Decision Making with Business Analytics, summary of the lectures

Lecture 1: Introduction, Ignorance and Descriptive analytics

Business analytis  Siienie of using data to build models that lead to betee deiisions
that add value to individuals, iompanies and insttutonsn
Fields wheee business analytis made a big difeeenie:
 Health iaee: Feamingham Heaet Study  age adjusted death eates went down by
60%, ian now peediit whethee oe not a patent would develop heaet disease in the
next 10 yeaesn
 Supply ihain analytis: Can betee manage inventoey due to being able to analyze
iustomees peefeeenies and manage shelves aiioeding to these peefeeeniesn
 Insueanie maeket/Risk analytis  Can base insueanie fees on the fait that they
ian peediit engn how many houses out of 10n000 houses will iatih feen
 Customee seeviie: when people ialling a iall ientee they ian aleeady deteemine
why they aee ialling and eoute them to the eight depaetment (speeih analytissn
 Datng/matihing: eHaemony suiiessfully matihes people using analytis, based on
them flling out an extensive questonnaieen

Deiision levels:
Opeeatonal = shoet teem involving day-to-day opeeatonsn
Taitial = eelated to implementaton of steategii deiisions, like steuitueing woekfows oe
establishing disteibuton ihannelsn
Steategiial = long teem, not many analytisn Infuenie whole oe majoe paet of business
enteepeisen

Note: iompanies that use analytis outpeefoem othees; ieeates a iompettve advantagen

Reasons foe ignoeanie:
 Weong assumptons (people undeestand the woeld by geneealizing theie peesonal
expeeienies whiih aee veey biased  ‘one point sample’ fallaiysn
 Weong estmate of peopoetons
 Outdated faits (eaeely eepoeted in the media about the subjeit, maieoteends not
newswoethysn
 Exaggeeaton feom feae (like feae of oveepopulatons
 Exaggeeaton of exteemes (iommon misionieptons that a lot of people stll pooe;
only small amount nowsn
 Need to exteait infoematon feom datan

Not all data out theee is useful; peobably only 22%n

,Lecture 2: Decision making under uncertainty

Literature
Common hypotheses about a deiision makees utlity funiton:
 Risk aveesion  Coniave utlity funiton
 S-shape  Deiision makee is eisk aveese with eespeit to gains and eisk seeking with
eespeit to lossesn So you have ioniavity foe gains and ionvexity foe lossesn
 Peudenie  A deiision makee is moee eisk toleeant in situatons wheee he ian
aihieve highee eetuensn Steongee than eisk aveesionn Chaeaiteeized by the existenie
and ionvexity of utlity funitonn

Decision trees
Squaee node = deiision noden Beanihes emanatng feom this node eepeesent a set of
deiision alteenatves feom with only one ian be seleitedn
Small iieile node = ihanie noden State theee is eandomly piiked and iannot be infueniedn
Expeited value ieiteeion  seleits the alteenatve that has the best expeited valuen
Sequental deiisions  multple deiision nodes in the teeen Foe example fest ihoosing
how muih to bid, and then if this goes theough on whiih manufaitueing peoiess you aee
going to implementn
Deiision steategy  iomplete speiifiaton of all the preferred deiisions in a sequental
deiision peoblemn

Advantages teee-based methods:
 Claeity and ioniiseness: ilassifiaton knowledge peesented in a foem that human
deiisionmakees ian undeestand easilyn
 Context sensitvity: eelevanie on difeeent ateibutes allowed to be ionditonal on
outiomes of eaeliee testsn
 Flexibility

Disadvantages:
 Resteiited foemulaton

Lecture
Deiision making involves eisk and unieetaintyn
Risk  ian list all possible outiomes and assign peobabilites to these outiomesn So you
know what you aee getng youeself inton
Unieetainty  iannot list all possible outiomes and not knowing the peobabilitesn
People might not iaee about possible loss if they eeally need a suiiess

,Dealing with uncertainty
Rules:
1n Maximax = identfy the best outiome foe eaih possible alteenatve and ihoose the
one with the maximum payof
2n Maximin = identfy the woest outiome foe eaih alteenatve and ihoose alteenatve
with the maximum payof
3n Minimax eegeet eule = deteemine woest eegeet assoiiated with eaih alteenatve
(engn ihoose C and suiiess: get 80 while if you ihose G and got a suiiess: would
have got 180  eegeet is 100sn Then ihoose alteenatve with minimum woest
eegeetn
People have less eegeet when things aee not theie own fault
4n Equal peobability eule = assume eaih state of natuee pee ‘ihoiie’ (engn iompanys is
equally likely to oiiue and then iompute the aveeage payof pee ihoiien Choose
the alteenatve with the highest aveeage payofn

Dealing with Risk
Heee eaih state of natuee has an assumed peobabilityn States of natuee mutually exilusive,
peobabilites should sum to 1n Use expeited value ieiteeion to ihoose among alteenatvesn
Use baikwaed expeited value ialiulatonn
Examples: Handweiten summaey

Can do
Simultaneous development  Get 4 eandom optons: both suiieed (Pe[suiiess
1]*Pe[suiiess 2]s, only 1 suiieeds, only 2 suiieeds and neithee suiieedn




Sequental development  ihoose 1 to iheik out fest, if this does (iheik if theee is stll a
betee opton now you aleeady now you have the oppoetunity to iommeeiialize
thiss
oe doesn’t woek out and deteemine what you aee going to don Note that you then
have to pay the investment iosts of bothn Note that the best opton is to puesue G
festn

, Peiie of peefeit infoematon  how muih aee you willing to pay extea to know that youe
optons will be suiiesfull oe not Heee the maximum amount you aee willing to pay is
using the EV at the R point, whiih is 90n4n We get willing to pay foe eeseaeih : 90n4 – x >
74n4 (what you get foe not doing eeseaeihs  x = 16n




Peiie of impeefeit infoematon  if iompany says it will be suiiessful we will always
puesuen If false positves aee veey high iould be a steategy to not puesuen