FinalreportontheAmphiro-PWN-study
EffectsofReal-TimeFeedbackonHotWaterUse
Date:
March31,2016
Status:
DraftV0.99
Authors:
ThorstenStaake,UniversityofBambergandETHZurich
VerenaTiefenbeck,ETHZurichandBonnUniversity
SamuelSchöb,UniversityofBamberg
AnnaKupfer,UniversityofBamberg
Contactinformation
Forresearch-relatedquestions:[email protected]
ForPWN-relatedquestions:[email protected]
Internaldraft,notforpublication
FinalreportontheAmphiro-PWN-study
2016-03-31
Tableofcontents
Abstract................................................................................................................................................3
1. Motivationandstudyobjectives...................................................................................................4
2. Introductionofthefeedbackdevice.............................................................................................7
3. Studydesignandexecution..........................................................................................................8
4. Results.........................................................................................................................................11
5. Implicationsandconclusion........................................................................................................15
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page2of16
FinalreportontheAmphiro-PWN-study
2016-03-31
Abstract
Hot water consumption is the second largest position in a household’s energy budget, only outrankedbyspaceheating.Theaveragehouseholdusesmoreenergyforwaterheatingalonethanfor
lighting,refrigerators,andconsumerelectronicscombined.Thenexusbetweenhotwaterandenergyuse,however,isbarelyknownamongconsumers,andconsequentlyhotwaterisbarelyconsideredasastrategytoconserveenergyevenamongveryenvironmentallyconcernedcitizens.
Inafieldstudyamong637Dutchhouseholds,weinvestigatedhowreal-timeconsumptionfeedback
–theprovisionofinformationdirectlyintheshowerontheamountofwaterandenergyused–influenceshotwateruse.Participantsobtainedasmartshowerdisplay,whichtheywereabletoinstall by themselves between shower hose and showerhead. Data from 73’977 showers were recordedinamonitoringphaseoverthreemonthsstartingendofAugust2015.
In a baseline phase (feedback devices installed, consumption feedback deactivated), the average
consumptionwas3.2kWhand54literspershower.Withfeedbackinformation(waterandenergy
useplusenergyefficiencyrating),participantssavedonaveragebetween19%and21%oftheirenergyconsumptionintheshower.Absolutesavingspershoweramountto0.6kWh.Thesavingeffectsarestatisticallyhighlysignificantandstableovertheentireinterventionphase.
Projectedtooneyear,athree-personhouseholdintheNetherlands(with0.85showersperperson
andday)savesonaverage561kWhofheatenergyand8.7m3litersofwater.Forthesamehousehold size, monetary savings amount to86 EURperyear,leadingtoan amortization of the device
within less than one year. From an investment perspective, abatement costs are below 0.05 EUR
perkWhsavedforanaverage2.3-personhousehold(forcomparison:0.10EURperkWhforlarge
PVinstallations).
ThestudyconfirmsthepreviousresultsoffieldtrialsconductedinSwitzerland(togetherwithMobiliarInsurance)withsavingsofonaverage22-24%and100EURrespectively.Inhotel,evenwhen
theuserdoesnotpayforwaterandenergy,savingsofonaverage20%havebeenobservedintrials
conductedjointlywithMobiliar.Heavyusersandyoungindividualsevensavemoreonaverage.
Overall,thestudyconfirmsthatpersonalreal-timefeedbackonaspecific,energy-intensivebehavioronwhichthefeedbackreceiverhasastronginfluenceproducesverylargesavingeffects.Moreover,fortheapplicationintheshower,themeasurehasanexcellentcost-benefitratioandisapplicabletothevastmajorityofhouseholds(homeownersandtenants).
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page3of16
FinalreportontheAmphiro-PWN-study
2016-03-31
1. Motivationandstudyobjectives
Thewater-energynexus
In the Netherlands, water heating uses 2,460 kWh of energy per year
and household (see Figure 1), exceeding the total amount of energy
usedforcooking,lighting,andICTcombined.Whilespaceheating,lighting, and electric appliance have received much attention in the past
(e.g., in form of subsidized roof refurbishments, the prohibition of incandescent light bulbs, efficiency ratings, etc.), water heating only recentlymovedintothefocusofattentionamongpolicymakersandfirms
Waterheatingisthe
secondlargespositiononahousehold’s
energybill.
asanimportantfieldforimprovement.
Water heating deserves special attention in environmental campaigns,
asresidentialwaterheatinginCentralEuropeprimarilyreliesonoil,gas,
and electricity. This makes (hot) water generation very carbon intense,
andmeasurestoreducethedemandparticularlyworthwhiletopursue.
Waterheatingisvery
carbonintense.
Thefigureforwater
heatingdoesnotincludehotwatergeneratedbydishwashersandwashingmachines(partofappliances)andwater
heatedonthestove
(partofcooking).
1
Figure1:Consumptionperhouseholdbyenduse Despiteitsimportance,thenexusbetweenhotwaterandenergyuseis
barely known among consumers. While dedicated accounts for space
heating,smartelectricitymeterswithin-homedisplays,enhancedelectricitybills,etc.increasinglyinformconsumersandhelpthembuildupa
generalunderstandingonthosedomains,energyforpersonalhotwater
use is barely communicated to the end user. Consequently, hot water
conservation is not considered as an energy conservation target even
amongveryenvironmentallyconcernedcitizens,leavinglargesavingpo-
Citizensarebarely
awareofthelarge
amountofenergy
thatgoesintowater
heating.
1
Source:EuropeanEnvironmentAgency(2012)/SplitofICT&Lightinginferredfrombdew(2010).
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page4of16
FinalreportontheAmphiro-PWN-study
2016-03-31
tentials untapped. The understanding that drinking water is abundant
reducestheintentiontoconservehotwaterevenfurther.Nevertheless,
weexpectagrowingattentionpaidtohotwaterinthecomingyears:Its
shareamongtotalresidentialenergydemandwillgrow,asbetterbuildinginsulationandefficiencystandardsdrivedown energyconsumption
inotherdomains.Innewlybuilt,energyefficientbuildings,waterheatingevenexceedstheamountofenergyusedbyspaceheating.
Showerwaterconsumptionasanenergysavingtarget
Mostofthehotwater(about70%)isconsumedintheshower.Anaverageshowerrequires3.2kWh(includinglosses)inonly8minutes.There
isnootheractivityinanapartmentwithacomparablepowerlevel(i.e.,
thatusesmoreenergypertime).Atthesametime,showeringisanexcellent target for conservation measures: First, the energy intensity allows for large absolute savings, and the short time interval of each
shower only requires a short period of attention of a consumer to engageinenergyconservationbehavior.Second,showeringisadailyrou-
Showeringisextremelyenergyintense:1secondin
theshowerrequires
asmuchenergyas
workingonalaptop
computerfor2
hours.
tine of almost all citizens, so campaigns have the potential to address
virtually all households. Third, only a very small fraction of the citizens
knows how energy intense the daily shower is. Thus, information campaigns may leverage a steep learning curve. Forth, consumers have a
huge influence on their shower water use. Cutting one’s shower duration by only one minute leads to savings in the order of 13%. Fifths,
shower consumption is something that can be easily attributed to one
Showeringisanexcellenttargetforbehavioralefficiency
campaigns.
individual person (unlike electricity that is partly determined by other
household members and by devices that have a base load), making it
possibletoprovidepersons-specific,verytargetedandthuspowerfulinterventions.
Supportingconsumerswithfeedbackinterventions
Feedback interventions inform users about their behavior and/or its
outcomes. The idea behind the concept is that the feedback receiver
getsinformationontheoutcomeofabehaviorsheperformedinaway
thatiseasytoremember(oftennumericalorsymbolic).Astepcounter
is a prominent example for physical activity during the day. The feedbackhelpsthereceivertoevaluateherperformance,compareittoprevious days and to the performance of others, to set personal goals, to
sharetheinformationamongpeers,etc.Ifprovidedinaneasytounderstand, motivating way, feedback interventions can have a large impact
on behavior. Depending on the domain, feedback can support physical
exercise, diet plans, the attempt to quit smoking, vocabulary learning,
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Feedbackinterventionsinformthereceiverabouttheoutcomeofabehavior.
Page5of16
FinalreportontheAmphiro-PWN-study
2016-03-31
teethbrushing–andenergyconservation.Intheenergydomain,feedbackinterventionshavebeenshowntoleadtoeffectsthatareequivalentineffectsizetolargepriceincreases.2Theconceptoffeedbackisoftenappliedinordertocurbelectricityconsumption(e.g.,intheformof
in-homedisplayswithdataprovidedbysmartelectricitymetersorinthe
form of paper-based home energy reports) or as means to reduce fuel
consumption(intheformoffuelgaugesonacar’sdashboard).
Feedbackcanbecategorizeddependingonitstimeliness(e.g.,monthly
vs.dailyvs.directlyafterabehaviorhasbeenperformedvs.alreadyduring a behavior). It can also be categorized according to its specificity
(e.g.,providedforasinglebehaviorvs.foraggregatedbehaviors/providedperpersonvs.perhousehold).Ingeneral,timelyfeedback(ideally
givenduringthebehavior)andperson-specificfeedbackismorelikelyto
yieldlargechangesinbehavior.
Themajorinfluencecitizenscanhaveontheirshowerconsumption,the
limitedawarenessamongconsumers(i.e.,thesteeplearningcurve),and
thepossibilitytoprovideperson-specificfeedbackduringthebehavioris
performed makes showering an excellent candidate for feedback interventions.
Real-time,person-
specificfeedbackis
expectedtobeespeciallypowerful.
Showeringisanideal
targetforfeedback
interventions
Studyobjectives
Theresearchgoalofthestudywastoquantifyandtobetterunderstand
theeffectofreal-timefeedbackonshowerbehavior.Wewantedtofind
out (1) how feedback changes the amount of hot water and thus the
amountofenergyconsumed,(2)iftheeffectsarestableovertime,and
(3)ifspecificsubgroupsofthestudyparticipantssavemorethanothers.
AdditionalgoalsfromPWNhavebeentolearn(4)howreliablethefeedbackdevicesworkedinthefield,(5)howtheparticipatinghouseholdrespondedtothedevices,and(6)iftheratiobetweencostandsavingper
deviceiscompetitive.
Themainstudyobjectivewastoinvestigatetheeffectof
real-timefeedback
onhotwateruseand
thepracticabilityof
thefeedbackdevice.
Theobjectiveshavebeenaddressedinalarge-scalefieldstudyinvolving
637 Dutch households in 2015. The study was conduced by a research
teamlocatedattheUniversityofBamberg,ETHZurich,andtheUniversity of Bonn. PWN at Velserbroek financed the study and supported its
implantation.
2
Source: Allcott, Hunt (2011). Social norms and energy conservation. Journal of Public Economics 95(9),
1082-1095.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page6of16
FinalreportontheAmphiro-PWN-study
2016-03-31
2. Introductionofthefeedbackdevice
Thecentralelementofthestudywasasmallmeasurementanddisplay
devices(“amphirob1”)thatprovidesreal-timefeedbackonenergyand
waterconsumptionintheshower(Figure2).Forthestudy,the637participatinghouseholdsinstalledstudyversionofthedevicebetweentheir
showerhead and shower hose. The devices turn on automatically as
soonaswaterflowsandprovideinformationonwaterconsumption(in
liters)andenergyconsumption(in[k]Wh)foreachindividualshowertogetherwithanefficiencyratingfromAtoG(Aindicatingaveryenergy
efficient shower). The information was provided on a per shower basis
(startingatzeroforeachshowerextraction),andextractionsthatshortly
followedeachotherweregroupedintoonelargerextraction(toprovide
one number per shower even if the user makes short breaks, e.g., to
shampooherhair).
Thesmartshower
meter“amphirob1”
providesreal-time
feedbackonwater
andenergyusedirectlyintheshower.
Figure2:Schematicrepresentationofthefeedbackdeviceamphirob1
Ingreaterdetail,displayarea1toggleswatertemperatureandtheenergyefficiencyrating.Whennowaterflows,thedisplayremainsonfor
abouttwominutes,waitingwhethertheshowerisbeingcontinuedand
allowing the user to still see the final shower results. Display area 2
showswaterconsumptionoftheongoingshowerextractionduringthe
shower,andtogglesbetweenwaterconsumptionandenergyconsumptionwhennowaterflows.Displayarea3providesanemotionalrepresentationoftheefficiencyrating(apolarbearstandingonameltingice
floe;thesizeoftheicefloeiscoupledwiththeenergyefficiencyrating).
Area3doesnotprovideadditionalinformation,buthassomeemotional
valueandappearstomakeusertalkmorefrequentlyaboutthedevice.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page7of16
FinalreportontheAmphiro-PWN-study
2016-03-31
Asthedeviceautomaticallyturnsonwheneverwaterflows,itdoesnot
require user action to start the measurement process. Nor does it require batteries, as it generates its energy from the water flow; this increases the environmental appeal of the device and makes battery
changesabundant.Thepressuredropissmall–thedevicedoesnotreducethewaterflowinaverynoticeableway.Unlikemechanicalflowrestrictors that face much criticism for restricting consumer choice and
Thedevicedoesnot
requireabatteryasit
generatesitsenergy
fromthewaterflow.
shower comfort in a paternalizing way, the feedback device lets users
freelychoosetheflowrateanddurationoftheirshowers.
Besides displaying the information outlined above, the amphiro b1
stores(foreachshower)theduration,volume,numberofinterruptions,
andaveragetemperatureperextraction.Overall249suchdatasetsfitin
the memory. Base temperature (for the energy calculation) was set to
12°C.Deviceaccuracyis+-6%betweendevicesataflowrateof12l/min,
with a much smaller deviation for repeated measures of one device.
Flowratescanbebetween2l/minand22l/min.
The devices were built in Austria according to common environmental
standardsanduseonlydrinkingwatercompliantmaterialsforallwaterconductingparts.
Theinstallationdoesnotrequireanytools(seeFigure3).amphirob1is
compatibletoallcustomary1/2"handshowersystems.
Figure3:Installationofamphirob1
3. Studydesignandexecution
Experimentaldesign
Thestudywasorganizedasafieldexperimentinordertoexaminethe
effectofthefeedbackinterventionintherealworld(i.e.,notinanartificialsettinginalaboratory).Participantswererandomlyassignedtotwo
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page8of16
FinalreportontheAmphiro-PWN-study
2016-03-31
differentgroups3,theso-calledtreatmentandthecontrolgroup,which
received group-specific devices. The devices handed out to the control
group displayed only information on water temperature (i.e., no feedbackonwaterorenergyuse).Thedevicesgiventothetreatmentgroup
also displayed only water temperature during the first N*10 showers
(referred to as baseline phase; N describes the number of household
members using the shower), but thereafter automatically switched to
Thefieldexperiment
wasorganizedasa
randomizedcontrolledtrialpreceded
bybaselinemeasurements.
feedbackmode(theinterventionphase).Intheinterventionphase,the
devicesprovidedthefullsetofreal-timefeedbackonwaterandenergy
consumptionasdescribedinSection2.
This design is referred to as randomized controlled trial with baseline
phase. It allows us to investigate changes in consumption once the interventionofinterest(here:feedbackonconsumption)becomesactive
by observing the difference between baseline and intervention phase.
Moreover,byobservingthecontrolgroup,thestudydesignalsoallows
us to subtract non-intervention related influences (such as changes in
outdoor temperature or changes in the behavior that stem from the
feelingamongtheparticipantsofbeingmonitoredinastudy).Thestudy
designisillustratedinFigure4.Anonlinequestionnairewasconducted
bothatthebeginningandattheendofthestudy.
Figure4:Experimentaldesign
3
Randomizationwasperformedwhileconsideringthetargetgroupsize.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page9of16
FinalreportontheAmphiro-PWN-study
2016-03-31
Recruitmentofparticipatinghouseholdsanddatacollection
637householdsparticipatedthestudy.TheywererecruitedamongPWN
employees including subsidiaries, PWN customer panels, PWN volunteers,andagroupreferredtoasnudgepanelthatwasavailableforresearch purposes. Prior to the first questionnaire, a short online survey
637householdsparticipatedthestudy
wasconductedtoidentifyhouseholdsthatanticipatedtorelocateorto
beabsentforlongerperiodsduringthestudyorwhohadheadshowers
(where the feedback device could not have been installed). This was
done to avoid distributing devices to households that could not completethestudy.Participationwasvoluntarily(“opt-in”)andfreeofcost
totheparticipants.
Shower data was collected over a period of three months. Participants
whereaskedtoinstallasmartphoneappthatcollectedtheshowerdata
from the feedback devices and uploaded the retrieved data to a cloud
server for subsequent analyses. These steps required a compatible
smartphonewithBluetooth4.0connectivity.(iPhone>4Sandselected
Android phones). In case of problems during data upload, the research
teamsentreturnenvelopesandaskedtheparticipantstoreturnthedevicesviamail.PWNemployeesalsohadtheopportunitytodropthede-
Datawascollected
overaperiodof
threemonths.
viceoffatthePWNheadquarters.Theresearchteamthenreadoutthe
devices,setthemnormaloperationmode(sothatcontrolgroupparticipants received consumption feedback from then on) and retuned the
devicestothehouseholds.TheprocessstepsareshowninFigure5.
Figure5:Studytimeline
Returnrateanddatacollected
Outofthe637participatinghouseholds,503provideddataeitherbyusing the smartphone app or by shipping the device back for readout by
the research team. The return rate of 80% can be considered as very
good. In total, the datasets includes 73’977 shower extractions. From
Intotal,63’206data
pointswereavailable
forthesubsequent
analysis.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page10of16
FinalreportontheAmphiro-PWN-study
2016-03-31
these,63'206extractionscouldbeusedinthesubsequentanalysis.4This
makesthedatasetoneofthelargestonescoveringshowerbehaviorin
theworld.
4. Results
Graphicalrepresentation
The plot of the measured data nicely illustrates the effect of real-time
feedbackonenergyconsumption(Figure6).Wedescribeitstep-wiseto
easeitsinterpretation.
Thetwolinesshowsthemeanenergyconsumptionpershoweroverthe
course of the study of the two groups (blue = control group, red =
treatmentgroup).Duringthebaselinephase(nofeedback,from0%to
10% of study completion), the energy consumption of control and
treatment group participants is almost identical, showing us that the
randomdistributionoftheparticipantsintothetwogroupsworkedwell.
Thisisanimportantindicatorthattheparticipantsaresimilarregarding
important characteristics, and it increases the level of confidence that
the effects observed in the subsequent intervention phase can be attributed to the intervention and not to group-specific differences. The
firstdatapointisnoticeablylowerthantherest.Weassumethatisbecause many participants who installed the feedback device and tried it
outwithasmallerwaterextraction,withoutactuallytakingashower.
During the intervention phase (between 10% and 100% of the study
completion),controlgroupparticipants(bluedots)continuetoseeonly
water temperature. The consumption increases over time, as indicated
bytheupwardslopeoftheblueline.WeattributethistrendtotheHawthorneffect:atthebeginning,participants“feelobserved”andthustake
shortershowersthantheyusuallywould;overtime,theygetusedtothe
deviceandreturntotheirnormalshowerhabits.Thisisnotinterfering
withthestudyresults,astheeffectispresentforbothgroups.
Withtheonsetoftheinterventionphase(feedbackisshownforthefirst
time, study completion rate 11%), treatment group participants (red
dots)immediatelyreducetheirenergyconsumption.Thisdecreaseisat-
Withoutfeedback,
boththecontroland
thetreatmentgroup
consumethesame
amountofenergy.
Theupwardslopeis
attributedtothe
Hawthorneeffect.It
doesnotreducethe
absolutesizeofthe
savingeffects
Theimmediateeffect
offeedbackiseasily
visible.
4
Mostofthenon-usabledatapointscamefromhouseholdswerethenumberofshowersexceededtheinternaldevicememory,leadingtocorrupteddata.Forthesakeofbriefness,wedonotdiscussoutlierremoval
herebutincludeadiscussioninanextendedreporttoPWN.Thedetailedanalysisshowsthatoutlierremoval
doesnotaffectcontrolandtreatmentgroupdifferently.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page11of16
FinalreportontheAmphiro-PWN-study
2016-03-31
tributed to the feedback intervention. Savings are represented by the
differencebetweenthetwotrendlines.Thetrendlinesarealmostparallel: the treatment effect remains constant during the experiment. If
thereisachange,thenthegapseemstowiden;thatwouldmeanthat
thesavingsevenincreasethelongertheparticipantsreceivefeedback.
Thefollowingsubsectionquantifiestheeffectsize.Thegraphicalrepresentation,however,alreadyconveysthelargeeffectssizefromreal-time
feedbackonshowerbehavior.
Figure6:Feedbackeffectsonper-showerenergyuse
Calculatingtheeffectsizewithadifference-in-differencesmodel
In order to quantify the effect size, we calculated the changes in consumption with a difference-in-differences (DID) analysis. A DiD analysis
compares the mean energy use of the two groups (control and treatment)duringbaselineandduringtheinterventionphase.Thisrelatively
Thedifference-indifferencesmodel
showssavingsper
showerof0.64kWh.
simpleapproachhastheadvantageovermoresophisticatedregression
modelsthatitismorestraightforwardtounderstandandverify.
For a DID analysis, one derives the difference between control and
treatmentgroupduringthebaselinephaseandsubtractfromitthedifference between control and treatment group during the intervention
phase.Inourcase,thisrevealsaveragesavingspershowerof0.64kWh,
or20.8%.TheanalysisisillustratedinFigure6.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Consideringalsothe
resultsofaregressionanalyses,weregardpershowersavingsof0.6kWhas
veryreliableestimate.
Page12of16
FinalreportontheAmphiro-PWN-study
2016-03-31
AnalternativetoDIDanalysisistoestimateamorecomplexregression
model.Usingafixedeffectregressionmodel,wefoundthesavingstobe
0.55 kWh, or 19.6%. Given the inherent error margins of field studies,
thisvirtuallythesameresultasshownbytheDIDanalysis.
Figure6:Calculationoftheeffectsizewithadifference-in-differencesapproach(energyusepershower,nominimumthresholdfilter)
Theenergysavingsalmostcompletelyresultfromareductioninshower
duration. The treatment group only slightly reduced the flow rate, and
tooktheirshowersatalmostthesametemperature.Thisisnotverysurprising:reducingthedurationofashowerbyaminuteortwoishardly
noticeablegivenahuman’ssensefortimeatthesescales,whileareductionofthewatertemperaturewouldresultinaseverelossofcomfort.
In addition to the energy savings, the average amount of water saved
pershoweris9.3liters.Whiletheenergysavingsaccountforaconsiderableshareofthehouseholdstotalenergybudget,thewatersavingsonly
accountforaverysmallproportionoftotalwateruse.
Inaddition,householdssaveonaverage9.3litersper
shower.
Stabilityoftheeffects
Forboth,thedifference-in-differencesandtheregressionmodel,theresults are robust to different filters (e.g., excluding or not excluding extraction with unbearably hot shower water temperatures or very short
(4.5 liters) extractions that probably result from bathroom cleaning rather than from showering). Moreover, the regression shows no significanttimetrendintheeffectsize,indicatingthatthesavingsremainconstant.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Thesavingeffects
arestatisticallyhighlysignificantandstableovertime.
Page13of16
FinalreportontheAmphiro-PWN-study
2016-03-31
Attempttoexplainthelargeeffects
Savingsfromsmartelectricitymetersaretypicallyintherangeof2%to
5% (Schleich et al. 2013, McKerracher & Torriti 2013), amounting to
about80to240kWhperyear.Thisraisesthequestionwhythesaving
effectsobservedinthisstudyaremuchbigger.Severalfactorsmayex-
Thesavingseffectsa
muchhigherthanfor
smartmeteringfor
electricity.Itiseasier
forconsumerstorespondtofeedback
providedinthe
showerthantokeep
trackoftheirelectricityusethatisinfluencedbymanyfactors.
plainthelargereffectsobservedhere:First,thefeedbackonhotwater
useisgivenalreadyduringthebehaviorisperformed,allowingtheuser
to immediately respond to it. Second, the feedback is person-specific
anddirectedtoaspecificbehavior,makingiteasytounderstandandto
relateto.Third,thefeedbackreceiverhasahighlevelofcontrol:Unlike
feedback on electricity, the energy consumed in a shower is not influenced other household members, base load from appliances, etc. and
thevalvecontrollingtheconsumptioniswithinreach.Andforth,thecurtailmentbehaviorrequiresonlyshorttimespanofattention,asalotof
energyisconsumedinonlyafewminutes.Thesefactorsappeartomake
showeringanexcellenttargetforfeedbackinterventions.
Totalenergyandwatersavings/abatementcosts
Sofar,wehavereportedthesavingspershower.Thedatasuggestthata
persontakesonaverage0.85showersperday(oralmostsixperweek).
FortheaverageDutchhousehold(2.3persons),thisresultsinsavingsof
428kWhplus6.7m3ofwaterandwastewaterperyear.
Table 1 summarizes the energy and water savings and the abatement
costs(investmentperkWhsaved)overathree-yearperiod.Threeyears
havebeenchosenasaconservativeestimateforthedevicelifetime.For
theabatementcostcalculation,devicecostsof60EURhavebeenused.
Note that the numbers only reflect the pure cost side: They do not include the households’ savings on their energy and water bill. If those
savings are taken into account, the costs per kWh saved are negative;
thecost-benefitanalysisispresentedonthenextpage.
Householdsize
1person
2persons
2.3person
3persons
4persons
Energysavings
558kWh
1’117kWh
1’284kWh
1’675kWh
2’233kWh
Watersavings
8.7m 17.5m 20.1m 26.2m 35m Investmentper
kWhsaved
0.107EUR
0.054EUR
0.047EUR
0.036EUR
0.027EUR
3
3
3
3
3
Table1:Waterandenergysavingsfordifferenthouseholdsizesandabatementcostwithinthreeyears
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page14of16
FinalreportontheAmphiro-PWN-study
2016-03-31
Monetarysavingsfordifferenthouseholdsizesandheatingscenarios
Forindividualhouseholds,themonetarysavingscanbeestimatedusing
current water and energy prices. For the calculation, we differentiate
between two heating types (gas and electricity), which are both common and which have different cost of energy. Table 2 summarizes the
monetarysavingsoverathree-yearperiod.Cellshighlightedinboldindicateareturnoninvestmentinlessthanoneyear.
Thefollowingutilitycosthavebeenusedforthecalculation:
•
•
•
•
Costofdrinkingwater:1.83EUR/m3
Costofwastewater: novariablecost
Energycostgas:
0.079EUR/kWh
Energycostelectricity: 0.23EUR/kWh
Householdsize
1person
2persons
2.3person
3persons
4persons
Heating:Gas 60EUR
121EUR
139EUR
181EUR
241EUR
Heating:Electricity
145EUR
290EUR
333EUR
435EUR
580EUR
5
Table2:Monetarysavingsfordifferenthouseholdsizeswithinthreeyears
5. Implicationsandconclusion
Theresultsshowthatreal-timefeedbackonshowerwateruseleadsto
verylargesavingeffects.Thisespeciallyholdsforhouseholdswithmore
thanonperson.Forfamilies,thepaybackperiodsarelessthanoneyear,
even at today’s low energy prices for fossil fuels. The effects are also
noteworthygiventhatthetechnologyisenablingratherthanpatronizing
bynature:Itgivestheuserstheopportunitytoactinlinewiththeirintentions rather than relying on higher energy prices or restrictive
measuressuchasflowrestrictors.
Large-scalesavings
canbeachievedwith
abatementcostof
lessthan0.05EUR
perkWh.
Another strength of the feedback intervention is its compatibility with
the vast majority of showers. Unlike most other technologies for heat
energyconservationthatsufferfromsplitincentives(wherehomeownershavetoinvestwhiletenantsbenefitfromlowerenergybills),itsin
5
Note:Thecalculationusesaveragevaluesforheatingefficiency.Mostlikely,gasheatinghasalower-than-
average efficiency (savings are slightly higher than in the table), and electricity heating has a higher-thanaverageefficiency(savingsareabitlower).
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page15of16
FinalreportontheAmphiro-PWN-study
2016-03-31
stallation makes sense both for tenants and homeowners. This makes
thetechnologywellsuitedforlargescalecampaigns.
From the perspective of administrators of an energy efficiency campaign, the abatement costs (i.e., the investment necessary to save one
kWh)areverylow comparedto othermeasures.Whenspecificallytargeting families (three-person households), the investment per kWh
savedis0.036EUR.
A challenge related to the adoption of the feedback intervention remains.Consumersrarelyrelatehotwatertoenergyuse.Consequently,
hot water conservation is not considered a saving target even among
very environmentally concerned citizens. This makes marketing (hot)
water saving technologies a challenge, probably requiring an enduring
approachtoraisetheawarenessamongcitizens.
BitstoEnergyLab,ETHZurichandUniversityofBamberg
Page16of16