Q-LSMs - SA Marketing Magazine

Report
Q-LSMs
A PROUDLY South African Solution
to an uniquely
South African Market Research sin
Developed by:
Millward Brown:
Erik du Plessis
Kamendra
TNS:
Neil Higgs
A Definition of Market Research
• “Purveyors of Consumers’ Valuable Time to Marketers”
• We Persuade Well Selected Respondents to give some of their time to
answer questions,
• Which we analyse, interpret, present to marketers
• I.e. We add value to the time we got from consumers
IT IS A RESEARCH SIN
TO WASTE THIS VALUABLE TIME OF THE RESPONDENTS
AMPS is proudly unique to
South Africa
• Nowhere in world does marketers and media
owners combine to have an industry
respected single source database,
• Unilever (Eddie Schulze) and SAARF developed
LSM’s as a shorthand to differentiate between
the haves and have-nots based on their living
standard
• Prof. Jackie Galpin did the statistical analysis
• Has become a South African industry norm
BUT:
This can take 10-20 minutes.
Sufficiency versus Redundancy
• Determining the LSM algorithm
– When Prof Galpin does the analysis the data has been collected.
There is no need to use as few as possible variables in the
classification algorithm,
– Whether the algorithm uses 28 variables, or all the variables in
the AMPS survey does not matter,
– Much rather err on the side of redundancy.
• Using LSM inside the AMPS database
– When using LSM for analyses inside the AMPS database the LSM
is coded per respondent.
– It does not matter how many variables are used in the algorithm
• Using LSMs to collect primary data from respondents
– It really matters how much time you are wasting
– SUFFICIENCY is a real priority
This
‘wasting my time’
is coming back to bite us
Respondents do not
feel it is impolite
to end a cell phone or on-line interview
that wastes time.
JUST CLICK THE END BUTTON!
LSM 1
LSM 2
LSM 3
LSM 4
LSM 5
LSM 6
LSM 7
LSM 8
LSM 9
LSM 10
Refused
R 50000 +
R 40000 - R 49999
R 30000 - R 39999
R 25000 - R 29999
R 20000 - R 24999
R 16000 - R 19999
R 14000 - R 15999
R 12000 - R 13999
R 11000 - R 11999
R 10000 - R 10999
R 9000 - R 9999
R 8000 - R 8999
R 7000 - R 7999
R 6000 - R 6999
R 5000 - R 5999
R 4000 - R 4999
R 3000 - R 3999
R 2500 - R 2999
R 2000 - R 2499
R 1600 - R 1999
R 1400 - R 1599
R 1200 - R 1399
No Income
R 1 - R 499
R 500 - R 599
R 600 - R 699
R 700 - R 799
R 800 - R 899
R 900 - R 999
R 1000 - R 1099
R 1100 - R 1199
Is Personal Income a good surrogate?
Total SA - Income versus LSM
3 000 000
2 500 000
2 000 000
1 500 000
1 000 000
500 000
0
Is Personal Income a good surrogate?
Hardly
• 11 million out of 37 million said ‘No Income’,
but we know they live and consume stuff and
have things (Living Standard),
• 5th largest income category, nearly 1m.,
refused to answer the question
• Not much correlation with LSM
LSM 1
LSM 2
LSM 3
LSM 4
LSM 5
LSM 6
LSM 7
LSM 8
LSM 9
LSM 10
Refused
R 50000 +
R 40000 - R 49999
R 30000 - R 39999
R 25000 - R 29999
R 20000 - R 24999
R 16000 - R 19999
R 14000 - R 15999
R 12000 - R 13999
R 11000 - R 11999
R 10000 - R 10999
R 9000 - R 9999
R 8000 - R 8999
R 7000 - R 7999
R 6000 - R 6999
R 5000 - R 5999
R 4000 - R 4999
R 3000 - R 3999
R 2500 - R 2999
R 2000 - R 2499
R 1600 - R 1999
R 1400 - R 1599
R 1200 - R 1399
R 1100 - R 1199
R 1000 - R 1099
No Income
R 1 - R 499
R 500 - R 599
R 600 - R 699
R 700 - R 799
R 800 - R 899
R 900 - R 999
Is it better when we look at
Metropolitan?
Metropolitan Income versus LSM
1,400,000
1,200,000
1,000,000
800,000
600,000
400,000
200,000
0
Is it better when we look at
Metropolitan?
• 4.4m. out of 15m. has no income
– LSM 6 : 1.4m.
– LSM 7 : 871 000
– LSM 10: 407 000 out of 1.6m!
• 544 000 refused to answer the question:
– Mostly LSM 9&10,
Where would you make the income cut-off for the different LSMs?
900,000
800,000
600,000
LSM 1
LSM 2
500,000
LSM 3
LSM 4
400,000
LSM 5
LSM 6
LSM 7
300,000
LSM 8
LSM 9
200,000
LSM 10
100,000
50 000+
30 000-49999
20 000-29999
14 000-19 999
10 000-13 999
6000-9999
4000-5999
2000-3999
1 000- 1 999
600-999
0
1-599
MONTHLY PERSONAL INCOME
700,000
Introducing Q-LSMs
Toward Better Market Research practices
via good science
Research Question:
I want to split my sample into:
LSM 8-10,
LSM 5-7
LSM 1-4
OR:
I want to interview LSM 7-10
How do I not waste 20 minutes of interviewing
time just to do this?
23,9 m is the number of people in non-rural South Africa,
Solve this problem for a non-rural sample
Research Question:
I want to split my sample into:
LSM 8-10,
LSM 5-7
LSM 1-4
OR:
I want to interview LSM 7-10
How do I not waste 20 minutes of interviewing
time just to do this?
23,9 m is the number of people in non-rural South Africa,
Solve this problem for a non-rural sample
Splitting Sample into 3 Supergroups
(Non-Rural Sample)
C: 1,3m
B:14.1m
A: 8.5m
(sulsmr1101) Tap water in house/on plot
2. COMMUNITY SIZE
(sulsmr1102) Hot running water from a geyser
(commun1 ) (250 000+) - Part of Large Urban
(sulsmr1103) Flush toilet in/outside house
(commun2) (40 000 - 249 999) - Part of Large Urban
(sulsmr1104) Built-in kitchen sink
(commun3) (500 - 39 999) - Part of Small Urban+Rural
(sulsmr1105) Home telephone (excl. cellphone)
(commun4) (Less than 500/Rural) - Part of Small Urban+Rural
(sulsmr1106) 3 or more cellphones in household
(sulsmr1107) 2 cellphones in household
(sulsmr1108) 0 or 1 radio set in household
(sulsmr1109) Swimming pool
(sulsmr1110) TV set
(sulsmr1111) M-Net/DStv subscription
(sulsmr1112) Air conditioner
(commun5) (40 000+) - Large Urban
(commun6) (Less than 40 000 + Non-Urban) - Small Urban+Rural
3. METRO AREA
(armetro01) Cape Town
(armetro02) Cape Town Fringe Area
(armetro03) Port Elizabeth/Uitenhage
(sulsmr1113) DVD/Blu Ray player
(armetro04) East London
(sulsmr1114) Home theatre system
(armetro05) Durban
(sulsmr1115) Computer - Desktop/Laptop
(armetro06) Bloemfontein
(sulsmr1116) Refrigerator of combined fridge/freezer
(armetro07) Greater Johannesburg (Alex, JHB, Sandton, Soweto, Randburg)
(sulsmr1117) Deep freezer - free standing
(armetro08) Reef (Urban Gauteng, excl. JHB, PTA, Vaal)
(sulsmr1118) Electric Stove
(armetro09) Pretoria
(sulsmr1119) Microwave oven
(armetro10) Kimberley
(sulsmr1120) Vacuum cleaner/floor polisher
(armetro11) PMB
(sulsmr1121) Washing machine
(armetro12) Soweto
(sulsmr1122) Tumble dryer
(armetro13) Vaal (De Deur, Meyerton, Sasolburg, Vanderbijlpark, Vereeniging)
(sulsmr1123) Dishwashing machine
(sulsmr1124) Motor vehicle in household
(sulsmr1125) No domestic workers or household helpers in household
(sulsmr1126) Home security service
(sulsmr1127) House/cluster house/town house
(sulsmr1128) Metropolitan dweller
(sulsmr1129) Rural rest (excl. W. Cape & Gauteng rural)
(armetro14) eMalahleni / Witbank
(armetro15) East Rand
(armetro16) West Rand
4. PROVINCE
(prov) Province
5. GENDER
(sex) Gender
Splitting Sample into 3 Supergroups
(Non-Rural Sample)
C: 1,3m
B:14.1m
A: 8.5m
Desktop/laptop?
C: 1,3
B: 12.8
A: 2,2
N
Y
C: 0
B: 1,3
A: 6.3
Splitting Sample into 3 Supergroups
(Non-Rural Sample)
C: 1,3m
B:14.1m
A: 8.5m
Desktop/laptop?
C: 1,3
B: 12.8
A: 2,2
N
C: 0
B: 1,3
A: 6.3
Y
Motor Veh.
C: 0
B: ,9
A: ,6
N
Y
Vacuum
5-7
C: 0
B:,8
A: ,3
Geyser
Y
N
8-10
C: 0
B: 0
A: .3
C: 0
B: ,4
A: 5.7
Y
N
8-10
5-7
C: 0
B: ,2
A: 5.6
C: 0
B:,2
A: ,1
Splitting Sample into 3 Supergroups
(Non-Rural Sample)
C: 1,3m
B:14.1m
A: 8.5m
Desktop/laptop?
C: 1,3
B: 12.8
A: 2,2
N
Y
Motor Veh.
TV Set?
Y
C: ,7
B:12,4
A: 2,2
C: ,7
B: ,3
A: 0
N
Vacuum
N
5-7
C: ,7
B: 11,5
A: ,8
C: 0
B: ,9
A: ,6
Y
Y
8-10
5-7
C: 0
B: ,2
A: 0
N
Y
Vacuum
Stove
C: 0
B: ,9
A: 1,3
C: 0
B: 1,3
A: 6.3
N
1-4
C: ,6
B: ,1
A: 0
5-7
C: 0
B:,8
A: ,3
Geyser
Y
N
8-10
C: 0
B: 0
A: .3
C: 0
B: ,4
A: 5.7
Y
N
8-10
5-7
C: 0
B: ,2
A: 5.6
C: 0
B:,2
A: ,1
Accuracy of the rule:
• 23.9 million people in the non-rural universe.
• 8.5 m. (36%) are LSM 7-10, this decision rule will:
– Correctly identify 7.4 million (87%) of these as LSM 810,
– Classify .78 m. LSMs 5-7 incorrectly into this category
(i.e. 9.6% of the sample)
– Clasify ZERO LSM 1-4 as LSM 7-10!
• 14m. are LSM 5-7, this decision rule will:
– Correctly identify 13.2m. (94%) of these as LSM 5-7,
– Classify 1.1m. LSM 8-10 incorrectly in this category
(7% of sample)
Can the accuracy be improved?
YES,
Just ask a few more questions!!
(I.e. Run the analysis for another layer)
Why don’t we show this?
Limitation of a PowerPoint slide!
Splitting LSM 9-10 from Rest
(Non-Rural Sample)
Desktop/laptop?
N
Y
Vacuum
Domestic
Worker
N
N
Y
Rest
Rest
Motor
Veh.
Domestic
Vacuum
N
Y
Y
9-10
N
Rest
Y
9-10
N
Rest
Y
9-10
Finding LSM C versus Not
Audience
4,500,000
4,000,000
3,500,000
LSM 1
3,000,000
LSM 2
LSM 3
2,500,000
LSM 4
LSM 5
2,000,000
LSM 6
LSM 7
1,500,000
LSM 8
LSM 9
1,000,000
LSM 10
500,000
0
Metropolitan Areas (250 000 +) Large Cities & Large Towns (40 000 - 249 Small Towns & Villages (500 - 39 999) Settlements + Non-Urban (Less than
Urban
999) Large Urban
Small Urban
500) Rural
Isolating LSM C (or Not)
The Q-LSM ‘law of diminishing returns’
• Example: if, to identify LSM 7-10, we just
interview everyone at random, 36% will e
correct,
• Every ‘decision rule’ will increase this
percentage correct,
• With our three question decision rule we get
87% correct,
• To get 100% correct we need to ask all 28 LSM
question and do the weighting,
Does it Matter?
• LSMs is a continuum!
– Misclassification happens close to the borders of
the categories.
• The whole SA population is continuously
migrating up the LSM scales!
My Suggestions:
Timing:
• Implement every time a new study is done,
• Especially when migrating from F2F
interviews.
Plans:
• Neil Higgs:
– SAMRA
– PAMRO
– ESOMAR
• Africa
• Other Emerging
• Unilever (Source: Neil)
– Own LSM questions: tried once in two countries
– Big Segmentation
– Interest in Rural increased
Q-LSMs
A PROUDLY South African Solution
to an uniquely
South African Market Research sin
Developed by:
Millward Brown:
Erik du Plessis
Kamendra
TNS:
Neil Higgs

similar documents