1、Two leading researchers discuss the value of oddball dataUSINESSES OFTEN FACE big,messy problems-challengesMARTIN:So you attacked thethat defy precise definition,change constantly,and have numer-mystery by asking what could beous causes.To find solutions,says Roger Martin,dean of thetrue,rather than
2、 what was true.AndUniversity of Torontos Rotman School of Management and author ofyou narrowed the field of inquiry andThe Design of Business(Harvard Business Press),managers must movecame up with an evidence-basedthrough aknowledge funnelcomprising three stages:staring into theheuristic that whittl
3、ed the mysterydown to a manageable size.mystery;developing a heuristic,or rough rule of thumb;and creating analgorithm,or step-by-step formula for addressing the problem.To exploreSCHERER:I focused closely onthis process,Martin turned to someone who has wrestled with a con-what children with autism
4、do-re-founding problem for two decades:world-renowned autism spectrumpetitively overanalyzing numericaldisorder and genomics researcher Stephen Scherer of Torontos Hospitalpatterns,for example-and why theyfor Sick Children.This is an edited version of their conversation.do it.And I inferred that the
5、 geneticdeletions and duplications I saw,known as copy number variations,MARTIN:Im very interestedjust noise,and we should focus onpredisposed some children to havethe results we see again and again,developmental imbalances,whichin how managers,scientists,anddesigners tackle so-called wickedthe conf
6、irming data.What did youled to behaviors typical of autism.problems-big,unwieldy puzzlessee in the oddball data?Since then weve been able to iden-that look utterly unmanageable.tify a few autism-susceptibility genes.SCHERER:I noticed a pattern inTake the origins of autism:YouMARTIN:Most research isg
7、enetic variations:Children withcould spend years just figuring outautism had more deletions and dufocused on creating reliable out-what angle of attack to take-is itcomes-outcomes that can be conplications of genes than usual,andgenetic,environmental,viral?Mostsistently recreated.Thats a big partthe
8、 variations tended to occur onproblem solvers would start with ahypothesis,test it,and then look forspecific genes.Scientific literatureof the scientific method.But I thinkcontended that 99.9%of the humanthe major leaps forward in knowl-data that confirm or disprove it.Butedge come from focusing,as
9、you did,genome-the genetic instructionyou took a different approach.Canon achieving a valid outcome-oneyou explain your process?that actually answers the questionSCHERER:Autism is a vastto give us the information we reallyAnswers to difficultneed.It seems to me that reliabilityproblem;no single rese
10、archer or laband validity are inherently incompat-can take on its full breadth.I focusedproblems can oftenible-that to achieve a valid outcome,on just one piece of it:the data thatone must incorporate some aspectseverybody else was throwing away.be found in the datapoints that dontof the subjectivit
11、y and judgmentI call it the garbage-can approach.that are typically eradicated in theMy belief is that answers to reallyseem to fit.quest for a reliable outcome.I feeldifficult problems can often be foundthat the distinction between reliabil-in the data points that dont seem toity and validity is at
12、 the heart of thefit existing frameworks.To me,thosebook that tells our bodies how toinnovation dilemma.little variations are like signpostssaying Dont ignore me!Evolu-develop-was exactly the sameSCHERER:I agree.Quantitativetion doesnt tolerate useless junkfor all of us and that the variationsmeasur
13、es that produce reliabilityfor long,so all data points,even theamounted to just an altered letteroften strip away nuance and contextoddball ones,need to be consideredhere and there,so to speak.But ourand thus sometimes even preventseriously.group found that the deletions andthe discovery of a valid
14、outcome.Byduplications were important-thatMARTIN:I think most peopleemphasizing validity,even at thethey were more like textual differ-would sweep the anomalous stuffexpense of reliability,a researcherences than mere typos.We set outunder the carpet-theyd say itscan get to a heuristic that movesto explain them in autism.the work forward.Focusing on the26 Harvard Business Review I November 2009 I hbr.org
