TURF Breathing new life into an old technique

1
TURFBreathing new life into an old technique

• Ray Poynter
• Director, Virtual Surveys

2
A typical research problem

• Gelati Sons make ice cream
• In a typical store they sell 8 flavours and they
  have lots of data about how well they sell
• They have a new contract to supply a national
  supermarket
• But they are only allowed to offer 4 flavours
• Which flavours?
• The simple answer?
• The best selling 4
• The research answer
• TURF Total Unduplicated Reach and Frequency

3
TURF a bit of background

• Dates back to the late 80s
• Many research companies offer it in their
  toolkit
• Only a handful of papers over the last 20 years
• Rarely used these days
• BUT
• With a dusting of Internet-based data collection
• And exposure to Excel-based modelling
• A powerful tool for portfolio management

4
Why TURF?

• Consider the matrix below, with 3 flavours
• The data shows whether a flavour is bought by
  each respondent

Coffee
Banana
Almond
1
1
1
R1
1
0
0
R5
2
3
5
Gelati Sons
Almond Banana Coffee Damson Elder Fig Grape Hazel
R1 1 1 0 0 0 0 0 0
R2 0 1 1 1 0 0 1 0
R3 0 0 1 0 0 0 0 0
               
Rn-1 1 1 0 1 1 0 0 1
Rn 0 1 0 0 0 1 0 0
There are 70 different ways to choose 4 flavours
from these 8, which 4 maximise the reach?
6
Solver

• Excel Add-in
• Check you have the Solver Add-In enabled
• Choose a cell to maximise
• The Reach value in our case
• Create constraints
• Each flavour is either in or out (integer values
  in the range 0 to 1)
• The number of flavours must equal the number
  requested
• Solver will then search for the best solution

7
Solver example 1
Number of flavours Number of flavours 4
Reach Reach 95
  1 1 1 0 0 0 1 0 4
  Almond Banana Coffee Damson Elder Fig Grape Hazel Reached
R1 1 1 0 0 1 0 1 1 1
R2 0 1 0 0 0 0 0 0 1

R20 0 1 0 0 1 0 1 0 1
Total 5 13 3 2 4 3 8 6 19
8
Different scenarios
Flavours Unduplicated Reach Flavours
1 65 Banana
2 80 Almond Banana
3 90 Elder, Almond Banana
4 95 Grape, Elder, Almond Banana
5 100 Hazel, Grape, Elder, Almond Banana
Sub-samples can easily be set up Either as
sample selections Or, as separate Excel pages,
one per key sub-sample
9
Simple to Collect

• Each respondent sees all the scenarios, randomised

If definitely or probably buy
10
Frequency, thats why its not TUR

• Only people who are going to buy the product have
  a frequency greater than 0
• Definitely buys have a frequency
• Probably buys have a frequency only if you are
  counting probably buy as people who are buying
• Frequencies need converting to a common base
• In our example we might use the values as
  purchases per year
• Frequencies may need re-scaling
• Ideally using calibration data or norms
• Rough rule of thumb
• Square root of definite buy frequencies
• Cube root of probably buy frequencies

11
Choice and Frequency

• The questions were monadic
• So, what do we do if we have a respondent who
  says
• If Almond is offered I will buy 4 per year
• If Banana is offered I will buy 12 per year
• If we offer him Almond and Banana?
• If the products are comparable?
• As in this example
• Usually safe to assume he/she will buy 12
  products
• Some unknown mixture of Almond and Banana
• If necessary, keep the ratios, e.g. Almond 3,
  Banana 9
• If the products are not substitutable?
• e.g. some last longer, or are twice as big
• Then more complex assumptions have to be used

12
Simple example, re-visited

• Almond has more people who would buy, but they
  would buy less
• Almond Coffee meets everyones needs, but with
  the lowest frequency
• Banana Coffee has the highest predicted
  frequency

13
Solving for Frequency
of flavours of flavours 4
Avg Frequency Avg Frequency 7.1
  1 1 0 1 0 0 1 0 4
  Alm-ond Ban-ana Cof-fee Dam-son Elder Fig Grape Hazel Frequ-ency
R1 8 8 0 0 4 0 2 2 8
R2 0 3 0 0 0 0 0 0 3

R20 0 8 0 0 2 0 10 0 10
Total 27 87 12 17 15 16 50 18 142
The system can be set up to report reach as well
as frequency, along with sub-groups etc.
14
Frequency solutions
Flavours Average Frequency Flavours
1 4.4 Banana
2 5.8 Grape Banana
3 6.6 Damson, Grape Banana
4 7.1 Almond, Damson, Grape Banana
5 7.4 Elder, Almond, Damson, Grape Banana
15
Improving the interface

• By using customised VBA and Solver a more
  complete solution includes
• Selection of sub-groups
• Dynamically switching between Definite Buys only
  and Definite plus Probably Buy
• Stepwise solution of 1 to N products, reporting
  reach, frequency, cumulative reach and cumulative
  frequency
• Dynamically switching between Reach and Frequency
• Ability to temporarily exclude products
• Ability to force specific products
• Ability to weight key sub-groups, e.g. to make it
  much more likely that longstanding customers will
  have a product they definitely like

16
The client experience

• Whilst traditional TURF approaches provide useful
  insight, it has often been static and dull
• What-if modelling allows the client to really
  understand the dynamics
• Extensions include
• Adding Value weights to the products
• Forcing specific items to be selected
• Asking for the next best solution
• Identifying the disenfranchised
• Modifying the rules so a solution that finds 2
  products for each respondent
• Multiple ranges, e.g. in chilled food the best
  ranges of Indian, Chinese, Mexican, and Italian

17
Definites versus Probables

• Should the analysis be based on Probably Buy or
  on both Probably and Definitely Buy?
• Cases vary but
• Which option is closest to sales data?
• Try it both ways, see what the difference is
• If you are getting enough definites use these
• If you are using frequency then either use only
  definites or down weight the probably frequencies

18
Key TURF Questions

• Why isnt TURF used more?
• Perhaps because it is a specific tool for a
  specific problem and is not readily converted
  into a general tool
• How might technology impact TURF?
• HB might remove the need for each respondent to
  evaluate all the scenarios
• When is TURF applicable?
• Flavours
• Products in a vending machine
• Travel and ticket options
• Pack and size variants (with care)
• Courses (including conferences)
• Menus and bundles

19
Thank you

• Questions?