Transcript

1|©2019PurdueUniversity

March2020

Farm Data Usage in Commercial Agriculture ByNathanDeLay,NathanaelThompson,andJamesMintert

IntroductionThereisalotoftalkabout“bigdata”inagriculturethesedays.Thefarmofthefutureissaidtousemultispectralimagery,soilandmicro-climatesensors,equipmenttelematicsdata,andGPStodriveyieldenhancingdecisions.Thegrowthofag-techstartupssuggestsinvestorsareoptimisticaboutthisfuture.Investmentintheag-techsectorgrew43%in2018tonearly$17billionaccordingtoAgFunderNews1.Thoughtheamountofdatabeingcollectedfromfarmsisgrowingrapidly,littleisknownabouthowfarmersleveragethisdatatomakedecisions.AccordingtheUSDA’sAgriculturalResourceManagementSurvey(ARMS),61%ofcorngrowersusedayieldmonitorin2010butonly34%usedthedatatogenerateayieldmap,indicatingadisconnectbetweendatacollectionanddataaction.Thenearly$1billionsaleofClimateCorptoMonsanto(nowBayer)highlightedthevalueofaggregatingfarmdatawithsoftwareplatforms,butquestionsregardingfarmers’useofdataservicespersist.Whattypesoffarmdatasoftwaredofarmerssubscribeto,andtowhatextentdoesthissoftwareinfluenceseed,nutrientandchemicaldecisions?

TheSurveyTobegintoanswerthesequestions,researchersfromPurdueUniversity’sCenterforCommercialAgriculturesurveyed800cornandsoybeanproducersabouttheircollection,management,andusageoffarmdata.Thesurveywaslimitedtofarmswith1,000acresormoretoproduceasampleoffarmsmostlikelytohaveactivedatastrategies.Thesurveyaskedrespondents32questionsregardingfarmdemographiccharacteristics,precisionagricultureadoption,typesofdatacollectedonthefarm,farmmanagementdecisionsinfluencedbydata(ifcollected),datamanagementpractices,andanydatasharingwithoutsideserviceproviders.Thegoalofthesurveyistounderstandthefarmdatalifecyclefromcollectiontodecisionmakingandthevariouschannelsthroughwhichdatabecomesactionable.

SampleDemographicsSamplefarmcharacteristicsaredisplayedinTable1.Fiftypercentoffarmssurveyedoperatebetween1,000and1,999acreswhile36%arebetween2,000and4,999acres,and15%have5,000acresormore.Thesampleismorerepresentativeoflargecommercialoperations(bydesign).About80%ofsurveyedfarmshaveowner/operatorsovertheageof50and35%areovertheageof65.GiventheaverageageofproducersintheU.S.is59,oursampleisgenerallyrepresentativeofthenationalagedistributionoffarmers.2Justunderhalfofsampledfarmshaveabachelor’sdegreeorhigherasthehighesteducationalattainmentamongfull-timeemployeesand9%haveapost-graduatedegree(Master’sandup),indicatingahighdegreeofhumancapital.

PrecisionAgricultureAdoptionHighspeedinternetaccessisslightlymoreavailableamongsurveyrespondentsthanruralAmericaasawholeat80%.Generally,adoptionratesofprecisionagriculturetechnologiesreflectthelargecommercialsizeandcropmixoffarmsinthesample.GPSguidanceorauto-steerforfarmequipmentisusedbyover90%ofthesurveyedfarms.Fifty-ninepercentoffarmsusevariableratetechnology(VRT)forseedingand71%useVRTforfertilizerapplication.Dronesorunmannedaerialvehicles(UAVs)areusedby26%ofsampledfarms.ThesurveysamplehassignificantlyhigherratesofprecisionagricultureadoptionthanthemostrecentestimatedfromtheUSDA

2|©2019PurdueUniversity

AgriculturalResourceManagementSurvey(ARMS)(seeSchimmelpfennig,2016)butarehighlysimilartoworkbyThompsonetal.(2018)whouseasimilarsamplingmethod.3

DataCollectionFarmerparticipantswerefirstaskedabouttheircollectionofthreecommontypesoffarmdata—yieldmonitordata,gridorzonesoilsamplingdata,anddroneorsatelliteimagerydata.Collectionamongoursampleiscommon—particularlyforyieldmonitorandsoildataat82%and77%,respectively.Satelliteordroneimagerydataistheleastlikelytobecollected(47%ofthesample)butgiventhenoveltyofthetechnology,thiscouldbeconsideredhigh.Thevastmajority(73%)createGPSmapsfromtheirdata,suggestingahighdegreeofengagementwithdataoncecollected.

Datacollectionisstronglyrelatedtofarmcharacteristics.Figure1displaysthepercentageoffarmscollectingeachdatatype,brokendownbyfarmsize.Datacollectionismostprevalentamonglargefarms—aresultconsistentwithpreviousfindings.3Therelationshipismostpronouncedfordroneorsatelliteimagery.Farmswith5,000acresormoreare51%morelikelytocollectimagerydatathanfarmsinthe1,000-1,999-acrecategory.Collectingimagerydata—particularlyviadrone—mayinvolvescaleeconomiesthatfavorlargeroperations.

3|©2019PurdueUniversity

Farmswitholderoperatorsaregenerallylesspronetocollectfarmdata,dependingonthedatatype.Figure2shows94%offarmswithoperatorsundertheageof36collectyieldmonitordatavs.80%ofthoseovertheageof65.Thisshouldbeinterpretedwithcautionhowever,duetothelowrepresentationofyoungfarmersinoursample(n=17).Thecollectionofsoilsampledatadropsoffsignificantlyforoperatorsovertheageof50,whileasimilarlysharpdeclineisobservedforimagerydataamongthoseolderthan65.Asagroup,operatorsage65andunderare29%morelikelytocollectdroneorsatelliteimagerydatathanthoseover65.

4|©2019PurdueUniversity

Figure3depictstherelationshipbetweenfarmeducationalattainmentanddatacollection.Collectionispositivelyassociatedwitheducation,thoughbeyondaBachelor’sdegree,yieldmonitorandsoilsamplingdatacollectionratesareindistinguishable.Again,imagerydatafromadroneorsatellitebearstheclearestrelationshiptoeducation.Thirty-eightpercentofthosewithahigh-schooldiplomacollectimagerydatavs.59%ofthosewithapost-graduatedegree—a55%difference.Simplygoingfromnocollegetosomecollegeraisesthelikelihoodofcollectingimagerydataby16%.Thisstrongcorrelationwitheducationalattainmentislikelyduetothenoveltyandtechnicalnatureofimagerydatarelativetootherformsofdata,favoringthosefarmswithtechnicalskillsattheirdisposal.

Ofthe800respondents,58(7%)donotcollectanyofthedatatypesincludedinthesurvey.Non-datacollectorsidentifiedtheprimaryreasonfornotcollectingfarmdata.Figure4showsthedistributionofresponses.Thirty-sixpercentsaiddatacollectionis“toocostly”while19%findthebenefitsofdoingsounclear.Takentogether,overhalfofnon-collectorsperceivefarmdatatobeun-profitable.Overone-thirdreportuncertaintyinhowtousefarmdataoncecollected—suggestingadisconnectbetweencollectionandaction.Surprisingly,only10%offarmscitedprivacyconcernsasthereasonfornotcollectingfarmdata.Privacymaybeofgreaterconcernwhenitcomestostoringandsharingfarmdatabutdoesnotappeartobeasignificantdeterrenttocollection.

5|©2019PurdueUniversity

Amongthosenotcurrentlycollectingdata,fewindicatedthattheywillbegincollectingdatainthefuture—thoughdifferencesemergeacrossdatatypes(seeFigure5).Seventy-sixpercentareunlikelytobegincollectingaerialimagerydatacomparedto43%foryieldmonitordata.Yieldmonitorsoftencomestandardonnewcombineharvesters—notrequiringadedicatedinvestmentoftimeandcapital.

6|©2019PurdueUniversity

DataDecisionMakingFarmersthatcurrentlycollectdatawereaskedtoratetheextenttowhichtheirdatainfluencestheirdecisionmakinginthreecropmanagementareas:seedingrates,nutrientmanagement/fertilizerapplication,anddrainageinvestments.Figure6summarizestheresponses.Farmdataappearstohavethelargestinfluenceonnutrientmanagementwith93%reportingtheirfertilizerdecisionstobe“somewhat”or“highly”influencedbydata.Theshareoffarmsreportingseedingrateanddrainagedecisionsasatleastsomewhatinfluencedbydatais81%and71%,respectively.Fertilizerapplicationdecisionsarenearlytwicelikelytobehighlyinfluencedbyfarmdataasseedingrateanddrainageinvestmentdecisions—reflectingthepopularityofvariableratefertilizerapplicationwithinthesample(seeTable1).

Farmsmakingdecisionsbasedontheirdataappearsatisfiedwiththeresults.Seventy-twopercentofthosemakingdata-drivenseedingratedecisionsreportapositiveyieldimpactvs.81%forfertilizerdecisionsand85%fordrainagedecisions.Interestingly,drainageisthemanagementdecisionleastinfluencedbyfarmdata.Butthosewhousedataintheirdraininginvestmentdecisionsreportthehighestlevelofsatisfaction.Levelsofsatisfactionriseasfarmerscollectmoredatatypes.Forexample,theproportionindicatingapositiveyieldresultfromdata-informedseedingratedecisionsis64%ifthefarmonlycollectsonlyonetypeofdata(e.g.justyieldmonitordata)butrisesto77%ifthefarmcollectsallthreedatatypes—a21%increase.Thissuggeststhatthereturnstodatacollectionmaybecomplementary,thatis,individualdatastreamsaremademoreactionablewhencombinedwithotherdatasources.

DataManagementPracticesThesurveybroadlyfocusesontwodatamanagementpracticesinthefarmdatapipeline:adoptionoffarmdatasoftwareplatformsandsharingofdatawithoutsideserviceproviders.Figure7shows

7|©2019PurdueUniversity

theadoptionratesoffarmdatasoftwarebyfarmsize.Overall,47%offarmsthatcollectdatauseatleastonedatasoftwareproduct,butadoptionratesaresignificantlyhigheramonglargeroperations—63%offarmswith5,000acresormorevs.36%offarmsinthe1,000-1,999-acrecategory.

Farmswithhighereducationalattainmenthavehigherratesoffarmdatasoftwareadoptionbuttherelationshipvariesbyoperatorage.Figure8showsthat,amongoperatorsover65,thosewithsomecollegearenearlytwiceaslikelytousefarmdatasoftwarethanthosewithahighschooldiploma.GettingaBachelor’sdegreehasasimilareffectonadoptionratesamongthoseage51-65.Softwareplatformsarepopularwithyoungoperatorsacrossalllevelsofeducationbutadoptionrisestonearly70%forthosewithapost-graduatedegree(e.g.Master’sorPh.D.).Focusingoneducation,largedifferencesinsoftwareuseacrossagegroupsaremostapparentattheendoftheeducationspectrum.Thesedifferencesmayhighlighttrendsineducationalattainmentovertime.Amongolderoperators,havingsomecollegerepresentsarelativelyhighlevelofeducationalattainmentwhileyoungeroperatorsrequiremoreeducationtodistinguishthemselves.Adoptionofsoftwareplatformsmayberelatedtothedegreetowhichoperatorsandoperationsspecialize.

8|©2019PurdueUniversity

Farmersthatuseatleastonedataserviceplatformwereaskstoidentifyalloftheproductstheycurrentlysubscribetofromalistofeightpopularbrands(seeFigure9).5ThemostwidelyusedsoftwareproductisClimateFieldView(Bayer),usedbyoverhalfofsurveyedsoftwaresubscribers.Forty-fourpercentuseJohnDeereOperationsCenterwhile22%useCaseIH’sAFSSoftwareplatform—generallyreflectingtheirrespectivemarketsharesforfarmequipment.Trimbleisthenextmostfrequentlyusedat21%,followedbyFarmersBusinessNetwork(FBN)(19%),Corteva’sEncirca(14%),FarmersEdge(10%),andGranular(alsoCorteva)at9%.6Nearlyonefourthofuserssubscribetoaservicenotlistedinthesurvey,suggestingalongtailinthefarmdatasoftwaremarket.Understandingwhattypesofsolutionsmakeupthistailisworthyoffutureinvestigation.

9|©2019PurdueUniversity

Thoughallofthesoftwarebrandslistedprovidefarmdatasolutions,thereisasignificantamountofproductdifferentiation.OperationsCenterandAFSplatformscollecttelematicsandas-appliedseedandchemicaldatafromtheirrespectiveequipment(thoughtheyarecapableofintegratingwithotherdatasources).FBNisprimarilyadataaggregatorforinputcostandperformancebenchmarking.Otherproductsprovidecloud-basedstorageandanalysisofagronomicdataforin-fielddecisionmaking(e.g.FieldViewandFarmersEdge)whileGranularprovidesabroadsetofsolutionsfromworkflowmanagementtolandvaluation.Itisnotsurprisingthenthat70%ofsoftwareuserssubscribetomorethanoneproduct(seeFigure10).Oursurveyindicatesthat63%ofsubscribersreceiveseedorfertilizerapplicationrecommendations(prescriptions)fromtheirsoftware.However,farmersdonottreatsoftwarerecommendationsasdirectives.Only44%followtheirsoftwarerecommendations“closely”while52%follow“somewhatclosely,”and4%donotfollowrecommendationsatall.

Onaverage,farmsusebetweentwoandthreesoftwareplatformsbutalmost90%subscribetothreeorfewer.Giventhegrowthofinvestmentinfarm-facingtechnologycompanies,itmaybedifficulttoincentexistingadopterstoaddanotherproducttotheirsoftwaresuite.Companiescouldinsteadtargetnon-adopters.Figure11showsthat,offarmersnotusinganyfarmdatasoftware,closetohalfindicateuncertaintyinhowtousethetechnologyastheprimaryreasonfornotsubscribing.Forty-onepercentofnon-adoptersperceivefarmdatasoftwareastoocostlyortheassociatedbenefitsunclear,indicatingabreakdowninvalueproposition.Privacyconcernsareagainsurprisinglyunimportantasadeterrenttosoftwareuse—only12%identifiedprivacyasthemainreasonfornotsubscribingtofarmdataservice.

10|©2019PurdueUniversity

Inadditiontowithin-farmdatamanagement,farmerparticipantswereaskedabouttheirdatasharinghabitswithoutsideserviceproviders.Specifically,farmerswereaskediftheysharetheirdatawithagronomists,agriculturalinputsuppliers,andequipmentdealers/manufacturersforthepurposeofgeneratingcropmanagementrecommendations.Over70%ofrespondentssharetheirfarmdatawithatleastoneserviceproviderandofthese,63%sharewithtwoormore.Figure12showsfarmersaremostwillingtosharedatawithserviceprovidersthatoperateclosetoon-farmcropmanagementdecisions.Fifty-eightpercentoffarmssharedatawiththeiragronomistfollowedbyaginputsuppliersat44%.Only12%reportsharingtheirdatawithequipmentdealersand7%sharewithaserviceprovidernotlistedinthesurvey.

Surprisingly,theshareoffarmsthatfollowrecommendationsprovidedbyoutsideserviceproviders“veryclosely”is31%,13percentagepointslowerthanthesharethatfollowtheirsoftwarerecommendationsclosely.Evenwhencomparingthesamesub-sampleofrespondentsthatgetrecommendationsfrombothsoftwareandserviceproviders(191farms)thedifferentialremainssignificant(about10percentagepoints).Thewillingnesstofollowsoftwaregeneratedrecommendationsoverthoseprovidedbyanoutsideadvisormaybeduetoaperceptionthatserviceproviders—particularlyaginputsuppliers—pairrecommendationswithproductsales.

11|©2019PurdueUniversity

About15%ofdata-collectingfarmsneitherusefarmdatasoftwareorsharedatawithoutsideserviceproviders.Forthiscohort,dataissiloedonthefarmandisunlikelytobemadeactionable.Only8%subscribetoasoftwareplatformbutdonotsharetheirdatavs.38%whoonlyshareanddonotusesoftwarewhile39%doboth.Farmsalreadyperformingonedatamanagementpractice(softwareorsharing)aremorelikelytoadoptanother.Thisisparticularlytruefordatasharingvs.non-sharing.Over50%offarmsthatsharedataalsosubscribetoasoftwareplatformcomparedto36%ofthosethatdonotcurrentlysharedata.

Tables2and3showhowcombiningdifferentdatamanagementpracticesisrelatedtoon-farmdecisionmakingandtheresultingyieldoutcomes.Table2showsthatfarmdataismoreinfluentialinthecropmanagementdecisionsoffarmsthatsubscribetoadatasoftwareplatformorsharetheirdatawithanoutsideserviceprovider.Farmsthatperformbothdatamanagementpracticesareoverfourtimesmorelikelytomakeseedingratedecisionsthatare“highlyinfluenced”bydatathanfarmsthatcollectdatabutdonotshareorusesoftware.Theproportionofdata-drivenfertilizerdecisionsamongthesoftwareplussharingcohortisovertwicethatoftheno-software,no-sharinggroup.

Table2indicatesthatsoftwareuseispositivelyassociatedwiththedegreetowhichdatainfluencesdrainageinvestments.Datasharinghowever,haslittletonoimpactontheimportanceofdataindrainagedecisions.Infact,amongsoftwareusers,thosethatalsosharedataareslightlylesslikelytoreporttheirdrainagedecisionsasbeinghighlyinfluencedbydata.Thiscouldbechannelingtheroleofaerialimageryandmappingindraintileinstallation(softwaresubscribersaretwiceaslikelytocollectdroneorsatellitedataand45%morelikelytocreateGPSmapswiththeirdatathannon-softwareusers).

12|©2019PurdueUniversity

Table3cross-tabulatestheproportionoffarmsreportingapositiveyieldimpactfromtheirdata-informeddecisionsbysoftwareuseandsharingpractices.Again,farmsthatsubscribetoadatasoftwareproductaremorelikelytoreportincreasedyieldsasarefarmsthatsharedatawithanagronomist,inputsupplier,dealer,orotherserviceprovider.Thedifferenceisespeciallylargeforseedingratedecisions.Comparedtofarmswithnoactivedatamanagementstrategy,farmsthatusesoftwareandsharedataare60%morelikelytomakeyield-increasingseeding-ratedecisions.

ConclusionDespiteintenseinterestinagriculturaldataamongthefarmpressandventurecapital,littleisknownabouthowfarmsactuallycollect,manage,andanalyzedata.Thissurveyprovidesausefulglimpseintothefarmdatalifecyclefromcollectiontoactiontoevaluation.Wefindthatamonglargecommercialcornandsoybeanoperations,datacollectioniscommon(92%collectatleastonetypeoffarmdata)andthatcollectionisstronglyrelatedtofarmcharacteristics.Largefarms,farmswithyoungeroperators,andfarmswithhigheducationalattainmentarethemostlikelytocollectandusefarmdata.Farmsnotcurrentlypursuingadatastrategyareunconvincedbythevaluepropositionandareunlikelytobegincollectingfarmdatainthenearfuture.Mostfarmsthatcollect

13|©2019PurdueUniversity

datasaythatithasatleastsomeinfluenceintheirseeding,fertilizer,anddrainagedecisionsandthevastmajorityreportapositiveyieldimpactfromtheirdata-informeddecisions.

Activedatamanagementpracticescanincreasetheseperceptions.Useofvariableratetechnology(VRT)isastrongpredictoroftheimportanceofdatainseedingrateandfertilizerapplicationdecisionswhiletheinfluenceofdataondrainagedecisionsriseswiththecollectionofdroneorsatelliteimageryanduseofGPSmapping.Resultsshowthatabouthalfoffarmscollectingdatasubscribetoafarmdatasoftwareplatformandusingmultiplesoftwareproductssimultaneouslyisnotuncommon.Again,larger,younger,andmoreeducatedoperationsarethemostlikelytoadoptafarmdataplatform.Privacyissues,bothatthecollectionandsoftwareadoptionstage,aresurprisinglyunimportantrelativetothoseregardingusabilityandprofitability.Over70%offarmsarewillingtosharetheirdatawithanoutsideserviceprovider—mostcommonlyagronomistsandaginputsuppliers.Farmersaremorelikelytocloselyfollowrecommendationsgeneratedbytheiragdatasoftwarethanthoseprovidedbyoutsideserviceproviders.Thisistrueevenamongthesubsetoffarmsthatreceiverecommendationsfrombothsources.

Overall,farmsthatuseanagdatasoftwareproductandsharedatawithanoutsideconsultantaresignificantlymorelikelytomakedata-informedmanagementdecisions.Theyarealsomorelikelytoregardthosedecisionsasyield-enhancing.Oftenagdataendsupsiloedonthefarm,storedondesktopsorUSBflashdrivescollectingdustinashopdrawer.Thesesurveyresultssuggestthatpro-activelymanagingandanalyzingfarmdatacanimprovedecisionmakingandgeneratepositiveyieldresults.Understandinghowdatamanagementpracticesshapeon-farmdecisionmakingisofcrucialimportanceinbringingthefarmofthefutureintoreality.Downstreamplayersintheagriculturalvaluechainmustrecognizethedataneedsofgrowersasdatatransparencyanddataintegrationdemandsrise.

Introduction[1]AgFunder.AgFunderAgriFoodTechInvestingReport:2018YearinReview.(2019).Availableathttps://agfunder.com/research/agrifood-tech-investing-report-2018/

[2]2017USDACensusofAgriculture

[3]Schimmelpfennig,D.(2016).FarmProfitsandAdoptionofPrecisionAgriculture.U.S.DepartmentofAgriculture,EconomicResearchServiceReportNo.217,Washington,D.C.;

Thompson,N.M.,C.Bir,D.A.Widmar,J.R.Mintert.(2018).FarmerPerceptionsofPrecisionAgricultureTechnologyBenefits.JournalofAgriculturalandAppliedEconomics51(1):142-163.

[6]Figure8excludestheunder20and20-35agegroupsduetoinsufficientobservationsineacheducationcategory.

[7]ThelistoffarmdatasoftwareproductswasdevelopedfromaprevioussurveyoffarmdatasoftwareusersandfrominternetsearchdataprovidedbyGoogleTrends.

[8]Recently,EncircawasmergedwiththeGranularsoftwaresuitetobecomeGranularAgronomybyEncirca.


Top Related