Data scientists andchiefdata officersare definitely thescorchinguselately, and governmentorganizationsat allstagesareperformingfor gettingextraaway fromtheirfastgrowingtroves of data.Figuring outtips on how toapproach all that data,having said that,can beoverwhelming.
Forcompaniesthat do not haveDJ Patil onspeeddial, amodernTDWI reportprovides agoodplace tostart off. (TDWI, like GCN, is owned by 1105 Media.) "SevenMethodsfor Executinga successfulData Science Strategy" is justjust what thetitleindicates-- a checklist ofsevenbasictenets ofcleverand results-oriented data science.
"Tofixbusinessproblems,developnewsolutionsand services and optimize processes," TDWI's Dave Stodder wrote, "organizationsprogressivelyneed to haveanalytics insightscreatedby data sciencegroupswhich has avariousestablishedofspecializedabilitiesand businessawarenessthat arealso good communicators."For makingthe mostofthis kind ofinvestments, the reportrecommends:
1.Establishyourcriticalbusinessdriversfor data science Just beforegettingstarted out,an organizationought toquestionwhatactualdata scienceattemptscan offerthatcommonbusiness intelligence and analyticsusually are not. Ifthere are actuallygaps,it iscriticalto rentstaffwithreal"knowledge of and curiosityabout thebusiness"to aidfill them.
2. Createan efficientstaff It will requiregreater thancuriosity,on the other hand. Andemployinga multitalentedsuperstar--that is"like chasing unicorns"to beginwith -- cango awayan agencythat has aone-off, artisanalprocedurewhose creator then leaves for greener pastures. "[A] wisertraining courseis usually tocreateasecureteamthatdeliverscollectivelythe skillsofa number ofindustry experts."
3. Emphasize communicationsexpertise "Organizations that use data sciencecorrectlyalmost universallyissueto communicationlike aessentialingredientfor theirsuccess," TDWIobserved.Companiesneed to"make it aprioritybecause theyconsidercandidates for data scienceteams."
4.Increasetheeffectthroughvisualization and storytelling "Data science thrivesin ananalyticssociety," Stodder wrote, but "not allpersonnel...are likely tobeelementof data sciencegroups, norreally shouldthey be."Gettingapproachesthat can helpnon-statisticians grasp the insightsfrom thedata isvitaltoreceivingactualpricefromtheexpense.
5.Provide thedata scientists all the data Even thoughcommonanalyticstypicallyconcentrationon avery carefullydescribedsetof structured data, data science has thelikelyto attractbenefitwithin thebroadmesses of unstructured datathat almost allcompaniescreate. "Data scientistsought toworkcloselywith data ateverystageso that they determine what they have got," Stodder wrote --and so theyreally need tohaveas muchof itas possible.
6. Pavejust howfor operationalizing the analytics Descriptive analytics arehandy, but predictive analytics areconsiderably morevaluable-- and prescriptive analyticsgiveby far the mostlikelyprofitby far.To generatethispossible, "data scienceteamscanmoveawayfrom uncoordinated, artisanaldesignimprovementandtowardtechniqueswhich caninclude things likehigh-qualityfeedbackclassestoproperflaws."