Practical Data Science: Building Minimum Viable Models

MVM is based on the principle that data-based startups need to have affordable data science models for their financial reality but also, these ... JoinNewsletter   PracticalDataScience:BuildingMinimumViableModels DataScienceforstartupsbasedondata:MinimumValuableModel,anewconcepttoavoidafullscale95%accuratedatasciencemodel.WanttoknowmoreaboutMVM?Havealookatthisinterestingarticle. comments ByErnestoMislej,Co-founder,DirectorofDataScienceGroup,7Puentes. Whenwetalkaboutinnovativeservicesorproducts,manystartupsfollowasmoothermodelofdevelopment.Thisallowsthemtominimizetherisktobeabletohaveimprovementswhencollectingcapitaltofinancethemselves.Oncetheyfoundthemarketfit,theissuewillbeaboutthegrowth,toachieveabalancepoint. Forthosestartupsbasedondata(nowadays,mostofthemconsidertheirdataasastrategicactiveforthedecisionmaking),tofindamodelthatinterpretsthemisadifficulttask.Extract/collectdata,measure,modelandmakingdecisionsisacommonroadforanystartupthataimstoadynamic,changing,fludidsectorofthemarket. Itisthedatascientistorthedatascienceteam’stasktofindthat/thosemodel/s,butfindingit/them(determinethemodellingtechnique,settingparametersandadjustment)maybeaverylong,andsometimes,non-alignedtaskwiththebusinesstimes.Forexample:itdoesnotmakesenseamodelto“predicttheresultsofafootballmatch”thatfindstheresultsafterthematchwasplayed.So,howstartupscanminimizethisriskwhenlaunchinganewapp?Dotheyneedsomuchdeploymenttoenterthemarket,dotheyhavethenecessaryresources?In7Puentesweunderstandtheydonotandthatiswhywecoinedanewconcept:MVM(MinimumValuableModel). MVMisbasedontheprinciplethatdata-basedstartupsneedtohaveaffordabledatasciencemodelsfortheirfinancialrealitybutalso,thesemodelshavetobeacceptableintermsofaccuracy.Thisworkingmethodologybasedonminimumandeffectivemodels,minimizestherisksintheeventtheproductdoesnotsucceedinthemarketand,therefore,isanobstaclelessinregardstothelaunching. Adatasciencemodelwith75%ofaccuracy,whichisacceptabletoguaranteethewell-functioningoftheapp,takes25%ofthetime.Toescalatetoa100%,i.e.,toaperfectmodel,exponentiallyincreasesthetimeusedandtherequiredinvestment.Ifwethinkamodelofrecommendationforanapplike“Tinder”,aMVMdoesnotneedthe10offerstobeideal,but,tohaveoutof10offersanaverageofgoodoffersand,maybe,alowpercentageofverybadoffers.Itisnotnecessarytodevelopapredictionalgorithm100%effectiveanditisnotfeasibleinfinancialterms. Everysector/projecthasits“good-enough”:sometimesthepriorityisaquickresponsebutinothercasesthecoveringisthefocus. TofindaMVM,itisnecessaryaconstantdialoguebetweentheareasthatdefinethebusinessgoalsandthedatascientist. Itisnousethespecialistworkingonlytwomonthswiththedata,sincefindingtheMVMrequirestopayattentiontowhatdataprovide.Manytimes,thebusinessareasrequireaveryprecisemodelwithtrainingdatathatisnotenough,theyarenoisyortheydonotadjusttothethoughtmodel. Maybeitisbettertoreducethescopeofthemodeltotheportionofthedatawhereitbetterworksand,inthefuture,expandthecoverageofthemodelwhenthestartuphasbetterfinancialresources. Morethan70%ofdatascienceproject’seffortsconsistondata-junk:collectionandcleaningofdata.Andthetimeformodeling,experimentingandcommunicatingresultsistooshort.SothatMVMmodelcomestoacceleratetheknowledgeextractionprocessfroma“lean”perspective. Bio:ErnestoMislejisaco-founderof7PuentesandDirectorof7PLabs,theDataScienceGroupof7Puentes.HeisalsoaMachineLearningandDataMiningprofessorintheMasterofDataMining&KnowledgeDiscoveryatBuenosAiresUniversity. Related: BigDataScience:Expectationvs.Reality HowtoStructureYourTeamWhenBuildingaDataStartup 4MajorTrendsDisruptingtheDataScienceMarket MoreOnThisTopicPracticalDeepLearningfromfast.aiisBack!KDnuggets™News20:n38,Oct7:10EssentialSkillsYouNeedtoKnow…WhatMakesPythonAnIdealProgrammingLanguageForStartupsCalculusforDataScienceALayman'sGuidetoDataScience.Part2:HowtoBuildaDataProjectAlternativeData,TextAnalytics,andSentimentAnalysisinTradingand… GettheFREEcollectionof50+datasciencecheatsheetsandtheleadingnewsletteronAI,DataScience,andMachineLearning,straighttoyourinbox. BysubscribingyouacceptKDnuggetsPrivacyPolicy Leavethisfieldemptyifyou'rehuman: <=Previouspost Nextpost=> TopPostsPast30Days FreePythonforDataScienceCourse HowtoSelectRowsandColumnsinPandasUsing[],.loc,iloc,.atand.iat 5TrickySQLQueriesSolved DecisionTreeAlgorithm,Explained TheCompleteDataScienceStudyRoadmap 7TechniquestoHandleImbalancedData FreePythonProjectCodingCourse TopProgrammingLanguagesandTheirUses 5DataScienceSkillsThatPay&5ThatDon’t EverythingYou'veEverWantedtoKnowAboutMachineLearning LatestNews DataAnalystSkillsYouNeedforYourNextPromotionTheMachineLearningLifecycleBepreparedtomanagethethreatwithanMSinCybersec...DimensionalityReductionTechniquesinDataScienceTheAbsoluteBasicsofMLOpsIMPACT2022:TheDataObservabilitySummit,onOct.25-26 TopPostsLastWeek HowtoSelectRowsandColumnsinPandasUsing[],.loc,iloc,.atand.iat FreePythonforDataScienceCourse 5DataScienceSkillsThatPay&5ThatDon’t 7DataAnalyticsInterviewQuestions&Answers 5TrickySQLQueriesSolved MoreRecentPosts IMPACT2022:TheDataObservabilitySummit,onOct.25-26TheMistakeEveryDataScientistHasMadeatLeastOnceFreeMicrosoftExcelforBeginnersCourseBuildaText-to-SpeechConverterwithPythonin5MinutesKDnuggetsNews,September21:7MachineLearningPortfolioPro...LearnHowDifferentDataVisualizationsWorkData-centricAIandTabularDataMorePerformanceEvaluationMetricsforClassificationProblem...HowToCalculateAlgorithmEfficiencyTopPostsSeptember12-18:HowtoSelectRowsandColumnsinP... RelatedPosts GettingDeepLearningworkinginthewild:AData-CentricCourseHowdoIdothatinPython?TopOctoberStories:DataScienceMinimum:10EssentialSkillsYouNeedto…DataScienceasaProduct-WhyIsItSoHard?KDnuggets™News20:n09,Mar4:WhenWillAutoMLreplaceData…BuildYourFirstDataScienceApplication GetTheLatestNews! GettheFREEcollectionof50+datasciencecheatsheetsandtheleadingnewsletteronAI,DataScience,andMachineLearning,straighttoyourinbox. BysubscribingyouacceptKDnuggetsPrivacyPolicy Leavethisfieldemptyifyou'rehuman: KDnuggetsHome»News»2016»Nov»Opinions,Interviews»PracticalDataScience:BuildingMinimumViableModels ©2022KDnuggets | About | Contact | PrivacyPolicy | TermsofService   PublishedonNovember8,2016by SubscribeToOurNewsletter (Get50+FREECheatsheets) Leavethisfieldemptyifyou'rehuman: GettheFREEcollectionof50+datasciencecheatsheetsandtheleadingnewsletteronAI,DataScience,andMachineLearning,straighttoyourinbox. BysubscribingyouacceptKDnuggetsPrivacyPolicy Leavethisfieldemptyifyou'rehuman: GettheFREEcollectionof50+datasciencecheatsheetsandtheleadingnewsletteronAI,DataScience,andMachineLearning,straighttoyourinbox. BysubscribingyouacceptKDnuggetsPrivacyPolicy Leavethisfieldemptyifyou'rehuman: