RZEEARCE ANZ :17=32?~Z";-, Ba!%renfeld, 2nd July -984
    TRANSLATING 7757 R---SUL-.S -rNTO PRODUCT
    One of the =ile$Z0=s for producz develc;ezs suzzess is
    the market zesearch test which shows, whether he has
    reached t"e target or failed. In both cases, when he is
    ai=ing to c:~ange an exist-;zq p=TW5z'ct without changing the
    taste and tze acceptance of-lZat product as well when he
    is aiming %= reach the briefizq?%.qf a bew product develop-
    meet, the c--nsumer test result QCU show how close he got
    to t'~tat tarTets.
    ror th.esAL two main objectives in a prc;k*ct development
    department the changing of existing brauft-and the
    development al. new products we have two difftrent market
    research tests,' v#ich use the same items and*'*Skales--
    a monadic test and-a pair comparison test.
    Slide 1 Our aim was to do som&,jaore systematic work to better
    understatd the consmar-z"t measurements as well as
    getting i--!ormations& wbick.product properties influence
    tJ2* senscry. judgement.
    Principially there are two uays.~p achieving a better
    understatdingr what the conswmar -tdst is measuringt
    Slide 2 By comparing the consumer judgeamt w:Lt# the result
    J
    from. expert smoker panel one can atranslimea one
    language ;.to another.
    The other way is to relate product or Saok& pi~*qerties
    with the czns=er judgement.
    0269
    cr
    BATCo document for Legal Services: Health Canada 21 October 1999

    




    - 2 -
    Both ways will i=prove our a*---J-1i-_y tc *--etter and faster
    reach t'-* taste-targets !or a given pr--duct development.
    The difficulty is that one can not get these infor=atioft
    only as & waste fro= nor--a! product testing. it. is
    necessav, to develop and test cicaret:es only f~r the
    parpose of producing this type of know '-Icw.
    2. B.A.T at7rcacb tc extlain-and aredict
    ccnsu--er test results-
    During --y presentation I %&ill try to quide you along
    the way we walked up to noie'~n explaining and predict
    -cons=er test results by using-product properties as
    predictcrs for sensory ratings::!-
    2.1 is there a relatinsh-;= betwe-en'-sr-cke deliveries
    and sensorv test results ?
    The starting point was the hypothesis thaCiX few of the
    ensory Lapresai.ons during smoking should hi~e something
    o do with the amount of smoke a smoker takes out of a-
    cigarette. The sensory judgement of cigarettes is done
    on a per puff basts.'Tbe result of the final judgement
    is a mental inteqratioa'~-af multiple single-puff
    judgements.
    One should not expect that the -relations between
    deliveries and sensory rati:q art always monotonous:
    in a set of pair comparison tests ue,,found that the
    Slide 4 sfullness4-rating was related with deckmasing per puff
    deliveries, but not related with f ea ifzq per puff
    deliveries. That means, that an increase in/per puff
    deliveries would not always increase that *ful.1hess'-
    rating whereas a decrease in per puff delive;i"-would
    always result in a decrease in the "fullnesSO-ratfng.
    Tne '3trengthP rating was different from that: hem
    it seems that increasing per puff deliveries also
    increase the strength rating.
    0270
    "No
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    3
    I must say, that we got th-s resu-t !rom a set 0! very
    i. ;.laz troducts, and it is not clear, whether this
    relationship is a more fundamental one.
    Another txample, which showed similar results, comes
    fr= a =:-a4ic test o! UB and KRONL amoig HB- ant
    KRONE-s=-~ker: R5 is a full flavour c.,garette,
    KRONE a -'2w tar cigarette.
    Slide 5 In these tests EB-smokers judged KRONE lower in st:enct-1h
    and less "full* as one. would expect. KRONE-smoker judged
    EB hic~-er in strengtt, wiLICh is &130 Aot unexpected
    but eq-za-1 it 1!ullnesso as their own brand KRONE.
    In ancther monadic test with cigarett*s with dif!erent
    =-'=t4.- deliveries we did not fi~cf an influence on the
    fullness rating but a clear relatioir.,-.betweea per puff
    nicatin and the strength rating.
    Putting all =r datas from product testing-." ether,
    which means comparing a sample of very dif!ec-ent products
    there is noaxly n a correlation between fullness and
    deliveries, but a satisfactory correlation between
    strength and per putidelivezies. The per pu."! deliveries
    explaim roundabout, 40 %-4d the total variance of the
    streogth rating (and other related dimensions
    lik* i:ritation).
    suzMarising these results one caz-- state:
    Slide 6
    CD
    BNW 0271 CO
    011
    Cn
    113
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    2.2 Influence of oer-au-f!nicatin on sensorv test results
    A!ter we got to ttz: ;oinz we wanted tc know, whether tar
    or ticati-n is more responsible fcr the strength rating.
    We d av a!: ;-ot ttx---* ciq a rettez with di f f ere nt
    nicetin but equal pez-pu-4! tar. All three sam;les were
    evaluated in a Monadic test by a BB- and a KRONZ-sxoker
    panel.
    There was one surprising result: the smokers only
    di3tinguished between t!=Ze C-4g&-'ettSZ whcae tic-otin
    deliveries were close to their mostly smoked brand.
    slide 7 An increase of 0,015 =q nicotim per puff caused a
    si.Sni!-cant -nerease in the strv~agth raLt-Ing of
    0,5-to 0,7 and an decrease in mildstess of about
    1,0. We also found that the higher atcotin figures
    were rakated with higher sharpness ratij%Q3.
    There was also a signif icant change in thij,'"tirq
    for the item hurb.
    Up to now we do have indications that nicotin has a
    much bigger influence an the strength rating than tar,
    but I hope during tho rAzt month we will got a much
    clearer picture, after we-iave analysed the results
    from the just running tests.
    In sumarising this results one:Ccan state:
    Slide 8
    I
    1 ri
    ZD
    a ~w 0272
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    2.3 influence cf sroking benavicur on senscrv
    test resu:zs
    One pzz~>La= in a -laining sensory results by smoke
    deliveriee- is that we dc not know the pu!! vo-ume c---
    s=zckars. Zzcauza a ch.ange in the pu!f vo~-Ume wLll
    change the deliveries, any explanation of sensory
    results by standczd,deliveries will be i=Poss-!:_'e !cr
    a product, which be:,r&ve different under changing pu!!
    vclu=es.
    The foi!owing example will ithow ycm such a dilemma:
    Slide 9 -.Two low tar products with neajdY ident4cal per pu!!
    (last tvc~ daliverles are significantly diifeient in the strength
    rows ratizg.
    masked)
    in this dase we measured the puff volumv-9f an smokers
    who judged Jim two cigarettes and wo analyi4d t3w%.
    Slide 10 cigarettes at &-f!erent puff volumes. The two-:cigarettes
    behave quite diff4rent at higher puff volumes: cigarette B
    has the higher per-puff-nicotin with increasing puff
    volume. At the me= piig volume of the smoker group the
    diff-feronco is in tha zanis.of 0,014 mg nicatia per pu!!.
    1, r= the investigation aboii the influence a! p*r-pu!!
    nicotin an the stzength z&tiA4),w* would expect for this
    pair of cigarettes a diffezance:i'n the strength rating
    of about 0,5 to 0,7. The diffecanceefound was 0,8
    V_
    Slide 9 Because we normally do not have measured puff volumes
    Clast we U212 a mathematical te= called 0ela3iiaitya as a
    two correlation factor. Dirk Pangritz will qivf~.-F his
    rows) presentation same more information abaut.61astiaity and
    how it is used in explaining sensory test resultit,,
    '3 Nw 0273
    BATCo document for Legal Services: Health Canada 21 October 1999

    




    There is anotner interesting as;ect in this a;;roach.
    if one knows tte a=cunt of tar or nicotia whicn is -
    slide -10 necessary to g4-.e a defined difference in the strangtt
    rating itzhculd '.a possible to calculate consu-7er puff
    vol,=25 fr= Sensory result. To do this wo, would need
    a set of cigarettes with di!ferent utepness in their
    delivery c~urve&, w.th different ricotin *last-city.
    in s-arisinc. the above results one can state:
    Slide II
    2.4 Influence of filter ventilatiem en
    sensory test results
    Vact.,lation has a big effect on the-~4ole smoke
    ncm; maition. Product developers faced -wilth tbe,
    objective to introduce filter ventilatiorLinto ezisting
    brands found a*lutions to nearly maintain tUy taste
    character and t!ie acceptance of that brands.
    DUZing this d*V&10J;!"t3 they changed blends and flavearse.
    introduced expanded t~"ccc, changed the porosity of the
    cigarette paper and the-'pr*S3Ur* drop and efficiency of
    tte filter.
    One should think that we b&ve vaough know-how about the
    influence of filter ventilat4 an -~e- sensory test results.
    But there is sill some contr*v=sy'a;xpngxt our product
    developers what f ilter ventilation rez2,--y does on seuxorT
    judgement.
    Ln
    13 Nw 0274 :Z:)
    .-N
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    7
    In tryirg to ce: this ;nfor-zation we compared tne res*,;Its
    fr= all product tests of those products,
    Z which had
    introduced f1.1tez vent.;!az.4cn or clanged the degree o!
    filter veznt4 Ution.
    Slide 12 The result olf this c==;az_4scn shows a quite syste~,natic-
    pattera;
    in a !ew cases there was a reduction in fullness and
    bitter, nearly always a reduct-on in strength, shar;ness,
    scratc.%-*ness and her'-, an izc:ease in emptyness and sweet.
    By compar-ng these result* with the results about the
    Slide 11 influence of per-puff nicot;c on sensory judgement one
    can see, th at the effects of ba.th treatments an the taste
    are contrary.
    And indeed: increasing the per-puff oicatin was one of the
    development tools most of the comparriv 4" uxecl when they
    introduced filter ventilation.
    In the =oma nt , we do not have more precise r~;sults*abo,,t
    the influence al- filter ventilation, but we h4re Nst..
    finished a test wLth a serie of cigarettes to study
    this in more detail.-
    2.5 Influence of blAmd c~umcez on sensorv test results
    In a set of product tasta.witg4air comparison consumer
    tests three samples out of eigbV,-were slightly changed
    in blend rectipe. All samples conta6knad stems treated
    by our STS-expansion process, tba camppprison cigarettes
    contained WTS. The per-puff deliverieslietvaen the
    azpl*s and between samples and compa iSCIV-Ciartttes
    are different.
    . ..  X1.
    "N
    8MV
    0275
    Co
    01\
    U1
    1-\J
    _41.
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    - 8 -
    The data base is sti:,~ to sr-a-111 to =ake proper statist;,ca.~
    analyses but 1 believe that it is worthwhile to look at
    1he results.
    I have platted the difference in fullness rating between
    Samples ard cc=pa- . son cigarette versjs the di!!ezrerce
    Slide 14 in per pu!_4 t&T. As I mentioned ea:lier the fullness
    rating of tne smaple will not he effected by deliveries
    as long as tne per puff deliveries of the sam;le is
    higher than the per pc.!f derlivery of the comparison
    cigarette.
    By looking at this plzt those. 2amples with the blend
    change were rated higher in ftllness. The blend change
    eflected about 5 1 of the whole ]:,lend ard was an
    iller
    .aase in !1avour grades and a deczeas* in f.
    grades.
    At tus point one would say we should havd'.expocted
    such a result-.
    But let as have- a look an the next plot:
    Slide 15 The strength ratl~q of these three samples was un-
    expectably high. The-same happened with the other
    dimension which are h-,gh correlated with strength like
    the irritation ratings, sweet, herb and bitter.
    It is too early to spec%LLatep -what that would mean for
    product development, but the result Is interesting encug-
    to continue the work about the influence of small blend
    changes.
    BNW
    027O
    Cx.)
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    7 2. 6- Influence-cf an exar.=Ie cf usina the
    described metnod In =rcduct devel-c--ment
    We get the omjective tc develic; an 9 mg pr:,_4,_ct wit,,*
    ;n L-ran-4s w-th
    nearly t.-e same taste as cne of cur =a-
    1 14 mg ta r.-,
    After a few t----als with di!ferent blends we 4_eveioped
    a sa==le, which. reached the taz;et of givin; 8 mg
    whic.-L'was rated equal in strength by our internal
    panel but still quire diffazent from the target brand.
    S!~de 16 The analytical figures tmtween target brand and sample
    I - are of course different. 3d_I_ we tried to match the -
    per-pu!! nicutin figures forra higher puff volume. From
    our limited infor=aticn about zean puff volu=e of full
    ffAvour cigarette smokers we thoug~= 50 JLI/2 sec would
    be apod, guess.
    I Both ciii~rettes were tested by a panel~_-saf swkers of the
    target brand.
    S1.1do 17 The result of- the development sample was noc-,very satis-
    (2. 3. f ying. Ifte. oni~ success: The strength and mildness rating
    row only) was not significa5it di!ferent from the target brand. But
    acceptance was very-olOW, as Was the fullness. The two
    irritation ratizm3s werr-higher, the same with most of
    the other dimensions ancf- -itams.
    The multiple regression betwien acceptance and the
    12 taste dimensions showed a vary high A-weight for the
    fullness dimension.
    The next step was to try to in=o&xe tlv~ fullness of
    the sample cigarette. we developed a flzi~vur which by
    -our internal Judgement increased the tobacc:~,note.
    i3o 0277
    N)
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    SI__e 17 The tes-t result of that sample is now much c-*:sez tc tre
    target: tbe sample is judged significant lower in !oreign
    raw only) taste am-4 higher in the item herb.
    S~ ;,;e 17 But nc-a we- nave more questions than we expecteiE: how can
    a flIa-wc%:r shculd or.. . Ly . rcrease the
    row or reduce t!%e 4r.-itation ratings.
    3. outlook* !--rther pla:?Ming
    Durins a%,.: research we arim ered & few quest.:.-s but
    simultaneously asked new qua~ktioax_
    We are planning to continue with this type c! inves-tI-ga-
    ttons because we believe that ~~.results of this
    syst=Atic approach to explain sensory taft, :esult3
    br pr6A-;K-- properties can be of help-kr product
    -f,-
    We will conceAtrate our future efforts &106k;-tva lines:
    ù Puzther modiii-r-ation of the test procedure, the
    seq=entatic=--of Ike panel group, further izprovamont
    of the mathematicaD.Rodels and the experiZental test
    design.
    ù Development of sets-at cf4orettes designed to isolate
    the effects of the differ eat::kproduct chat;es.
    All
    V_
    I N
    8NN o ~Jl
    278 zz-~
    I-)
    Z~11
    BATCo document for Legal Services: Health Canada 21 October 1999

    




    &A-T HAIAMPBURG' S CON SUM ~i--_R
    TE'ST METHODS
    10
    New Products
    existinq_prod;jr_ts
    p a comparison monodic test
    > existinj~ vs. > i~ew product
    product vdration'
    > smoker of that't~qnd > .,Standard"-
    (or from the same test group
    segment)
    (smoker f rom
    all brands)
    ean scores
    Results-.~'m Resiultse.~-,. mean
    > dif ferences
    scorts
    Both tests Llse the some items and a
    7- point skate
    RESEARME & DENMOPTWUMVT AND 1984
    0279
    BATCo document for Legal Services : Health Canada 21 October 199

    




       >
       1
       0
       0
       0
       _% MER TEST
       0
       THE TWO WAYS OF TRANSLATING CONSU
       RESULTS INTO PRODUCT DEVELOPMENT
       --7774-
       which product/
       what does Consumer smoke properties
       0 thot meon in Tra Test Explaining -
       in caused -this
       0 expert smoker
       Im Result to-ste-proffle ?
       Q. language'.)
       Pcodi,
       0 Pling
       Cr
       Q co 11 -
       , F~onsumer
       .,,Aim: reach a
       TL
       's
       )um'j
       's
       (v . t9sle-profil-targel,
       C>
       co
       C) j[tj9SjE:AUCjj & DZVELormENT AND PRODUCTION 19SY4
       6 ?, 6
       qpoo~,

       


    L" ING CONSUMOR TEST
    A ,, S AT
    I I . -10  i
    RESULTS IINTO FRODUCT O'EiVELOPMEE-NT
    I L- -  1 -10
    Is there a relationship between smoke
    deliveries and sensory test resU-tts?
    2. Influence of per puff n1cotin on
    s=-nsory test results
    1r),luence of smoking 'b~haviour on
    sensory ~est results
    4. Influence at; filterventilation on
    sensory test resu its
    5. IN luence 6 f blend 'changes on sensory
    test results
    B. Influence of f lavouring: an'tex. ample of using
    the described method in Pr8buct develop[n*.
    7. Outlock: further Planning
    RESEARCH & DEVELOPba= A= DDUCTION 1984
    BNw 0281
    V~,
    (J-1
    11C
    BATCo document for Legal Services : Health Canada 21 October 1999

    




       w
       0
       0
       CL
       0
       C
       r
       Ca INFLUENCE OF PER-PUFF-DELIV.EEIFS ON ,FULLNESS"AND .STRENGTH
       Testmethod: Poir Cornparisbn,
       A Fullness Strength
       =r
       0
       0 , . , .. ) I
       0
       0 INN
       Cr
       0
       Of fqrence of per-puff -deliveries Difference of pqr puff deliverie
       a4ainst comparison CIgGrette against comparison cigar elle
       0 CO
       N)
       njhFjjF;,.AYftC,II & IDEVELOPMENT AND pnODUCTION 1944
       I ~ (1 1. 7' 9 L) 0 0 5

       


       ED
       >
       0
       0
       06
       0
       6- L. 6. .6- L_ 6. 1 U*J W~d 101"i tT4 j _j __j _j
       U.,
       oil
       U) INFLUENCE OF PER-PUff,, I)ELIVERIES ON  FULL
       (D
       ANq.-,QSTRENGTH"
       Z ~1 ETSJ
       (D
       Testmethod: monadir,'j,`
       CD .011: .,~' rP_MWM A_:~L
       fullness difference strepgth difference I
       Cigarette S!"'~er PN P I"
       to own brand tq. own brand  between snrnp
       and own bior
       0 . -31, - , _r
       sk) . . KRO14E -~., ' HO I . . -.0.42 - 0,46 -0.37 -0.0.
       CL
       2) All
       fig. KRONE + 0. ok.- 0.86 +0.37 0.0
       I Y"
       0 V
       I,. ~%_, .. P C. I -if 10: !A,
       Jet
       A
       0
       Cr
       M 7?=
       RUSBARCIff DEVULO]PNINNT AND flikoDUMOM 1984
       to 0( 1 1 .. Ir
       ?,(e-6~9pooS

       


       0
       0
       CL
       0
       c)
       (D
       If
       IS THERE A RELATIONSHIP BETWEEN SMOKE OELIVERIES,
       AND SENSORY TEST RESULTS ? I),,-
       432 Y-
       CD
       THERE IS A RELATION BETWEEN STRENGTH AND PER-POFF DELIVERIES WHICH EXPLAINS
       ABOUT 40 % OF THE TOTAL VARIANCE.
       (D
       (D , 141~*
       2. BY COMPARING SIMILAR PROPUICTS,111E RELATION BETWEEN STRENGTH AND PER-PUfF
       DELIVERIES IS GETTING STRON60.
       0
       i 06
       3. FULLNESS IS NOT IJONOTONOUS RELATED Willi DELIVERkfS.
       Livolks ARE RELATED W1711 LOWER FUkjff S5# 11IGHER DELIVERIES
       0 LOWER
       JJAVR NO INFLUENCE ON FULLNESS,RATINGS*
       0 . :11
       cr
       CD
       to
       co
       w RESSARCIS 46 J)ZVFj,.6PUj1NT ANI) PRODUCTION io84
       co IR
       __j ___j -i I _-j

       


       0
       0
       CL
       0
       CD
       r*
       INFLUENCE OF PER-PUFF NICOTIN ONI'SENSORY JUDGEMENT
       M (significant differences. only)
       Ch
       Test method: monodic
       0
       M
       Smoker KRONE -- Smoker
       (D
       X
       Nicolin per puff 0.08 0.065 0,065 0.05
       0 V~ -
       0) )It . C ! I - ,
       lb mildnes!i 4,6 3. 4.1 3.1
       CL 01
       strenglh 4.7 5.1 5.5
       $,ha~pnOss 5. 1 .6 5.7 5.1 5.5
       %
       0
       5.2 5.3
       Cr h b' 4.8 .9
       SKALING 1 -7 1 very much
       C) 7 not(11 1311
       (0 co
       RESEARCH & DEVELOV*P:NT AND PnODUCTION 1084
       17~6(,(~00

       


       0
       0
       CL
       0
       r+ INFLUENCE OF PER-PUFF NICOTIN -ON SENSORY TEST
       RESULTS
       1. Increasing per-puff nicotin.,will effect
       [>Strength( )and' mildness(~)
       SharpnOess
       =r > thi? -items. sweet(~). herb bilter(j)
       0
       2. lncroa~ino per-puff nicolin ~qems to be more effective thon
       0
       incro~sing per-puff Aar on'lhe strength rating
       0 a) .. *%. .
       ty . . . , .
       (Ml
       Co RESEARCH & DEVELOPMENT AND PRODUCTION JOH4
       - %~ kw~ %---W W~w 6-..W _-_j . __j ~j --- j I __j I __j I  I
       ~, 1, 0 ~- 9 o, 0 0 S

       


       0
       0
       CL
       0
       -h
       0
       INFLUENCE OF SMOKING BEHMIOUR ON
       SENSORY JUDG~MENT-
       (n . I
       (D
       Testmethod: monadic
       W
       .IPMWNF N E. -E
       Sample strength
       X PIAWNF N
       PN PN
       0
       CL
       13 2,3 0.40 j0-047 2.54
       0.43 0.049 1.65-
       H 1.77
       0
       cr
       SKALING 1- 7 1 = NOT AT ALL
       CD
       7 = VERY MUC14
       00
       -4 RESEARCH & DEVELOPMENT AND PRODUCTION 1984
       9( 0 0

       


       CL
       INFLUENCE OF SMOKING 13EHAVIOUR ON
       Nicot n pet SENSORY TEST R~GULTS
       puil
       13
       0
       0 0.10 - 1
       0.014 mg
       X
       CL
       0 puif
       Cr IL
       CD Volume ('Mi/si
       17.5 23 22 35
       CD
       RESCARCIE & DEVELOPMENT AND PRODUCTION
       j~ 0 os

       


       >
       i
       0
       0
       CL
       0
       INFLUENCE OF SMOKING BEHAVI R ON' SENSQRY TEST RESULTS
       cn I , I
       I, THE POSSIBILITY OF SMOKING BE"AVIOUR UNFLUENCE ON,TllE TEST iESULT MUST ALWAYS BE CONSIDERED.
       (D
       IT IS HESSECERY TO HAVE:
       CD $HOKE ANALYSIS RESULTS AT DIFF&RENT PUFF VOLUMGS
       ELASTICITY TERM AS CORRECTION FACTOR
       CL 2. IT'SEEMS POSSIBLE VAT MEAN PUFF VOLUMES OF LARGER POPULATION CAN BE OMEASURED" BY SENSORY
       TESTS OF q;$OAETT 4*SFECIALLY DEVELOPED-FOR THIP,~611PO80.*
       0
       cr
       EINI
       to C=~
       w
       w 00
       (0 RPSEARCH & DEVELOPMZNT AND PItODUCTION
       9~6~j9ROO~

       


       m
       >
       1
       0
       0
       CL
       0
       21 INFLUENCE OF FILTERVENTILATION QN SENSORY
       TEST RESULTS
       Test method: air offlparison
       Difference in + 17% + 6% + 6% + 7% + 7%+ 22%
       fillerventilation
       FUllness 4
       Emptyness
       Mildness
       Strength 4
       0 Sharpness 1 4 1 4 4 4
       Scratchiness
       Dry 1. 1
       . q.,)"
       Foreign TO
       Clean 31
       0
       Sweel t f I f f
       0 orh . I
       Cr m H I 1 4 11 1 1
       4
       ImESEARCH & DEVELOPPARINT AND PnODUCTION 1064

       


       W
       >
       1
       0
       0
       0
       0
       EFFECT -OF FILTERVENTILATION AND INCREASE IN'
       (D PER - PUFF. NICOTIN
       INCREASE IN INCREASE IN
       (D FiLTERVENTIL'ATION PE11 P0F1:,NIC0"TIN
       Fullness
       cc
       Emplyness f
       Mildness
       X Strength
       0 Sharpness
       t
       Scratchiness
       Dry
       Foreign facfti
       Clean
       0 Sweet
       0
       cr
       (D
       Bill:&
       RESEARCH &DEVELOPMENT AND PRODUCTION

       


       >
       I
       0
       0
       0
       CD
       A FULLNES INFLUENCE OF s-LEND CHANGES
       (D ! ji~  I
       CA . I i
       (D
       CD
       CD
       +o,6
       0
       0 e
       -0,3
       PNVIRF
       40,2 PN
       0 +0,05 +011
       Cr co -0,05
       CD . I I
       VQ
       IOP4
       IftESSAItC11 & DEVU, LOPMENT ANDpnoDlUCTION
       1 V617MMO~;

       


       0
       0
       CL
       0
       STRENGT11 ANFLUENCE QF BLEND CHANGES
       0
       CL
       -014
       ID
       0 PMWNF
       Cr -011 -0105 +0;05 +6,1 +O'f 2
       PN
       co PO
       co CD
       to w RESEARCI~ & DEVELOPMEIVIr AND ]PRODUCTION
       1OQ4
       "o o

       


       >
       0
       CL
       0
       r
       SAMPLE SAMPLE
       CO)
       (D 527 150.
       PNWNF .1.48
       0,95.
       PN
       N -
       A, 01106 0,077
       (D PN
       X VENTILATION 21#4 1,2
       lu FD 84 1 77
       TAR A:
       \WIC 14.0 lil. 12,4
       *)AT 50 ML PUFF VOLUME: 527 00,111
       0 co ISO - 0,15
       cr
       ItESEAncu & DEVELOPMENT AND PItODUCTION
       2~6999005

       


       w
       >
       8
       0
       CL
       0
       INFLUENCE OF FLAVOUt AG
       Test method: monodic
       (D
       SAMPLE NO. 728 150 527
       NUMOER OF SMOKE 96 129
       ell
       :r ACCEPTANCE 3.00 3.64 3.05
       0 FULLNESS 3.16 3.09
       to 1 71 5.76 . 5.42 5.97
       EMPTYNESS I
       so MILDNESS 11.15 4.18 4. 06 1
       CL
       STREN6TH 5.23 5.30
       5.44 5.27 5.58
       L SHARPNESS
       0 SCRATC"INIIS 5.811 5.32 5.92
       4 -5.40 5.63
       5.61
       DRY 4.99
       0 5.22
       Foartlah TASTE 5 511
       Cr %
       (D 3:26 3.74 3.18
       $WE 5.57 5.53 'S. 50
       (0 HY(Ja 4.92 4184. 5.3G
       (0
       (0 1 1 BITTER G. 08 5.71
       (0 ItESEAltell &DEVI-' .1.01'MUNT AUD Piftol)XICTION 11)84
       VP699qooS

       


    -30-
    I
    R & D/MARKETING METHODS
    9 MARK!TING R SEARCH/SURVEYTNG TECKNMUES
    R.P. FERRIS
    SUMMARY
    01JECTIVE:
    To-illustrate, in a general retrospecttvi, SOM C*Mwft Sources of Was.' error
    and confusion in consumer research and to Tadicato poisiblo future' directions
    In such research. with particular emphasis CrLbshavioval measurement.
    - - - - - - - - - - - - - - --- - - - -
    Marketing btginsvith a fact: Consumer buys prcWuct,'~md the Usence of
    successful martatinj is understanding el buys jW proiacft.
    Marketing has drawn on Psyebology and Statistics to help clarify this
    relationship. Part rsmtly-is has become relevant to consider whether
    Dthavioural research can comtrOute practically to this understanding.
    It Is argued that In the absence of c rite behavioural observations of how
    people smake and how they choose to movs,.around the market place, most
    consumer research tends to measure consumert-jtheories about consumer
    behaviour. Project sinos, indicating significanj differences between the way
    people smoke (hold the cigarette) and the wky p*ool* think they smake is cited
    as evidence.
    Preference testing is discussed to indicate the inadequac ~'*f current
    Statistical models in handling the complexity of humn choice-49ta. The
    collaborative 38"illbank preference testing study Is outliftd;',Wustrating
    loptimall pr0orence testing principals and the point is made that--lactimall
    13 NW
    0296
    Ln
    %M
    Ln
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    -31-
    methodology design can only be considered agains.; the particular demand
    characteristics of s;ecifIc tesc situations. Pro3ect Prefab is briefly
    outlined as a mejMdologiCal exercise designed to Indicate the additional
    information avallabU from a repeated preference test des%qn where across
    trial stability of preference 3odgement Is also accounted for. It is
    considered that. althoug&of.thearotical Interest. repeat preference desi;ts
    art in practice time/cost Iftaffectivt.
    The neglected importance of relating sample composition to test results is
    emphasised in discussions of 62 & DC product evaluation exercises where sj=l*
    vafftbles such as usual brand smoked. deliver7 level habituation, ar4 levels
    of concept awareness are shaim to relate signiftcauxtly to differential
    peneptims of product. Examples art discussid from KLTA exercise% comparing
    middle and -UW tar smokers; a pro3ect developInt-1vw carboa monoxide desips
    (LOCO).* sidestriam ;roduct testing; And NILE, a prodUct test version of DELTA
    from whict eviden6e. Is put forward suggesting that sm*u4 adopt particular
    smoking 'styles$ In =der to receive from the product the;--type of Sensory
    experience they wtsh. i;d,that vithin certain bounds. this'irehavinral style
    will transcend inter product differences.
    Other sources of confusloo, cWatentional product testing art also
    Illustrated. Contract r1lbuTh erAzining the experience of middle tar swearT
    'down-shiftins' to flavoured and wflavoured cigarettes (PROJECT BOOTH) draws
    attention to the relative strength of &~ittudinal. brand imagery. and concept
    awareness variables in influencing perception-of product Sensory attributes.
    Results of studies an human perception of sensarl attribute changes cantisqtnt
    upon usully encountered levels of ventilation UrMbility suggest that
    0 Inherent product variability m&I also contribute soulmes of error in tht - .
    consumer test process.
    A popular trend to elaborate segmentation techniques b"ad tio"lly on
    AtittUdinal and lifestyle Measures is indiCatsd using the Stanfor-k-Researc.,j
    Institute IYALS, Model as an exumple. It is considered that the awit of
    B "W 0297 (Z)
    ui
    0`1
    U-1
    %-G
    -c~….
    G\
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    -32-
    I
    such awoachvs shoulo properly be rneasurel against their "11.1111ty
    ac:;.rt:aly ;redi&. waritet dynamics, and tnat any useful degree of finesse .n
    I achieving this ts likely to reculr* in-Ic"a oevelopment rat:iar 1~,an
    off-tme-pej; acquisition. ay the same t.-ken, such aporoaches are -likely to :e
    highly culture spacitic.
    GR & DC pro~ect NARINERS ft, detaiI44 as In examPle Of the use Of social
    pressure, smoking nd health.wd price sensitivity variables to predict
    quit.tng and switching patternTAn a cobort of UK smokers. A plan to
    Narli-trs by inecrporatlon of a smaktng behaviour perspective is outlined. :n
    tb1s approach, consu*r segments ar; d4flned rAt just in attitudinal terms.
    but 161 tertas of regularities in behavio I style with which they choo-ge
    smite' cigarettes. The behavioural measureS~,.l Ita. the attitudinal ones
    precad i nqpthem, will be related Predictively Tritat movemants I.e.
    quittiog/switchinq patterns).
    finally. hrtwe trends In consumer research are antictotad:
    1. Trend to total' pvopositlon testing and more tim/c"* effective
    methods.
    2. Increasing sophi%ticati-owof lam:ple segmentation.
    3. Increasing sophisticatUe.of ttatistical models to dial with
    contribution of variables/ulientL
    .4. Solsiantation for prediction, as w@41 jz~tharactarjjatjon, throup
    correlation of behavioural to sub3activeftUltudinal masures.
    /A
    5. Increasing Whasis an understanding swtickingf-oaterns.
    The key Proposition of this paper is reiteratedt th&tCwsuW-.rt%e&rcj
    sh:uld continue to ensure, wherever possible, the mchomip of
    wteption data to behavioural fact.
    Crl
    C)
    QD
    co
    ON
    6 " 0298
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    -33-
    DISCUSSION (KIAutes taken 4y 6.A. Read)
    Ir general terms the presentation served to identify shcr-.falls In consurwr
    research.
    A primary diffic0ty in these studies stems from the fact that the consutoer
    perception of a product Is not equivalent with consumer behaviour. A specific
    exainple. that of perceived hole blocAting uts identified as being factually
    incorrect based an an observational and Uestionaire study.
    In terms of product testing a number of signtficant observations in relation
    to sx;arimental design have been identified; sTiAdtarAmm product testing,
    preferwic* ratirbq following attribute rating using-tompamtiyt techniques
    improve test vali0ty and reliability. It has been ihWn that attribute
    assessment carrel atIC -better under circumstances where v&-JQg JVdgmntS were
    required for produc assessment. Conventional product test* asmms
    consumer consistency, thiF Is not the cast and therefore would-_1odi caft the
    need for repeat testing.
    In terms of a consumers response *natural product variability. consuu;k~ cm
    discriminate between these prod=U af*,therefore within assessment may wwql
    assess alternative products differently.- Sam interesting observations weraCe
    made in Project LOCO where Icrd tar swkerr-swkinq a reduced pressure drag-IQ
    mg product assessed it as too strong and waiceptable whereas the 6 mg low
    pressure drop product was judged as accepWIt. 'Project Nile was outlined
    which was designed to validate the use, of the selta7fischniqug using consumers.
    An outline of the advantages of generating a sub-gF a W ysis within the
    franwavrk of the Delta assessment with Increasing panel size was presented.
    An examl* of this sub-group analysis indicated Product amssment could be
    significantly influenced by the duration of specific chm bran* smoking. .
    Increasing own brand smoking duration reducing the ability to discriminate
    c:J
    BNW 0299
    a,,
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    att rItutes . '-In am Iti on. an effect of s Pee if I c procuct smoKlm; was apparent
    In terms of a -scoxers atility to assess alternative products. Similar 'own
    branc, effects were seen in terms of sensory profiling during product
    assessments. This otmervation was attributed to a smoker having a particular
    behavioural style associated with his products.
    Other studies (Pr,-.,ect Booth) bave indicated that smokers art sensitive to
    extraneous factars, i.e. the efTect of product wrapping affected product
    acceptability and strength assessmapt. Other research has Indicated that
    consumers can be sensitised to a product compt as indicated by a low
    sidestream produc: study. In this study-it cftld be demonstr&ted that %&.are a
    product concept Is Introduced and ex;wsed U r the consumer group, if there As a
    shortfall between expectation based on the coanpi: ttm this can reinforce the
    rejection of tht product.
    Project Marinert. designed to follow the trends of brin~P-swibcbtzg and
    quitting behaviow In relation to smokers. non-smokers andFsx~-smiokers in
    relation to a rarle of factors including smoking and health iftitudes.
    motivational factors and social pressures was outlined as macharKsm for
    developing a predictive model of switching behaviour. In addition-this
    project will be used to relate thw-markitt segmentation to particular bimvlamr-
    styles.
    A future direction for consumer resemb ww outlined:
    - total proposition testing.
    - to make tests tin and cost effective.
    - Increase sophistication of statistical models to with concept
    variability.
    correlate behaviour to subjective and attitude measures.
    increase understanding of switching patterns.
    0300
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    -35-
    The disc-.ssions covered the following points:
    Concern was expressed regarding the,use of total product testing In terzs of
    the inability tir identify the Importance of specific elements. It WAS
    Suggested that s*,&ttStiCAl models might help In .Nis contax- an~ outlined the
    ntt-d for a hoilstic 6*U base.
    Suggestec that const-ober research capabilities had been underestimated In term-S
    of their ability to resolve product stand-off situations. It was indicated
    that It is essentially a probleff Of Co~ication - if a test is properly
    conductod this can 1.1 - resolved. InOrAted that correct test design is more
    a W o;riate than repeat testing.
    Clarific4tion was Muired to identify the petfticalar criteria for adaptiog
    rtsurch tachniques for brand comarisOns and prOduct tasting in relatioa to
    pre,dwt deve-Topmat. It was suggested that iotai-'ptoducirpropasition
    assessmeat did'Ret. replace conventional product tesi~nj,,
    So= points,of clariftution were expressed in terms of pr*Ouct acceetabillty
    assossunt withis a 1;eCTFic Delta panel test and suggested that bl*W factors
    nay have lrlay*d a mom ta~"nt role than the physical design charactwittics
    of the product.
    The question was posed. what Is wri-4mportant. providing the consumer wits a
    product that fulfills his perception i0-.kq provide a product that is factX4111
    supportable and atts the consuftr PWcaft1qn. It was suggested that playing
    to the consumer perception is a marketi% ;r0psition although cautioa was
    exercised In this regard.
    t
    8 NW
    0301
    k_7
    BATCo document for Legal Services : Health Canada 21 October 1999

    




       0
       0
       CL
       0 L_ L_ L ftw L_ t k W. W.- ~ _j __j - -1 JLJ __j _10 _j _.-i __j j
       c
       (D
       -h
       0
       WHICH CONSUMER BUYS -WHICH PRODUCT ?
       SOCIODEMOGRAPHICS -SENSORY  'TAR"
       EVALUATION
       CD PERSONALITY NICOTINE
       PREFERENCE
       X I .. IMAGE
       LIFE STYLE
       TESTING
       0 VALUES PRICE
       PRODUCT
       ATTITUDES, TESTING SENSORY
       CL
       MOTIV.TIJ 13RA , NO  ATTRIBUTES
       ~Ns
       M16NITORING ELASTICITY
       PRIQE,ptNSITIVITY
       '8EGMENTATION
       JM~qe RESPONSIVENESS 6
       0
       cr
       FENS
       ENSI
       V
       EV A L I
       M6KING CAREER
       SMOKING STYLE
       9~0100 C

       


       0
       0
       CL
       CONSUMER BUIY$ PRODUCT
       ?
       to 6 . , ) I  i I
       X . PSYCHOLOGISTS .1.1. 1 '' STATISTICIANS
       MARKETlNC;,a*,"m=wmw-.
       PANDORA'$ BLACK BOX
       06
       0
       0
       0 03 RESEARCHERS
       (D
       BEHAVIOURAL BOX
       CD
       6 9 F,,; 0 0 c

       


       >
       1
       0
       0
       CL
       0
       .j lu __j __Rj ___j -i --i
       0 _L
       PSYCHOLOGIST: "DESCRIPTIVELY C!IARACTERISING YOUR
       CONSUMER WIT".'THE BROADEST RANGE
       OF ATTITUDINAL, MOTIVATIONAL AND
       CA
       PERSONAPTY TYPE MEASURES,
       'AND.
       RELATING, THIS TO SUBJECTIVF,'EVALUATION
       AND.lDRAND IMAGE MEASURES IS WHAT
       .4
       CqUNTS.
       (D
       STATISTICIAN: .,,,""THE PRODA131LISTIQ RELATIONSHIP
       BETWEEN KNOW".'CONSUMER VARIABLES
       AND OBSERVEWMARKET DYNAMIC VARIABLES
       INCID ENCE, SWITCHING
       (i.e. QUITTIN&
       INCIDENCE~ ANP` DIRECTION) IS WHAT COUNTS."
       0
       r)
       0 E30 -XVIOURAL: "THE
       co RtLATIONSHIP BETWEEN HOW THE
       cr , i.
       ,~PESEARCHER CONSUMER SMOKES TO GET WHAT HE
       co CD WANTS FROM THE PhODUCTS, AND HOW
       to w
       i co
       4:16 THE PRODUCT IS DESIGNED TO ALLOW HIM
       ~S6(;qeoo~ TO ACHEIVE THAT IS WHAT COUNTS. "  r

       


       w
       > 0 .0
       i
       0
       0
       CL
       0
       0
       C
       .10
       .--PROJECT SINOS SAMPLE STRUCTURE
       a,
       BRISTOL SHEFFIELD TOTAL
       co . . . I
       SEX Male 21 15  36
       Female 24 53 77
       (D
       TOTAL , 45 68 113
       0
       CL . k , '
       AGE 16-34 29 51
       31 It 6
       L 35-54
       0 55 + .8 0 16
       0 ED
       cr . 1~ 9
       (D
       ABCI 22 31 53
       C:) C2DE 23 37 60
       to
       9 ROOS

       


       I Im 0 0
       0
       0
       CL
       ~JECT SINOS MAINFINDINGS
       TOTAL PPE-FS (EXC? LIGHTING PUFF)
       UNBLOCKED 80%
       CD
       PARTIaLT BLOCKED 13%
       INDETERHINATE 7 %.
       CD I I
       X
       :r NORTHELN."LOCATION
       0
       UNBLOCKED 81%
       CL PARTIALLY BLOCKEP sm 10%
       I NDETERMI NATE
       9%
       0
       ca SOUTH9M LOCATION
       cr . N, -
       UNBLOCKED 78%
       PARTIALLY BLOCKED 17%
       cl) Cno
       to 0) INDETERMINATE I= 5%
       CD
       9 s o o s-

       


       0
       0
       CL
       fAp
       T S
       PROJECT S11jos TERMINAL PUFF ANALYSs.
       ERMINAL
       0 TOTAL (T T2u) I"
       UNBLOCKED 647.
       PARTIAL14Y PJ.OCKED 26%
       INDETERMINATE 10%
       (D
       =r NORTHERN WCATION
       0
       P$BLOCKED 71%
       CL ,lARTIALLY BLOCKED
       0) 1: 22%
       INDETERMINATE 7%
       0 'OUTIJERN LOCATION
       0 -, It, N
       cr UNBLOCKED
       C) PARTIALLY BLOCKED '167.
       cc w m
       40 C) 9 INDETERMINATE 13%
       40 4
       I
       9 rd 0 0 g

       


       >
       1
       0
       0
       CL
       -jai
       45S
       -h
       0 1 "1411AT DO YOU THINK THE FUNCTION'OF
       THESE SMALL HOLES MIGHT BE'?"
       ca i. . " . I % -
       TO STOP GETTING TOO MUCH
       NICOTINE 28
       TO STOP GETTING TOO HqQH^TAR
       VENTILATION (TAg, M,141COTINE)
       it
       0) 1 REDUCE smoKE ; ~
       =r INCRF-ASE AIR
       9
       0
       INCREASE FILT TtON 7
       DON'T TAYJI'13A K SO MUCHAEDUCE STR NGTH
       3
       LET INPURITZES OUT 2
       HELl,Fp,..THK,%~61GARETTE BURN
       STOrg YOO SNOKING'
       0 MORE
       a .4%8S,0kNG9ROUS  -
       0 .::--',,ALT40ilS THE NICOTINE TO COk- THROUGH
       cr (1) . .41. *
       yELPS TO TELL YOU WHAT THE CIGARETTE IS I
       'IOU HUST NOT SHOKE PAST IT
       DONIT KNOW 9
       00
       2~4~j9ROOS

       


       Im
       >
       0
       0
       CL
       01 "DO. YOU THINK DURING NORMAL
       (D SMOKING'YOU WOULD BE LIKELY-
       TO BLOCk THE~ ')HOLES AT ALL
       in
       m 52%
       ...YES
       w NO 27%.,,
       L DON'T.POW  21%
       0
       Cr Vq (N = 97)
       MA
       9~6,1-qs)oog

       


       0
       0
       0
       0
       C tu __j .-a
       PO I
       (D Is*
       ABOUT WHAT PERCENTAGE OF THE TIME YOU SPEND PUF'FING
       ON THE CIGARETTE WOULD YQJJ'THINK SOME OF THOSE Jjbj.,FS
       cn COULD BE BLOCKED
       (D - 0, 25, 50, 75 or 100
       cc DON'T,KNOW - 12.5%
       w . 10 0%.: 1 1
       0 25 152
       0)
       CL 50 4 2. 5%
       75 0, 20% 72.5%
       100 loz
       0 Ct. I
       IN ft 4 0)
       cr (33
       to CD
       6,7(~Jqfjoosc

       


       I
       >
       I
       0
       0
       CL
       0
       Product testing methodolow
       0
       19
       r- Collaborative studly
       92
       CA Sample specificatiorr. 2,678 smokerp6 male/female, Marl boro/Winsion
       CD ~Imultaneous
       1. Product placement type
       Sequential monadic
       0
       Pre-attribute rating
       CL Methodologllc~! 2. Preference positlonl~q
       Im variables
       Post-altribute rating
       wa t1l. 1 1.
       0 Descriptive attributes
       0 3. Attribute question scaling Comp,,-xrative attributes
       CD
       CO
       w
       V
       096 ~, n,()O~

       


    P:.AC X-V~ 77?E * S=-f%--7ANE0US -V-
    i AnAlTtix Method L- "tem Preference)
    C ZT, =- ; - Frexerence ;  I S&=;.Ie
    Pref Pref P1.0, IT , V al, %40
    A  TC
    1 Sj=u-ItA=eouZ : Before De3;,-;--ve 46.7 45 2 343 0.30
    2 silzu-ltazeous Z. Before Cam rative 48.5 41:8 S 342  I.Z5
    49.Z
    sizuItiLzeaus After 37 9 9 325  Z.i 9
    1
    simultilleaus After Co=&rzt.'v* 1 48-2 40-5
    &VIAL StWIC-LZAnOus 48.1 41.11 10.5 1346  2.60
    5 Sequential : Bef are : Dist" 49.2 42.9 7.9 3 '79 1..!9
    6 Sequez--i'al. : Before : COwF;;a;9v*, /49.1 42.5 8.1 312 1.20
    7 S4RUCZt!Al After : Descrtptivs -0 .0 46.0 7.0 328 a to
    8 Sequential After * c4aparstive 49.7 43.7 7.6 343  0:96
    --TT 7.7
    Acm SEQU-.% AL ~4~- 3 43.3 1332 1 79
    MERAI -1 7=AL 9.1 2679 3 10
    value 1.65 a 90Z confidence
    ,1-96 - 952 can-oldence
    An&-',vgis wetbod Z-.- (excluding Non Preference)
    CZLL Pro ?Taf sa=;Ia Significance
    I Size 2
    I I  Qhi Value
    I siaultansous : lef Ora : Des 50.3 49.2 313 .05
    - 34f Ora : Cam  ve .
    2 Simultaneous 53 46 309  1.57
    r DesEte, 5  43:3
    simultalsous Afte .- 56:7 5 233 4.58
    After coupamlya '-- .54.4 45.6 298  Z.to
    TOTAL S=T.TARZOUS . ~Z-$ 46.2 1205  6.72
    3 Sequential Before : Descriptive 53-3 ": 46.5 303 1.32
    5 Sequential, Before : Cmarative 33.AL- lis. 6 305 1.31
    7 Sequential I After . Descriptive 50.S.. AgA3 305 .01
    8 Sequential * After : Comparative 52.7- . A7. 317  .31
    TOTAL SEQU=AL
    32.5 47.3 1221?~.  3.03
    OVIRALLTOTAL 53.1 46.9 Zw  9.49
    Ch-, value 2.71 90Z confidence
    3.84 95-^ confidence
    BNW 0312
    CN
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    11
    FIt-rF--RE%,CS OLTES-zION Pos--- -1
    BE70RE -y- AMR
    ~~is Method Nor. Preference)
    ?-eierence
    CELT- I - Ss--;ItsigtU,-Cancl
    Value
    e f I P
    A 31
    SL=4-1-.araou2-- : Before Descri;tIlve 46.7 45 - 2 8.,. 1 343 .20
    12 SizuItaneous'! Ilefore Co=-~aratjve 48.5 41.8 9.7 .342 I.Z5
    5 Sequential 1--Aef ara 1 49.-)4Z.9
    7-i 3:5 1 .'. 9
    6 Sequential : 'Wore CocParacive 49.1 42.8 8.1 332 I.ZO
    TOTAL IEFORZ 49.4 43.2 8.4 1346 2.00
    3 Si=ultaneous - After D`G7&cri7CIva 49.2 37.9 12, *9 325 2.19
    4 si=U11tanecum , After : cez;aZative 43.2 40.5 11 .3 336 1.50
    7 Sequent:ULL After : Descriptive 47.0 46.0 7 . 0 328 0.1.9
    8 Sequential After : c6c;pArAtIve 48.7 43.7 7.6 343 0.96
    TOTAL AFT-1.1 &8-3 42-0 9.7 1332 2.42
    OVERALL-TOTAL 1.1 2678 3.10
    'T' value -L.65 - 9OZ confldence
    I-U a 95Z confidence
    Analysis Metbod 2. (amn-Iuding Non Pref arenct)
    Preference
    N.- sampleSISMUIcanc
    CELL Praf A Pref Size 2
    Chi Value
    ,e~ -, I T I--
    1 SLISUltateous Were DascrI:ptIvi0- 2 315 .05
    2 simItanso" Uf or* camp 50 11 41&96:3 309 1.57
    arative 53:7
    5 Sequential Befora k Descriptive 53.3 46.5 303 1.32
    6 Soquantlal. Before . Comparative -`-,&3.4 46.6 305 1.31
    TOTAL UJOU P 52.IrI, 47.2 1232 3.36
    3 31waitsueous : After Descriptive 163 4.59
    its
    6
    4 SIMItanscus : Af tat cowpantiv* 541. 6 Z9 2.10
    7 Sequential : Af tat : Descriptive 50.S- 49.3, 305 .01
    8 Sequential : After : Gazparative 52.7 q.3 ;17
    TOTAL AF= 53.4 46. 61 1203i,- 1 5.59
    OVERALL TOTAL 53.1 46.9 2433 9.1.9
    i, I I I . -
    Chi value 2.71 a 90*. confidence
    3.84 - 95. confidence
    0313
    0\
    CIN
    r\-,
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    - US - .
    Aealvais Method I- (including Not- ?reference)
    Pre--creace
    Sa::; e I-. -,!Icance
    P-e 7' Va' ue
    ref I P Bf 1401
    A
    I si=ultaneaus Before Des 46.7 45 2 8 343 0.30
    3 S-Imu*:aneous A.Itar 49.2 37:9 12:9 342 z 19
    -.Scforu D zrivriv* 49.2 42.9 7.9 2:9 i.0
    Sequential -Idter D%scriptive 47.0 46.0 7.0 3,21.8 0.18
    D ES CRI?I74.VX 9.0 13:5 1.91.
    Q-0
    2 Simultaneous Before : Cospa 5 41's 9 7 342 1.25
    ra
    4 Simultaneous A 1-0 t a r : C rl= 'Ili 88 . 2 40.5 11:3 336 1 50
    6 Sequential Before : coupastive 49.1 42.8 9.1 332 1:20
    8 Sequential z Alter : Cowparative A&7 43.7 7.6 343 0.96
    TO-al CCMFARAT4,rVZ t.8-6 42.2 9.2 1.153 Z.48
    OV12at IXTIAL As J- 42.9 9.1 2678
    valU02, 1.65 - 90' confidence
    .1.96 - 95i Confidence
    Analysis xdftho~2~ Ok=ludinz Not Preference)
    Sample Su"ficames
    ?-.af A Prof a Size 2
    VALUG
    I Simultaneous : Before Dem=IP&I'Ve 50.3 49.2 315 .05
    3 Sizzatannous : Af ter Deemipctft 56.5 43.5 283 4.39
    5 Sequential Before D:SMITtive 53.5 46.5 303 1.32
    7 Sequantial After 0 a c3dptive 50.5 49.5 3C5 .01
    I- I
    "M DzSCtIPTT7z 47.2 1 1206 3.50
    2 Simultaneous Before Comparative S2.7 46.3 309 1.57
    :98
    4 Simultaneous After comparative 54.* :44.6 2 a 2.10
    6 Sequential Before Comparative M*e. 40-A16 305 1.31
    Sequencial After Comparative 52.7- , 47A 317 .81
    I - - ~^ r
    = AL CWARA%Tn 53.5 A" 6.02
    C ---U.LL TOTAL 53.1 46.9 M35 -j- 9.49
    I , 4 - - --- -- I I  T
    C--4 value 2.71
    901 conf idence.
    3.84 : 951 conlidance
    Nw VI
    0314
    ON
    ~-,j
    BATCo document for Legal Services : Health Canada 21 October 1999

    




       0
       0
       CL
       0
       r+ .40 - . . . I i
       ATTRIBUTE CORRELATION WITH PREFERENCE
       Correlation
       Coefficient
       Overall qualiq .74
       (D
       0 QualitY of TalPte 6 7 0%
       X Qualfty ~f Afttri'aste .56 *
       Natural/Artittetal Taste
       0 .54 *
       Amollat ;of Tjbte .39
       .36
       CL lfosatqi~ .) . .41 1
       )(in I i Y .35
       L4fricturing Quality
       R. * r ~ of burning
       w Effort
       r .23
       0 C) .24
       IV, N Strength
       .06
       0
       Cr
       (D M .6 \11 I
       High correlation wIth preferanq*
       1.B The size of the correlatloq coefficient Ls the Important factor,
       (D not Its sign  or
       t"96~' qpoog

       


       >
       f
       0
       0
       CL
       0
       . r _j
       (D
       TOM ATTRTBUTE DATA'
       C FA21PR bNkLYSIS SUMARY
       (D _,;Xactor I Factor 2 Factot 3
       Overall Quality .72 * - .05 .05
       (D 'Quality of Taste .85 * - -.04 - o6
       QuAlity of Aftertaste .73 * - .25 - .00
       "aturol/Artificial Uste. 7 .73 * - 07
       (D I
       4) 0 Amount of Taste .68 .46
       X Smoothness q, .44 .72 o4
       Moistness .58 . 4 - 17 .26 1
       0 1
       Msnufactu~ing Qu .09
       ,plity .57 .01
       Rate of 1prat to I
       ,pg .76 *
       CL Draw .09 62
       Strf"St "')".'.09 .87 k .01
       L %
       0 jj4fign Aggotinted for 14.7% 9.9%
       IN- C7 -
       0 , r
       0 60%
       Sub Tot4j lpf 3 factors
       Cr ED .1 V.
       ort-
       CD
       Primary contributor to factor In question
       Cn
       9~)00~1

       


       0
       0
       CL
       0
       (D
       Tho . se attributes whlc~ require the respondents to ma.ke' . some
       value Judgement have the highest correlation with preference,
       viz. overall quality, qua~jty of taste and aft~rtaste, natural
       taste
       Those attribute$ wrth require the respoqoents to make a
       0 m '~
       2
       definitive-judge. upon., the product c~aracteristtcs have the
       lowest cor1rala.k1on with preference vii. strength, draw effort,
       CL
       rate of bprn'.*hg. A
       0
       0
       0 k)
       CD
       CO
       cc
       CO CA3
       A.
       99659ROO_~

       


       0
       0
       CL
       0
       JLJ
       --h
       0
       r
       The study demonstrates that Sim4lianco"o Placement is a more
       sensitive method than Seqipentipl Mqnadic.
       CA
       2,, The stuay demonstrates., that asking the prefereqqe question
       (A
       Aftor the attribute.lVdations to likely to be-,a more sensitive
       method than 4xkjn4jt Before
       =r
       0 3. The study I fcates that although th~ri is little to choose
       .0
       between DeOcOptiye attribute scalps and Comparptive attribute
       CL sc&jes,.'-Tkt lattei seems to be m-b~rie asneitive.
       0
       4@ jh Is study plus other wo~T& undertaken on other 1)rodtict tests,
       0 strongly indicates a high degree of attribute redundancy.
       Cr
       ED
       C)
       CO
       Z9( *
       ) s- 9 e 0 0

       


       >
       1
       0
       0
       CL
       0
       Project PREFAIEk, Summar
       V of data
       IN W.
       CO Firpt Test
       refiK. "Total
       0
       preference
       CD
       7;;~ Prefer -123 i~
       =r 4) 1  252
       0
       0 13 1 .'1"" "1"
       No preferqobe 32 15 34 81
       0
       CL 0
       ID
       Prlpfe~-.13 100 26 112 238
       0
       0 . ! T tal 255 5~ .261 571
       P* 10 .. I
       Cr
       (D
       Preference changea: 56% R.eversals: 38%
       w
       40 C)
       A - B & H Special Filter (Export)
       CD -
       v
       9 so 0 0

       


       CL
       Analysis by Binomial Distribution
       19
       r- . 1~
       CD
       F1178
       i6e -4'tes't
       CD
       j~ii~r- Prefor -~r e -fe r
       .PtAndard Prefer St6ridard
       A error A B Ilerror
       Frequency ~61 25'2 .4"
       :r Percentage 49.4 50.6 2.20 Jc4 48.6 2.20
       0
       95% CIL Lovy!)V)'~ 45.1. 46.3 47.1 44.3
       Opper 53.7 54.9 55.7 52.9
       0.62
       0-41. Z -0.27
       0
       Cr
       M ai 0 .. t"I - -
       A = B & H Special Filter (Export)
       C)
       B = State Exprims 655 King Size
       C:)
       (j 9 (1 1! 9 "C" 0 0 S

       


       0
       0
       CL
       0
       (D
       Analysis by Trinomial Dis
       #ibution
       Z
       First Test Second Test
       Prefer No Prefer A- 13 Prefer No Prefer A - 13
       (D A prekgy B A prefqr B
       0 Frequency 265.,,),. '66 261 252 ~ 181 238
       CL Percenia9p.:,i )'-',~44.7 9.6 45.7 -1.0 , '44.1 14.2 41.7 2.4
       St"
       0
       qxrd.,f Mir
       0
       Cr jC.L. Lower
       Upper 6.7 10,0
       to C)
       46 A = B & H Spqclal Filter (Export)
       13 = Slate Expross 555 King Size

       


    Illust.ration of the significance of
    a.
    Preference Distribution
    Case I
    2.5-
    2.0-
    I Case IF
    0.5-
    L
    0 0.2 0.4 0.6. 1.0,
    p
    Strong preferers
    Ln
    0322
    01,
    ID\
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    ay-Preference Distribution;:
    !5~
    SE - 8 (20.16, 25-98).
    4i-
    3'-
    ~P)f
    B & H - B (5.54,6.94)'!
    . , :1 0 0.2. *0'4 - - -d.6 -V8 -. .1'.0:
    p
    b) CumUffiative Preference Disffibution:
    0.989  --0.999
    '0.80~1
    0.8-
    0.659 -""~B& H
    0.6-
    g(p) -
    OA,
    02-
    0.075-
    7 0.003.
    0. 0.2'; 0.4' p, 0.6 0.8. 1A.
    g(P) = Proportion of consumers with preference probabifit~ <, p
    B NW 0323
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    - 1%.
    Repeated Preference Test
    I
    Consumers in Test  571
    Guessers 95%
    Guessers stating 1:Weference for A 43%
    Guessers stating preferdnce for B  43%
    GuesseM stating no prefereme 9%
    TrUe prefere,~s of A 2. 7 0/c
    True preferersoPp 0-80/c
    True non-preferers
    1-61yc
    4X4
    :X
    A B & H Special Filter (E
    B State Express 555 IQng Size.
    . I
    BNw 0324
    BATCo, document for Legal Services : Health Canada 21 October 1999

    




       0 0
       >
       1
       0
       0
       CL
       0
       -h
       0
       r Project PREFAMe Conclusions
       CD
       U2
       V.
       --P~&&r Pr'e'fe*r* Ao,
       I Guessers
       ~..B & H SE 555 prefer9nce
       '6in e test (biniFmig!f~' 66% 50%
       X
       Single test mlal) 42.5% 42.5%' .15%
       6
       CL
       n4lp
       t6st 24i~ 0.8% 1.5% 95%
       0 M I
       0 Iv7 -1; ~- .6~; ~J
       0
       Cr
       C>
       CA

       


       w
       >
       0
       a
       0
       L~ 6~
       -3
       Previous usum Middle Tar
       r
       : ca. .
       0
       r Smell
       so maw eflon
       Rd Merit
       0 Allerlasle
       Taste
       IV
       CL
       10 Smoke, texture
       -& kiisi Ord . .) Mouthburn
       Thfo 11'6064
       a
       SlIk
       Throalburn
       .. wuly
       i~.
       cr m . 9, . .
       Vin
       Im
       i""c
       Ou"
       Kent. L
       -16
       co C>
       cz
       Q6S9&oog

       


    C cz. -
    cc
    I= 2 0~
    2.
    NO
    .2 1 Ld
    ENW 0327
    BATCo, document for Legal Services : Health Canada 21 October 1999

    




       0
       0
       CL
       JLLJ __j -im __j
       CD
       -h Usual brand smokers ea
       0
       CID
       cn
       CD
       Sinell
       CD John Pla moo-it
       ciriww"od, Taste
       R6
       0
       -4~
       0. ;~- Thioal calcWhil
       CL
       Smoke lexturo
       -A Lord.ExtFl I I Throalburn
       0 MOLIffibuin
       Kent G.L
       CD
       co
       r ( f) 0 o

       


       0
       0
       0
       Vanguard Smokers
       2
       r
       cn % ),
       0 Gin,011
       44
       % . Aftertasio
       (D o " wc;~e
       Tasio
       CD Met t'.
       91)
       X
       0 Vanuard I.AM
       Im
       0 Job" piq
       D) Smoko lexhire
       Urn assy Kent G. L
       IN) Throal catch/hil
       & Diroalbtim
       0
       0 Silk Cut
       0 a)
       Aoco
       RZ6S980o~

       


       M
       0
       0
       CL
       0
       .6- --j -J
       CD . U‹ - -
       uallIraƒdsm'olcorsO
       CD 1
       0
       . , . ) b .1~l,
       CD "!-d- D' - t M', , - -
       John r
       ne Mer II
       . 1. Tasiti
       =r Mrone
       0
       CL
       0 Ici .'. Motithburn
       16 Throalliurn
       0 Lam Throal calcIVIiIi
       Kent G.L
       bllk Cut
       Cr .. : .1 \.
       CD CD . . , . ,
       4
       Accepla gly

       


    ~E
    cn ce a 1.9 -. -
    1 CI
    J2
    E
    ig ! CI,. .
    0331
    CI.
    BATCo document for Legal Services: Health Canada 21 October 1999

    




    4Sz
    3
    '20
    LU
    a . - - :9
    5 93.
    BNW 0332
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    <
    M Sfroke t‰~t–re!
    0 M Mcuthbum 1 M;
    BON
    7=atbum
    M
    Cr
    0  7r...'Oat catchffift
    M  Taele
    CD zw -  =
    0 . CL -
    srnell -0
    Il . CD ~
    ait
    Ac-
    -apt…bility
    0
    ;r
    0333
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    . 11 - .. PA I
    Smckeiterure.
    CD
    D raw
    Mouthbu-n.
    NMI
    ~21 ~ MW
    I ;f - ,
    Throatbo
    M FWP
    Cn  Throa~ CaMM*
    co
    E. Taste  z
    ... .1 - . - - - rM
    rn  Smell
    3
    Cr Q I
    11) . /-
    Ca
    Attertarm
    CM
    a
    0
    Pr
    M ptnifity
    %  Acce
    Cn
    13 W' 0334 C:)
    co
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    Sr-,:.ke te_xturei
    C-:
    Draw eff ~rt
    M, C b U M Fri
    11 -- -- MI V* :
    0 C
    .9 . .~D wk
    Th,-:at catCtVWtF-
    Tazza
    Smell
    CL
    Aftertaste
    Acceptability'
    M 0335 CD
    0--
    -rh..
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    Smoke'texture
    9
    M. . Draw effort,
    Ir
    Mouthburn
    -0 .2
    Throathum
    . - 1,- 04
    Throat catc."t
    Taste
    5'
    Smell
    Attertaste
    Acceptability
    (JI
    B Oq36 CD
    CD
    Co
    BATCo document for Legal Services : Health Canada 21 October 1999

    




    Smoke. texture
    004
    Draw e
    ffortj
    0 Mouthburn , - . I (P I
    0
    Throatbum
    Throat. catdVhlt-- U)
    . if
    Taste
    Pr Smell
    Aftertaste
    Cn -- rt
    Acceptability
    I
    BNW 0337
    11.0-
    cz
    cI~-
    BATCo document for Legal Services : Health Canada 21 October 1999