-Martinez et al. 2010). Theterms direct feedback and real-time (or near real-time) feedback are therefore taken to be synonymous.

The terms in-home display, in home energy display(Ehrhardt-Martinez et al. 2010), and home energy dis-play (LaMarche et al. 2011), are all used to refer to anindependent display that provides real-time energyconsumption information. These systems tend to becomprised of a sensor as well as a display, whichcommunicate wirelessly. The sensors tend to use cur-rent clamps to monitor the home’s main circuit panel,though some systems use optical sensors to track thepower meter. They tend to provide whole home energyfeedback, though some systems have extra clamps formeasuring individual circuits and are therefore capableof providing a breakdown by circuit (Donnelly 2010).Darby (2006) uses the terminology direct displays,which depicts a free-standing display, supplementalto the electricity meter, providing information on elec-tricity and gas consumption.

Wood and Newborough (2007a) use the term Energy

Consumption Display to refer to anything that providesenergy feedback using a technological format. They fur-ther distinguish between central displays (i.e. displaysplaced in a central location in the home) and activitybased displays (i.e. displays located next to the activityabout which feedback is provided). Activity-based dis-plays, defined as devices which sit between the walloutlet and an appliance (or group of appliances) andmeasure the energy consumption of that appliance (orgroup of appliances), have also been called plug in elec-tricity usage monitors and watt-meters (Hochwallner andLang 2009), plug monitors, outlet level monitors andoutlet readers (LaMarche et al. 2011), plug-in devices(Fitzpatrick and Smith 2009) and distributed direct

Energy Efficiency


sensors (Froehlich et al. 2010). When these plug-loadmonitors also offer control or automation, they are some-times called smart plugs/sockets/outlets/strips (Donnelly2010; LaMarche et al. 2011), a type of smart device.

Other types of smart devices incorporate novel sens-ing and control algorithms for direct feedback andautomation (Badami and Chbat 1998); these includesmart thermostats, smart lights and smart appliances(Ehrhardt-Martinez et al. 2010). The most basic smartdevices have sensing and/or communicating network-ing chips, enabling data collection and automation;more advanced options enable higher degrees of auto-mation with wireless two-way utility communicationfor demand management control, delayed start func-tions and pricing signal control (Donnelly 2010).

Often, smart devices form part of a Home AreaNetwork (HAN). Donnelly (2010) uses the termsHome Automation Network and Home Area Networkinterchangeably and states that the simplest HAN is asmart thermostat that enables heating/cooling controland communicates with a central computer and/or theutility’s metering system. However, she notes that acomplete HAN includes (1) smart devices withembedded/attached networking and/or communicatingchips for automation, (2) advanced network systems andsoftware using mesh networks to provide measurementand feedback of appliance specific data, (3) the potentialfor two-way communication with the utility and (4)some kind of consumer interface for direct, real-timefeedback. Hochwallner (2009) defines a home automa-tion system as one that “consists of “smart” devices anda communication bus that connects all devices in ahome”. The communications bus is used to both controlappliances, and to receive information from the appli-ances about their current power consumption.

Typologies of feedback

Darby (2001, 2006) proposed a typology of feedbackthat focused on direct and indirect feedback but alsoincluded three additional categories: inadvertent feed-back (learning by association, e.g. through micro-generation in the home or community), utility-controlled feedback and energy audits. EPRI (2009)subcategorised feedback into six types: four indirectand two direct. They divide indirect feedback into (1)standard billing: traditional source of feedback thathouseholds receive from their utility company, gener-ally in the form of a monthly bill or statement; (2)

enhanced billing: provision of more detailed informa-tion about consumption patterns from the utility, suchas historical or social comparison statistics; (3) esti-mated feedback: data supplied by the user is analysedto produce estimates of aggregate (e.g. carbon foot-print) or disaggregated (e.g. appliance-specific) usage;and (4) daily/weekly periodic: presents energy infor-mation to the user that is time-delayed by a day ormore, but is provided more often than the traditionalenergy bill. Direct feedback is further categorised as(5) real time: delivers overall consumption level on areal-time or near-real-time bill and (6) real-time plus:provides disaggregated (e.g. individual appliance) en-ergy feedback and/or allows users to control appliancesin the home.

The Ehrhardt-Martinez et al. (2010) typology alsobegins with Darby’s definition of indirect and directfeedback and constructs an analogy based on an onionmetaphor. The layers of the onion are defined as: (1)utility delivered (utility bill or website), (2) vendor de-livered (whole home information), (3) deeper contextualinformation (e.g. includes statistical analysis), (4) in-home energy display (real time or nearly real time feed-back), (5) “Smart” devices (e.g. provide simple automa-tion), (6) disaggregated and contextual (informationabout individual appliances) and (7) automation (wholesystems that include disaggregated real-time feedback,home automation and sometimes energy generation andstorage systems). The three outer layers of the onion(1, 2 and 3) correspond to indirect feedbackmechanismsand the three inner layers (4, 5 and 6) correspond todirect feedback, with home automation at the core (7).As the layers of the onion are peeled away, the feedbackbecomes progressively sophisticated.

In a meta-review of feedback studies based on thesecategories, Ehrhardt-Martinez et al. (2010) found “dis-tinct differences in the average and median energysavings associated with different types of feedback”.However, they do note that significant variation existswithin each of the feedback categories. Whilst theauthors attribute this “within category” variation todifferences in study methodology, it is also possiblethat there are significant differences between types offeedback within these broad categories as well. Forexample, within the real-time plus category, a feedbackintervention may or may not be electronic and may ormay not provide appliance-specific information, bothof which are variables which may impact the effective-ness of the feedback intervention.

Energy Efficiency


The classification “taxonomy” proposed byLaMarche et al. (2011) takes a different approach,consisting of three basic categories intended to captureessential components of a typical feedback device suchthat feedback can fall into one or more of these cate-gories: (1) control devices (allow the consumer orutility to actively control energy use), (2) user

Interfaces (provide energy feedback to consumers)and (3) enabling technologies (underlying supportframework). Control devices can be centralised (com-municate with multiple devices), device level (usercontrols a single device) or on-board (control is inte-grated into the device). User interfaces can provideraw, i.e. direct feedback (e.g. real time or historic usagedata), or processed, i.e. indirect feedback (e.g. compar-isons, advise, goal setting). Enabling technologies in-clude sensors, communications (e.g. gateways) andcommunications protocols (e.g. Zigbee).

Limitations of previous research

Whilst the literature to date presents significant evi-dence supporting the potential effectiveness of energyfeedback to promote conservation, further research isstill needed to answer the questions of how and forwhom feedback works best. As seen in the previoussections, the results of past energy feedback studieshave varied, with effects ranging from negative (i.e.increase in energy consumption) to large effect sizes(accounting for nearly 50 % of the variation in energyuse). These results suggest that the effectiveness offeedback varies based on the type of feedbackreceived.

In addition, although past literature reviews haveproposed categories to distinguish between these typesof feedback, current categorisations lack the techno-logical sophistication to account for the diversity inavailable technologies and are not systematic in theirclassification of specific feedback technologies.Classification, or categorisation, is the process ofgrouping like objects into categories based on theirproperties (Cohen and Lefebvre 2005). Categorieswithin a classification structure should be clearly de-fined (e.g. new objects can be easily categorised),mutually exclusive (e.g. each object fits in one andonly one category) and collectively exhaustive (e.g.all objects fit into a category); the result is that everyobject within a classification structure fits in one andonly one category. When categories are based on a

fixed set of characteristics in parallel, the resultingstructure is a typology; when these characteristics areconsidered in succession, the resulting classificationstructure is a taxonomy (Marradi 1990). A review ofexisting classifications (see above) identified the fol-lowing issues:

1. All existing products and platforms are grouped intofour (or fewer) categories, which leaves single cat-egories containing upwards of a hundred technolo-gies, making distinction and selection difficult.

2. Categories focus primarily on the type of informa-tion provided and ignore physical design and op-erating differences.

3. None provides a systematic description of the spe-cific characteristics that vary by type, makingcategorisation of emerging technologies difficult.

The goal of this project is to analyse the currentstatus of feedback technologies in the marketplace witha focus on the specific characteristics of various tech-nologies. This paper presents a taxonomy of techno-logical feedback, derived theoretically and empiricallyfrom a review of 196 feedback technologies coded onover 100 device characteristics.

Methods

The study utilised content analysis and classificationmethodologies to derive a taxonomy of feedback tech-nologies. Content analysis is a technique of compres-sing large amounts of text into a manageable data setby creating and coding the text into categories based ona set of specific definitions (Neuendorf 2001; Stemler2001). Descriptive data on over 200 specific feedbacktechnologies were identified and collected from Marchto August 2011. After similar/identical devices wereremoved, product information was analysed qualita-tively using open coding followed by axial coding;themes were constructed from analysis of the codesin consultation with previous literature (Corbin andStrauss 2007; Creswell 2009). The set of final charac-teristics were screened for relevance and a taxonomystructure was derived to categorise all products andplatforms such that the categories were mutually ex-clusive and mutually exhaustive to the dataset. Thefollowing four sections (data collection, inclusion,coding and analysis) describe the methodology of thisstudy in further detail.

Energy Efficiency


Data collection

The following four methods were used to identify andcollect data about feedback technologies: (1) review ofrelevant literature, (2) Internet keyword search, (3) retailwebsites and (4) personal contacts. As products andplatforms were identified, a raw data file was createdfor each product with any available information (e.g.user manuals, product summaries, new articles, photos).

Data collection began by compiling a list of identi-fied feedback technologies from the following reports:Anderson and White 2009 (7 devices); Ehrhardt-Martinez et al. 2010 (12 devices); Herter 2010 (49devices); Herter and Wayland 2009 (35 devices);Hochwallner and Lang 2009 (4 devices); LaMarcheet al. 2011 (38 devices); Stein 2004 (11 devices); andStein and Enbar 2006 (27 devices). One hundred andone unique feedback devices were identified from the-se reports.

Next, general searches were conducted in Google aswell as Amazon.com using the keywords energy andfeedback simultaneously. In addition to identifying addi-tional technologies, these searches also uncovered third-party websites that specifically market and/or sell feed-back technologies, including: www.powermeterstore.com, www.mymeterstore.com, www.smartgrid.gov andwww.home-energy-metering.com.

Further feedback technologies were identifiedthrough informal inquiries via email and discussionwith colleagues and personal contacts, including col-leagues at our universities, researchers at energy-related conferences and colleagues at the ACEEE.The total number of feedback products and platformsinitially compiled and reviewed using all four of theabove search strategies was 259.

Inclusion

For the purpose of this work, we defined energy feed-back as information about actual energy use that is

collected in some way and provided back to the energy

consumer. As such, our inclusion criteria for determin-ing whether a feedback technology meets this defini-tion were as follows:

1. The feedback technology collects informationabout actual building electricity use.

2. The feedback technology provides this actual us-age data back to the user.

3. The feedback technology is an actual product orprototype (not concept).

4. Sufficient information is available to describe thefeedback technology.

Additionally, this work considers only those com-mercial or pre-commercial ventures in which the com-mercial product or platform provided is energy feed-back; feedback systems provided to consumers by elec-tric utilities, whose primary commercial venture is theretail of energy, are not included here. Commercial andpre-commercial feedback technologies that met all fourof the above criteria of energy feedback were subject toinclusion in our analysis. Among the initial 259 devices/systems collected, we identified 196 that met all of theabove criteria.

Coding

Code development was iterative and utilised the con-stant comparison method and multi-phase coding(Corbin and Strauss 2007; Creswell 2009). Each feed-back technology was treated the unit of analysis forcoding and the content analysis utilised a manifest ap-proach, such that the exact information was pulled fromthe data. An initial set of codes was developed based onprevious literature (e.g. frequency, immediacy, content,medium—see review above); additional codes werecreated, as needed, as technologies were added.Variables relating to key hardware and system propertiesof the feedback technologywere added to ensure that wewere able to account for both physical design and oper-ating conditions (and differences). Further characteris-tics were added in an iterative procedure; opening cod-ing from the 196 products and platforms resulted in atotal of 117 distinct codes, divided into five primarycategories: development, hardware, system, data collec-tion and data presentation.

In the second round, the 196 technologies were re-coded according to these characteristics. All coding wasthen reviewed for accuracy; discrepancies were resolvedthrough discussion between all three authors. In somecases, information being coded for a particular technol-ogy either was not obtainable (e.g. total number ofsubjects contacted) or was somewhat ambiguous (e.g.random assignment); therefore, not all studies could becoded on every variable. When information was missingin a study and there was no clue to support a reasonableestimate, the information was coded as missing data.

Energy Efficiency


http://www.powermeterstore.com

http://www.powermeterstore.com

http://www.mymeterstore.com

http://www.smartgrid.gov

http://www.home-energy-metering.com

Because the coding process involved some degree ofsubjectivity, all study variables were coded by at leasttwo authors and results on 10 % of the data were com-pared for reliability. Inter-rater reliability was acceptablyhigh (kappa >.700) for all variables (Cohen 1960).

Analysis

During axial coding, the 117 distinct codes werereviewed and collapsed into 36 primary characteristics.For example, codes that represented multiple levels ofthe same characteristic were condensed (e.g. Linux,Mac, and Microsoft combined as levels of the charac-teristic, “operating system compatibility”). The nextphase of analysis distinguished those codes related tothe primary goal of the study (e.g. categorising feed-back devices) from others related to the quality ofpersonal preference (Corbin and Strauss 2007). A tax-onomy classification structure based on these charac-teristics was then constructed and all technologies werereviewed for fit. Final categories were derived based onan integration of analysis results with previous litera-ture as well as data regarding the most important devicecharacteristics for consumers.

Findings

Feedback characteristics

After both open and axial coding, 36 feedback character-istics within five broad categories were identified; thesecharacteristics grouped by category are listed in Table 2.

From this list of characteristics, we identified bothtyping characteristics and quality characteristics; qualitycharacteristics are those that pertain to the quality of thefeedback system whilst typing characteristics are thosevariables necessary to distinguishing between categoriesof feedback. For example, some individual technologiesprovide feedback data at multiple levels of granularity,so it made sense to amalgamate the different levels ofgranularity for categorisation purposes, and instead usethis variable to define the quality of the feedback tech-nology within type. To determine the categories for ourtaxonomy, we identified those that were:

1. Stable and inherent to the technology in itself2. Consistently identifiable for at least 80%of the devices3. Theoretically relevant

4. Had an even distribution of products/platformsacross the variable options (i.e. no more than 80 %of devices belonging to one type).

Those that met these criteria are identified in bold textin Table 2. These were further grouped into six primarytaxonomy characteristics: product hardware (sensor units,external transmitters and physical displays), communica-tions (communication protocol), control (appliance con-trol), display (medium of presentation), collection (col-lection point) and protocol (communication protocol).These characteristics, their definitions and levels are listedin Table 3, and further information is provided below.

Hardware

Hardware describes the physicality of the feedback tech-nology, asking whether or not the technology requiresthe purchase of any new sensors, transmitters and/ordisplays. Any system that provides feedback via existingchannels (i.e. does not require the user to purchase newhardware devices) and collects data via existing sources(e.g. utility meters, data loggers) does not have hardware.Any system that requires the purchase of a device ordevices, such as a display or a sensor, has hardware.

Communications

Communications refers to whether or not the physicalcomponent or components of feedback systems are ableto communicate with each other and/or pre-existingelectronic devices. These communications can be eitherwired or wireless. Feedback systems that consist ofvarious hardware components that communicate withone another, or consist of a single hardware componentthat communicates with other electronic devices in thehousehold or building have communications capabili-ties. Individual feedback products that collect and pro-vide data within the same device do not have commu-nications capabilities. This characteristic does not applyfor feedback systems that have no physical components(i.e. have no hardware).

Control

Control refers to whether or not the feedback systemenables remote control of electronic devices within thehome and/or building. This includes automation, i.e.setting devices to turn on or off or change setting at a

Energy Efficiency


specified time, as well as the ability to manually turndevices on and off from a remote location.

Display

Display refers to the physical medium on which thefeedback data is presented to the user. When feedback

is displayed via existing channels, such as a utility bill,website, computer software or phone, it has no display.When feedback is presented on an independent display,whether it is wall mounted or portable, it has an auton-omous display. When the display is built into the devicethat collects feedback data (i.e. the sensor), it is classi-fied as an embedded display.

Table 2 Characteristics identi-fied during coding

aInsufficient data—missing datafor 20 % or more of datasetbInsufficient variation—over80 % of dataset fell intoone category

Category Attribute Category Attribute

Development Status of technology System Technology requirements

Costa OS compatibilitya

Target audience Amount of memorya

Hardware Sensor units Memory locationa

External transmitters Integration w/other systems

Physical displays Documentation availability

Power supply options Data collection Data granularity

Measurement capabilitiesa Collection point

Monitoring channelsa Data presentation Medium of presentation

Measurement resolutiona Display update frequencyb

Voltage/current rangesa Temporal granularityb

Collection update frequencya Comparison messages

Power factor correctiona Units of measurement

Communication channelsa Appliance control

Communications rangea Visualisations used

Communication protocol Level of configurability

Installation protocol Recipient of feedbackb

Additional components Provision of advice

Table 3 Taxonomycharacteristics Characteristic name Characteristic definition Characteristic levels

Hardware Does it have physical hardware? No (platform)

Yes (product)

Communications Does it have communicationsabilities?

No (monitor)

Yes (network)

Control Can it be used to control electronicdevices remotely?

No (information)

Yes (management)

Display What type of display is feedbackpresented on?

None (existing channels)

Embedded (within device)

Autonomous (standalone display)

Collection Where does the data come from? Grid

Sensor

Appliance

Protocol Does the system use proprietarycommunications protocols only?

Yes

No

Energy Efficiency


Collection

Data collection describes where the feedback informa-tion comes from. Data collected by a meter or providedby the utility is classified as grid. Data collected by thefeedback product is classified as sensor. Data thatcomes from an existing home appliance or device, suchas refrigerator or home thermostat, is classified asappliance.

Protocol

Protocol refers to whether or not the feedbacksystem uses only non-standard communication pro-tocols such that it can only communicate withitself. Feedback systems that are capable of usingpublic communications standards can communicatewith other devices, such as smart meters or smartappliances that use the same communication proto-col. Feedback systems that use only proprietary com-munications cannot.

Feedback taxonomy

The six characteristics described in Table 3 wereanalysed with the dataset in order to identify meaningfulcategories. Although a simple factorial typology of thesix variables reveals a possible 144 combinations, manyof these are not possible (e.g. feedback technologies thatdo not have hardware by definition do not have sensorsto collect data); therefore, a taxonomy structure wasderived to create mutually exclusive and exhaustivecategories whilst retaining parsimony in the final con-struction. Figure 2 presents the final taxonomy structureof feedback technology, which was constructed fromanalysis of these characteristics with respect to bothexisting technology as well as past literature on the mostmeaningful characteristics of feedback.

Our taxonomy of feedback technology is comprisedof nine categories, which are divided into two groups,platforms and products. A platform does not require thepurchase of any new hardware; instead, it integrateswith existing hardware that users already have (e.g.

Platform Product

Monitor

ApplianceMonitor

Load Monitor Grid DisplaySensor Display

In HomeDisplay

NetworkedSensor

InformationNetwork

ManagementNetwork

Network

InformationPlatform

ManagementPlatform

Is new hardwarerequired?

Can the technology communicate?

Can devices be remotely controlled?

What type ofdisplay does thesystem have?

Where does thedata come from?

Are only proprietary communications used?

Feedback TechnologyStart

DistributedDisplay

No Yes

N/A

No Yes

No

No

Embedded

Appliance Sensor

Yes

No Yes

Autonomous None

Yes No

Various

Various

Grid Sensor Sensor

ClosedManagement

Network

Open Management

Network

Fig. 2 Taxonomy of energy feedback technology

Energy Efficiency


smart appliances, smart meter) and provides energy usedata to consumers via enhanced energy bills or reports,mobile apps, web browsers or computerised software.Feedback platforms are broken down further into twogroups: information platform and management plat-form. The key difference between an information plat-form and a management platform is that a managementplatform can be used to remotely control appliances andan information platform cannot. Examples of informa-tion platforms include forms of feedback such as en-hanced energy bills and customer web portals, providedby companies such as Opower and Efficiency 2.0PEER. Both of these platforms rely on smart meter datafrom a partner utility, which is then processed andpresented to consumers in a paper-based report and viaan online web portal. Additional services such as com-parisons to other users, advice about energy savingbehaviours, indications of approximated appliance en-ergy consumption and potential savings and rewardsprograms are other characteristics of this type of feed-back and can be used to distinguish between the differ-ent information platforms available to users.

Management platforms allow the user to automate“smart” electronic devices connected to the platform,including smart lights, a smart thermostat or smart appli-ances. Examples of management platforms include SilverSpring Network’s Smart Energy Platform and theFutureDash Greendash Hub. These technologies rely onsmart meters and smart devices already in the home forinformation and control. The information is provided toconsumers via a web-based portal, and users are able toremotely control their smart devices via theweb interface.These devices may also be controlled via a utility deliv-ered demand response program. Figure 3 illustrates thetype of network architecture involved. Different compa-nies use different protocols to transmit information fromthe smart meter and smart devices to the consumer portal.The Silver Spring system provides information to con-sumers via the utility, whilst FutureDash is aims to workwith consumer electronic manufacturers to skip this linkand enable data transfer over the Internet.

Feedback products, unlike platforms, do require theuser to purchase some sort of hardware, and are furthersubdivided into two categories based on whether or not

Fig. 4 Examples of energy load monitors

Smart meter

Smart device

Smart device

Smart device

Web portal

User

Fig. 3 Network architectureof a management platform

Energy Efficiency


they have communications capabilities. A productconsisting of a single hardware component that doesnot communicate data with any other electronic deviceis a monitor. Feedback monitors contain sensors (tocollect energy use data) and a display (to provide databack to users) in a single piece of hardware. A productwith multiple hardware components that communicatewith each other, or that are able to communicate withthird-party electronic devices, is classified as a network.These networks tend to be confined to a physical spacewithin a single building and may therefore be thought ofas local area networks, or, in the residential setting, ashome area networks (Donahue 2007).

Feedback monitors are further broken into two cat-egories, appliance monitors and load monitors, both ofwhich contain inbuilt sensors and embedded displays.Because they are not capable of communications, theydo not enable remote control of appliances; however,some products are fitted with timers that can be pre-programmed to allow some amount of automation.

An appliance monitor collects data from and displaysdata about an individual appliance (i.e. the appliance hasinbuilt energy sensors and an embedded display show-ing this information). Fridges, freezers, washing ma-chines and tumble driers that have an embedded displayto present their energy use to consumers, are all classi-fied as appliance monitors.

A load monitor is a separate piece of hardware thatserves as a proxy between the energy source andenergy-consuming device. Most load monitors collectdata at the plug level (although some collect data at themeter level), and sit between the socket on the wall andthe appliance plug. Some load monitors offer the op-tion of additionally viewing the data on accompanyingcomputer software (facilitated via a USB or SD card

connection). These features, and others such as theviewing options (i.e. ability to see instantaneous powerconsumption, total energy use, energy use over a pre-defined period), memory availability and cost, distin-guish different products in this feedback category.Examples of load monitors include Belkin’s ConserveInsight Monitor, the Kill-a-Watt, and Watts Up, asillustrated in Fig. 4a–c, respectively.

Networks are feedback products that have commu-nications capabilities. Similar to platforms, networksare broken down into two categories based on whetheror not they enable the user to remotely control appli-ances. Information networks do not enable remotecontrol of appliances whereas management networks

do. Information networks are further broken down intoin-home displays (grid displays and sensor displays)and networked sensors, and management platforms arebroken down into open management networks andclosed management networks.

In-home displays are feedback products that have aphysical display providing energy usage information tousers in real (or near real) time and collect data fromeither a sensor (sensor displays) or the user’s electricity

Fig. 6 The TED 5000G, including (left to right) one set ofcurrent clamps and measuring transmitting unit for breaker panelinstallation, and one gateway to transmit data externally

DisplaySensorUtility meter/

elecricity consuming device

Data transmitted

Sensor collects

data

DisplaySmart utility

meter

Data transmitted

Grid Display

Sensor Display

a

b

Fig. 5 Basic system archi-tecture of in-home displays

Energy Efficiency


meter/utility (grid displays). The basic architecture ofgrid displays and sensor displays are shown in Fig. 5aand b, respectively. Some in-home displays offer theoption of additionally viewing the data online or viaaccompanying software.

All grid displays provide whole-home level energyusage information and some receive and display peakdemand pricing and other messages from the utility.AzTech’s In-Home Displays, Ambient’s Energy Joule,and GEO’s Duet II for Smart Meters are all types of griddisplays. The main difference between these displays ishow the information is presented to users (including theunits used, the availability of historical data, features ofthe display in terms of size and colour, and so on) andwhat additional features the display provides, such aspricing information from the utility.

The most common type sensor display consists of atleast one pair of CTclamps, a transmitter, and a portable

display. Whilst most sensor displays transmit data fromsensor units to displays using wireless or powerlinecommunications, some rely on wired communications,using, for example, CAT-5 LAN cable installations.Examples of sensor displays include Current Cost’sENVI, Ewgeco’s Electricity Monitor, and the TED1001. Of all the types of feedback identified in thisresearch, sensor displays formed the largest category,containing 61 separate products.

Networked sensors, the third type of informationnetwork, are feedback products that have a sensor orsensors, but no physical display. Physical sensors col-lect energy usage data and communicate it to externalservers where it is processed for viewing on a webbrowser, app or computer software. The TED 5000 Gseries, depicted in Fig. 6, falls into this category.

The two forms of management platforms, closedmanagement networks and open management networks,

Smartmeter

Smartappliance

Smartthermostat

Web portal

User

Communication network

Physicaldisplay

Appliance

Appliance

Appliance

Sensor

Sensor

Information

Control

Information

Control

Information

Control

Sensor

Utility

Fig. 8 Basic system architecture of open management networks

Utility meter

Appliance

Appliance

Web portal

UserSensor

Sensor

Information

Control

Information

Control

Information

Control

Communication network

Physicaldisplay

Sensor

Fig. 7 Basic system archi-tecture of closed manage-ment networks

Energy Efficiency


are feedback products that enable users to remotelycontrol connected devices. Data are collected from avariety of sources, including smart meters, sensors andsmart appliances, and presented to users on a combina-tion of physical displays and existing channels (i.e. webportals, computer software or mobile app). The definingfeature that distinguishes between closed managementnetworks and open management networks is the type ofcommunication protocol utilised by the system. Closedmanagement networks communicate using only propri-etary protocol and form closed communication networksto which only proprietary devices can join. Open man-agement networks may use proprietary communicationprotocols on some layers, but they are also capable ofusing public communication protocols to form opennetworks to which any device communicating with thesame protocol (e.g. smart meters, smart appliances, etc.)can join. This also means that utilities can send demandresponse signals to devices and appliances on the net-work via the smart meter.

The majority of closed management networks,such as AlertMe Smart Energy, Plugwise andthinkeco Modlet, use plug-in sensors to measure andcontrol the energy usage of plug-in devices, as shown inFig. 7.

The majority of open management networks (seeFig. 8), such as EnergyHub andGreenwaveReality, havea physical display and sensors, and some offer control toboth users and third parties, thus enabling utilities tomanage large connected loads such as pool pumps.

Table 4 summarises the characteristics of each of thenine types of feedback with respect to hardware, abilityto communicate data, options to control appliancesremotely, display options, data collection and the com-munications protocol used by the system.

Conclusion

Although feedback has been widely studied and is amuch-anticipated part of our national and global transi-tion to the Smart Grid, there have been few attempts toclearly distinguish among the hundreds of feedbacktechnologies and their unique characteristics. This studyis a vital first step towards an energy feedback “market”,in which consumers can feel confident to select andpurchase products and platforms that help them under-stand their energy usage in the home.

This work provides a novel contribution to the energyfeedback literature by linking together the theoreticalunderpinnings of feedback technologies with actual com-mercial or pre-commercial ventures. However, it must benoted that the energy feedback market is a fast pacedsector with many companies entering and leaving themarket each year. This taxonomy was derived empirical-ly from data collected in 2010–2011; since then, a num-ber of players have left this space (e.g. Google) and othershave entered (e.g. Chai Energy, Bidgely). This is not aproblem in itself, as the main aim of this work wasdevelopment of a taxonomy of energy feedback technol-ogies and not to compile a current and complete list ofthem. However, a potential knock on effect is that thesechanges could result in the creation or disappearance ofwhole categories of feedback technologies. The feedbacktechnology market does not present a unique situation inthis respect and problems can be managed here byreviewing key characteristics and resulting categorisa-tions as technologies develop over time, as has been donefor classification structures of other technologies. Forexample, as camera technology has advanced in recentyears additional categories, such as Megazoom” and“Interchangeable-Lens Camera (ILC)”, have been added

Table 4 Characteristics of each feedback type

Feedback type Hardware Communications Control Display Collection Protocol

Information platform No NA No − − −

Management platform No NA Yes − − −

Appliance monitor Yes No No Embedded Appliance −

Load monitor Yes No No Embedded Sensor −

Grid display Yes Yes No Autonomous Grid −

Sensor display Yes Yes No Autonomous Sensor −

Networked sensor Yes Yes No None Sensor −

Closed management network Yes Yes Yes Various Various No

Open management network Yes Yes Yes Various Various Yes

Energy Efficiency


to existing categorisations to account for the differencesin key camera characteristics described by the new tech-nologies. This means that the taxonomy must be viewedas dynamic rather than static, requiring regular reviewsand revisions to ensure that the categories describingcommercial and pre-commercial feedback technologiesremain fully representative of the marketplace. Althoughthis presents a potential time sensitivity to the uselessnessof the current categorisations, without them it would bemuch harder for consumers and reviewers to comparedifferent models and determine which is most appropriatefor a given situation. Furthermore, the identification ofkey characteristics development of the taxonomy struc-ture provided in this work can help guide the creation ofadditional categories as needed.

This work is distinct from much of the existingliterature; rather than try to evaluate the effectivenessof a particular type or characteristic of feedback, weinstead aim to develop a categorisation to assist prac-titioners and researchers organise future work on ener-gy feedback. This research does not provide a compar-ison between technologies, which we envisage as anadditional research project. Rather, we have developeda categorisation that can serve as the basis for publiclyavailable product information on feedback devices andsystems, much like that which is available for otherconsumer electronics (e.g. televisions, cameras, etc.).This work did exclude home energy management tech-nologies that gave users control of appliances withoutfeedback. Whilst this study was aimed specifically atfeedback technologies due to the potential savings thatcan be induced from increased awareness and knowl-edge about personal energy use, an extension to othertypes of home energy management system offeringautomation and control would be a useful next step.

A consolidation of all available data on energy feed-back devices will be useful, not only to the averageconsumer but also to researchers interested in energyfeedback technology. There is currently a great deal ofredundancy within the research community, with multi-ple teams across the country (and world) collecting thisinformation on their own in order to test and comparedevices or systems. Our taxonomy report and subsequentdatabase allows researchers to focus on designing andimplementing studies, rather than attempting to scour theweb looking for basic information about these devices.

Feedback has become ubiquitous in many otherareas of our lives (e.g. food, transportation), and thereis a clear indication from policy leaders and private

industry that energy information should and will bemade readily available to consumers in the comingdecade. A better understanding of the diversity andsimilarity of devices is vital to this growing field.

Appendix A

Name Developer or retailer Type

4Home Motorola Open managementnetwork

Aclara’sENERGYprismCustomer CareSolutions

Aclara Informationplatform

The Agentis Platform Agentis Informationplatform

Acuview EnergyRecording &ReportingSoftware

Accuenergy Informationplatform

Agilewaves BuildingOptimizationSystem

Agilewaves Open managementnetwork

Akuacom DemandResponseAutomationServer

Akuacom Informationplatform

AlertMeSmartEnergy

AlertMe Closedmanagementnetwork

AmWatt ApplianceLoad Tester

Reliance Controls Load monitor

AzTech AZ Tech Associates,Inc.

Grid display

Battic Door HomeEnergy Monitor

Battic Door Sensor display

Belkin ConserveInsight Monitor

Belkin Load monitor

Black & DeckerPower Monitor

Black & Decker Sensor display

Blue LinePowerCostMonitor andEnergy Meter II

Blue Line Innovations Sensor display

Brand Electronics20-CTRWholeHouse

Brand Electronics Sensor display

Brand ElectronicsDigital PowerMeter 4-1850

Brand Electronics Load monitor

Brand ElectronicsDigital PowerMeter 20-1850


Energy Efficiency



Brand ElectronicsDigital PowerMeter 20-1850CI


Brand ElectronicsDigital PowerMeter 21-1850CI


Brand ElectronicsONE meter

Brand Electronics Sensor display

Brultech ECM-1220 Brultech Sensor display

Brultech ECM-1240 Brultech Networked sensor

Brultech GreenEyeMonitor

Brultech Networked Sensor

Cent-a-Meter Island Power Pty Ltd Sensor display

CentralisedElectricity EnergyManagementSystem

Shenzhen SailwiderElectronics

Closedmanagementnetwork

Cisco BusinessEnergyManagementServices

Cisco Informationplatform

Control4 EnergyManagementSystem 100

Control4 Open managementnetwork

Current Cost EnviR Current Cost Sensor display

Current Cost TheClassic

Current Cost Sensor display

Current Cost TREC Current Cost Sensor display

Customer InterfaceDisplay

Dent Instrument Grid display

CustomerIQ EnergyPortal

Silver SpringNetworks

Informationplatform

Digi X-Grid Solutions Digi Open managementnetwork

DreamWatts Makad Energy Open managementnetwork

E-Mon EnergySoftware

E-Mon Energy Informationplatform

E-Watch Secure Together Informationplatform

Ease II Manager Secure Together InformationPlatform

Eco-Eye Elite Modern Moulds andTools Ltd’s

Sensor display

Eco-eye Elite 100 Modern Moulds andTools Ltd’s

Sensor display

Eco-eye Elite 200 Modern Moulds andTools Ltd’s

Sensor display

Eco-Eye Elite Mini Modern Moulds andTools Ltd’s

Sensor display

Eco-eye Elite Mini 2 Modern Moulds andTools Ltd’s

Sensor display

Eco-Eye Plug-In Modern Moulds andTools Ltd’s

Load monitor

Eco-Eye Smart Sensor display


Modern Moulds andTools Ltd’s

Eco-Eye Smart PC Modern Moulds andTools Ltd’s

Sensor display

Eco-Eye Smart PV Modern Moulds andTools Ltd’s

Sensor display

EcoDog’s FIDOHome EnergyMonitoringSystem

EcoDog Networked sensor

EcoManager EDF Energy Closedmanagementnetwork

ecoMeter Ampy EmailMetering

Grid display

EcoView Commercial Advanced Telemetry Closedmanagementnetwork

EcoView Residential Advanced Telemetry Closedmanagementnetwork

Efergy E2 Efergy TechnologiesLtd

Sensor display

Efergy Elite WirelessMonitor

Efergy TechnologiesLtd

Sensor display

Efergy EnergyMonitoringSocket

Efergy TechnologiesLtd

Load monitor

Efficiency 2.0’sPEER

Efficiency 2.0 Informationplatform

eGauge eGauge Systems Networked sensor

ELOGsoftware Dent Instruments Informationplatform

EM 2500 Energy MointoringTechnologies

Load monitor

eMonitor Powerhouse Dynamics Closedmanagementnetwork

EMS-2020 Energy Cite Grid display

EnergateHomeEnergyManagament Suite

Energate Open managementnetwork

Energy Engage eMeter Informationplatform

EnergyFlowMonitor Noveda Technologies Networked sensor

EnergyHub HomeBase

EnergyHub Open managementnetwork

Energy Joule Ambient Grid display

Energy ManagerEXT Software

Electro Industries/GaugeTech

Informationplatform

Energy Master eQ-3 Load monitor

EnergySmart Monitor British Gas Sensor display

EnergyView Online Schneider Electric Informationplatform

Energy Efficiency



eSight eSight Energy Informationplatform

Ewgeco B100 Tayeco Ltd. Sensor display



Ewgeco H300 EEE Tayeco Ltd. Sensor display

Ewgeco H300 ERG Tayeco Ltd. Sensor display

Ewgeco H300 EWG Tayeco Ltd. Sensor display

Ewatch 100—Secure Secure Together Informationplatform

Flukso Flukso Networked sensor

Home EnergyController (HEC)

Secure Together Open managementnetwork

Freedom Secure Together Grid display

GEO Chorus Green Energy Options Open managementnetwork

GEO Duet Green Energy Options Grid display

GEO Ensemble Green Energy Options Closedmanagementnetwork

GEO Minim Green Energy Options Sensor display

GEO My Energy Green Energy Options Informationplatform

GEO NpowerMonitor

Green Energy Options Sensor display

GEO Prelude Green Energy Options Sensor display

GEO Quartet Green Energy Options Sensor display

GEO Solo Green Energy Options Sensor display

GEO Solo II Green Energy Options Grid display

GEO Trio & Trio+ Green Energy Options Open managementnetwork

Google PowerMeter Google Informationplatform

Greendash Hub FutureDash Corp. Managementplatform

GreenWave RealityEnergyManagementPlatform

Greenwave Realitty Open managementnetwork

Greenwire EnergyMonitor

Greenwire LLC Grid display

GridPoint EnergyManager

Gridpoint Managementplatform

Home EnergyManagementSolution (Cisco)

Cisco Open managementnetwork

iControlOpenHome—Utility

iControl Networks Open managementnetwork

In2MyHome In2Networks Informationplatform


Insteon Energy display Insteon Sensor display

Insteon HouseLinc Insteon Closedmanagementnetwork

Insteon SmartLinc Insteon Closedmanagementnetwork

Intamac Intamac Open managementnetwork

Intel Home EnergyDashboard

Intel Open managementnetwork

Kill-A-Watt P3 International Load monitor

Kill AWatt EZ P3 International Load monitor

Kill-A-Watt GraphicTimer & PlugPower Meter

P3 International Load monitor

Kill AWatt PowerStrip

P3 International Load monitor

Kill-a-Watt Wireless P3 International Sensor display

LS ResearchRateSaver display

LS Research Grid display

Lucid Design Group:Dashboard

Lucid Design Group Informationplatform

Lucid BuildingDashboard—B

Lucid Design Group Networked sensor

Lucid BuildingDashboard—C

Lucid Design Group Networked sensor

Mi Casa VerdeSmartSwitch &Vera

MiCasaVerde Closedmanagementnetwork

Microsoft Hohm Microsoft Informationplatform

Navetas EnergyMonitor

Navetas EnergyManagement

Sensor display

Navetas Smart Hub Navetas EnergyManagement

Sensor display

Needy Needs IncWireless EnergyMonitor

Needy Needs Sensor display

Nokia Home ControlCenter

Nokia Managementplatform

Nucleus General Electric Open managementnetwork

Onzo Onzo Sensor display

OpenFrame 7E(OpenPeak)

OpenPeak Open managementnetwork

OPOWER EnergyReports

Opower Informationplatform

OPOWERWebportal

Opower Informationplatform

People Power 1.0 People Power Informationplatform

People Power

Energy Efficiency



People PowerEnergy ServicesPlatform+SurfModule


PICOwatt/TenrehtePlug

Tenrehte Closedmanagementnetwork

Plugwise Plugwise Closedmanagementnetwork

Power Aware Cord Static! Load monitor

Power Cost DisplayMonitor

Energy ControlSystems

Sensor display

PowerLogic EPOEnergy ProfilerOnline

Schneider Electronics Informationplatform

PowerLogic IONEEM Software


PowerLogic IONEnterpriseSoftware


PowerLogicPowerViewSoftware


PowerLogic SCADASoftware

Schneider Electronics Managementplatform

PowerLogic SystemManager Software


PowerLogic TenantMetering Software


PowerPal Meter withCustomerInterface Display

Dent Instruments Sensor display

Power Player Quby Grid display

PowerStat Brunswick ElectricMembership

Grid display

PowerWatch-DR PowerWatch Inc. Open managementnetwork

Pulse EnergyManager

Pulse Energy Informationplatform

Pulse Check Pulse Energy Informationplatform

RCS Whole homemonitor & control

RCS Technology Closedmanagementnetwork

San Vision MobileEnergy Assistant(MEA)

San Vision EnergyTechnology Inc.

Open managementnetwork

San Vision PowerDashboard

San Vision EnergyTechnology Inc.

Sensor display

SaveOMeter Eco1 Sensor display

Scroller—Secure Secure Together Informationplatform

Seasonic Electronics Load monitor


Seasonic PowerAngelMonitor

Senquentric System Sequentric Open managementnetwork

Shaspa Smart Home Shaspa Open managementnetwork

Silk Powertech IST Otokon Informationplatform

Silver SpringNetwork’s SmartEnergy Dashboard

Silver Springs Managementplatform

SMARTware by Dent Dent Instruments Informationplatform

SolarCity’sPowerGuide

SolarCity Informationplatform

SRP M-Power Meter Salt River Project Grid display

Stanley 77–028Energy MeterEM100

Stanley Load monitor

SunPower Monitor SunPower Sensor display

Talking Plugs Zerofootprint Closedmanagementnetwork

Techtoniq EnergyStation

Techtoniq Informationplatform

TED 1001 Energy Inc. Sensor display

TED 1002 Energy Inc. Sensor display

TED 5000-C Energy Inc. Sensor display

TED 5000-G Energy Inc. Networked sensor







Tendril Tendril Open managementnetwork

The Energy Stick Quby Grid display

The Energy Valet Trilliant Managementplatform

The PowerTab Energy AwareTechnology Inc.

Grid display

The Owl ElectricityMonitors

2 Save Energy Limited Sensor display

UPM Dual RateEnergy Meter-EM130

UPM Marketing Load monitor

UPM EM100 EnergyMeter



Energy Efficiency



UPM Plug-in EnergyMeter and Electrici-ty Cost Calculator

UtiliFlex Juice UtiliFlex Informationplatform

UtiliFlex Joule UtiliFlex Networked sensor

U-Vue U-VUE Sensor display

WANF ElectricityEnergy WattUsage Meter

WANF Load monitor

Watts Up Standard Electronic EducationalDevices

Load monitor

Watts Up .Net Electronic EducationalDevices


Watts Up Pro ES Electronic EducationalDevices

Load monitor

Watts Up Pro Electronic EducationalDevices

Load monitor

Watts Up SmartCircuit 21

Electronic EducationalDevices

Networked sensor

Wattson 01 DIY Kyoto Sensor display

Wattvision EnergySensor

Wattvision Networked sensor

Wilting Flower Carl Smith Sensor display

Wireless Uni-directional ElectricalEnergy SavingMonitor


Sensor display

Wireless Bi-directionalElectricity EnergySaving Monitor &Control System



thinkeco Modlet thinkeco Closedmanagementnetwork

Wattsclever Compact wattsclever Sensor display

Wattsclever EnergyMonitor

wattsclever Sensor Display

Wattsclever EnergyMonitor for SmartMeter

wattsclever Sensor display

Wattsclever EnergyWatch Monitor

wattsclever Load monitor

References

Abrahamse, W., Steg, L., Vlek, C., & Rothengatter, T. (2005). Areview of intervention studies aimed at household energyconservation. Journal of Environmental Psychology, 25,273–291.

Anderson, W. & White, V. (2009). Exploring consumer prefer-ences for home energy display functionality. Report to the

Energy Saving Trust. Centre for Sustainable Energy.Åström, K. J., & Murray, R. M. (2009). Feedback systems: An

introduction for scientists and engineers. Princeton: PrincetonUniversity Press.

Badami, V. V., & Chbat, N. W. (1998). Home appliances getsmart. Spectrum, IEEE, 35(8), 36–43.

Bandura, A. (1969). Principles of behaviour modification. NewYork: Hold, Rinehart & Winston.

Campion, M. A., & Lord, R. G. (1982). A control systemsconceptualization of the goal-setting and changing process.Organizational Behavior and Human Performance, 30,265–287.

Carver, C. S., & Scheier, M. F. (1981). Attention and self-

regulation: A control theory approach to human behavior.New York: Springer.

Chopra, A. (2011). Modeling a green energy challenge after a

blue button. Washington: White House Office of Scienceand Technology Policy. http://www.whitehouse.gov/blog/2011/09/15/modeling-green-energy-challenge-after-blue-button. Accessed 20 April 2012.

Cohen, J. (1960). A coefficient of agreement for nominal scales.Educational and Psychological Measurement, 20, 37–46.

Cohen, H., & Lefebvre, C. (2005). Handbook of categorizationin cognitive science. New York: Elsevier.

Corbin, J., & Strauss, A. (2007). Basics of qualitative research:Techniques and procedures for developing grounded theory

(3rd ed.). Thousand Oaks: Sage.Creswell, J. W. (2009). Research design: Qualitative, quantita-

tive, and mixed methods approaches (3rd ed.). ThousandOaks: Sage.

Danahy, J. (2009). The coming smart grid data surge. Bloggingthe Grid. Retrieved from http://www.smartgridnews.com/artman/publish/News_Blogs_News/The-Coming-Smart-Grid-Data-Surge-1247.html.

Darby, S. (2001). Making it obvious: Designing feedback intoenergy consumption. In Proceedings of the 2nd internation-al conference on energy efficiency in household appliances

and lighting. Italian Association of Energy Economists/EC-SAVE programme.

Darby, S. (2006). The effectiveness of feedback on energy con-sumption.AReview for DEFRA of the Literature onMetering,Billing and Direct Displays. Oxford: Environmental ChangeInstitute, University of Oxford.

Donahue, G. (2007). Network warrior. O’Reilly Media,Incorporated

Donnelly, K. (2010). The technological and human dimensionsof residential feedback: An introduction to the broad rangeof today’s feedback strategies. In People-centered initia-

tives for increasing energy savings. Washington: ACEEE.Duffy, P. R. (1984). Cybernetics. Journal of BusinessCommunication,

21(1), 33–41.Ehrhardt-Martinez, K., Donnelly, K. A., & Laitner, J. A. (2010).

Advanced metering initiatives and residential feedback pro-

grams: A meta-review for household electricity-saving op-

portunities. Washington: American Council for an Energy-Efficient Economy.

EPRI. (2009). Residential electricity use feedback: A research

synthesis and economic framework. Palo Alto: ElectricPower Research Institute Report no. 1016844.

Energy Efficiency


http://www.whitehouse.gov/blog/2011/09/15/modeling-green-energy-challenge-after-blue-button



http://www.smartgridnews.com/artman/publish/News_Blogs_News/The-Coming-Smart-Grid-Data-Surge-1247.html



Faruqui, A., Harris, D., & Hledik, R. (2010). Unlocking the €53billion savings from smart meters in the EU: How increasingthe adoption of dynamic tariffs could make or break the EU’ssmart grid investment. Energy Policy, 38(10), 6222–6231.

Federal Energy Regulatory Commission. (2008). Assessment ofdemand response & advanced metering. Staff Report.

Fischer, C. (2008). Feedback on household electricity consump-tion: a tool for saving energy? Energy Efficiency, 1, 79–104.

Fitzpatrick, G., & Smith, G. (2009). Technology-enabled feed-back on domestic energy consumption: Articulating a set ofdesign concerns. Pervasive Computing, IEEE, 8(1), 37–44.

Froehlich, J. (2009). Promoting energy efficient behaviors in thehome through feedback: The role of human-computer in-teraction. HCIC ‘09 Workshop.

Froehlich, J., Findlater, L., Landay, J. (2010). The design of eco-feedback technology. In Proceedings of the 28th interna-

tional conference on Human factors in computing systems.Goetz, T. (2011). Harnessing the power of feedback loops. Wired

Magazine. http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1. Accessed 20 April 2012.

Goyal, S. C. & Bakshi, U. A. (2008). Principles of control

systems, Chapter 1. Pune: Technical Publications.Herter, K. (2010). Residential information and controls pilot:

Technology review memo – appendix A. El Dorado Hills,CA: Herter Energy Research Slutions.

Herter, K., & Wayland, S. (2009). Behavioural experimentationwith residential energy feedback through simulation gam-

ing. Sacramento: California Energy Commission.Hochwallner, T., & Lang, L. (2009). Approaches for monitoring

and reduction of energy consumption in the home. InSeminar-Thesis: Lecture series on sustainable developmentand information and communication technology.

Hofstede, G. (1978). The poverty of management control philos-ophy. The Academy of Management Review, 3(3), 450–461.

Institute for Electric Efficiency. (2011).Utility-scale smart meterdeployments, plans & proposals. Washington: Institute forElectric Efficiency.

Karlin, B., & Zinger, J. (2013). Residential energy feedback: ameta-analysis and methodological review. Under review.

Klein, H. J. (1989). An integrated control theory model of workmotivation. The Academy of Management Review, 14(2),150–172.

Kluger, A. N., & DeNisi, A. (1996). The effects of feedbackinterventions on performance: a historical review, a meta-analysis, and a preliminary feedback intervention theory.Psychological Bulletin, 119(2), 254–284.

LaMarche, J., Cheney, K., Christian, S., & Roth, K. (2011).Home energy management products & trends. Cambridge:Fraunhofer Center for Sustainable Energy Systems.

Latham, G. P., & Locke, E. A. (1991). Self-regulation throughgoal setting. Organizational Behavior and Human

Decision Process, 50, 212–247.Lawrence, J. W., Carver, C. S., & Scheier, M. F. (2002). Velocity

toward goal attainment in immediate experience as a deter-minant of affect. Journal of Applied Social Psychology,

32(4), 788–802.Locke, E., & Latham, G. (1990). A theory of goal setting & task

performance. Englewood Cliffs: Prentice Hall.Marradi, A. (1990). Classification, typology, taxonomy. Quality

and Quantity, 24(2), 129–157.McKerracher, C., & Torriti, J. (2013). Energy consumption

feedback in perspective: Integrating Australian data tometa-analyses on in home displays. Energy Efficiency, 6(2),387–405.

Neuendorf, K. A. (2001). The content analysis guidebook.Thousand Oaks: Sage.

Podsakoff, P. M., & Farh, J. L. (1989). Effects of feedback signand credibility on goal setting and task performance.Organizational Behavior and Human Decision Processes,

44, 45–67.Sánchez, M. (2012). Status on the roll-out of Smart Metering in

the EU. Presented at the Eurelectric and ESMIG event,Brussels, Belgium.

Schultz, P. W. (2010). Making energy conversation the norm. InEhrhardt-Martinez, K. & Laitner, J. A. (Eds.), People-cen-tered initiatves for increasing energy savings (251–262).Washington, DC: American Council for an Energy EfficientEconomy.

Skinner, B. F. (1938). The behavior of organisms: An experi-

mental analysis. New York: Appleton-Century.Stein, L. F. (2004). Final report. California information display

pilot technology assessment. Boulder: Primen.Stein, L. F., & Enbar, N. (2006). EPRI solutions custom report

direct energy feedback technology assessment.Stemler, S. (2001). An overview of content analysis. Practical

Assessment, Research & Evaluation, 7(17).Vallacher, R., & Wegner, D. (1987). What do people think they

are doing? The presentation of self through action identifi-cation. Pyschological Review, 94(1), 3–15.

Wiener, N. (1948).Cybernetics: Or the control and communication inthe animal and the machine (2nd ed.). Cambridge: MIT Press.

Wood, G., & Newborough, M. (2007a). Energy-use informationtransfer for intelligent homes: Enabling energy conservationwith central and local displays. Energy and Buildings, 39(4),495–503.

Wood, G., & Newborough, M. (2007b). Influencing user behaviourwith energy information display systems for intelligent homes.International Journal of Energy Research, 31, 56–78.

Energy Efficiency


http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1

http://www.wired.com/magazine/2011/06/ff_feedbackloop/all/1

2013 karlin ford squiers taxonomy-libre

Documents