About TURF Analysis
TURF stands for Total Unduplicated Reach and Frequency and is a powerful statistical tool developed to minimize repetition between respondent variable preferences. It was initially used for analyzing advertising channels but now has a variety of applications in market research often dealing with determining optimal product offering combinations.
Here are a few scenarios where TURF would be applicable: a store has to choose the top 10 products to offer from 30 possible choices, a company wants to add or remove products from its product line and is seeking information on relationships between existing/potential products, an individual wants to advertise in several magazines and wants to minimize duplicated readership, and other possibilities where repetition between variables may be affecting the data.
To demonstrate an example, consider the following table where respondents are asked if they would buy the following flavors from the local supermarket on a scale from 1 to 5, with a 5 being definitely will buy, and that the supermarket can only offer two flavors. Which flavors should they offer?
| Respondent |
Vanilla | Chocolate | Strawberry | Mint Chip | Rocky Road | Orange Sherbet |
|---|---|---|---|---|---|---|
| Bob |
3 |
5 |
4 |
1 |
2 |
3 |
| Tina |
5 |
5 |
5 |
2 |
4 |
4 |
| Mary |
5 |
4 |
4 |
3 |
3 |
4 |
| Mike |
2 |
3 |
3 |
5 |
3 |
2 |
| Total |
15 |
17 |
16 |
11 |
12 |
13 |
Based strictly to column totals, it seems like the store should probably offer chocolate and strawberry. However, if we assume that individuals that rate the flavors with a 4 or 5 (top two box), will be likely to buy the flavor, then an interesting situation arises. Consider Mike, who would not buy any ice cream if chocolate and strawberry were the only two flavors offered. If the store instead offered chocolate and mint chip, they would still satisfy the other three respondents because they would be satisfied with chocolate, but Mike would be satisfied as well and the store would have four satisfied customers instead of three that would occur in the other situation.
Another way of thinking about TURF is through the following graphical representation:
The Venn diagram illustrates the potential usefulness of TURF because some of the products that might not have been considered based only to the reach size have less overlap and could instead capture additional consumers. Arguably, one would want to consider both results, not just one or the other, in deciding the potentially best marketing mix.
TURF is a way of calculating these other potential combinations in an effort to eliminate repetition between respondents and maximize reach and frequency, and can be a great way to supplement an analysis that only takes the totals into account.
While these are basic examples, large scale analyses can become extremely complex. Part of the complexity of TURF is the large amount of combinations that can arise. For example, if a company has instead 30 ice-cream flavors, and wants to find the top 10 flavors, then there are slightly over 30 million different possible combinations of flavors to be considered (30 choose 10). So while the analysis itself is relatively straightforward, the large scales associated with TURF make doing it impossible without an efficient program.
Example Analysis
The following is a simulated analysis that is more like an actual analysis than the information discussed above and will be quite useful for any individual seriously considering TURF. Please know that since it is an example analysis, the data is simulated dummy data.
Download the dataset used in the analysis:
wrightw TURF Dataset, size: 29kb
Download the completed analysis file:
wrightw TURF Report, size: 423kb
The background: The client has 20 ice cream flavors (because Ice Cream makes such great examples), and has limited shelf space for only 7 flavors. Consequently, the client would like to determine the optimal flavor combinations using Total Unduplicated Reach and Frequency Analysis. The twenty flavors are Vanilla, Chocolate, Strawberry, Rocky Road, Mint Chip, Cookies & Cream, Orange Sherbet, Cookie Dough, Coffee, Reeses, Rocky Road, Fudge, Butterfinger, Rainbow Sherbet, Lime Sherbet, Cheesecake, Snickers, French Vanilla, Strawberry Banana, and Mango. The client has monthly purchase data on ice cream servings purchased by flavor, and would like results by both reach and volume.
In addition to determining the top 7 flavor combinations, the client would also like the following conditions used in the analysis:
- The classic flavors of chocolate, vanilla, and strawberry must all be chosen
- At least one of the sherbet types must be chosen
- At least one of the candy flavors must be chosen
I should also make note that this analysis was done using my own personal program, and other TURF analysis program outputs look significantly different (such as only including the top one combination instead of the top thousand and not including the similar choice combinations which I do with the 6, and 8 choices). Furthermore, many other providers and programs are unable to calculate the analysis with the variable forcing restrictions described above, so it is important to check with your provider to ensure that they meet your specific requirements.
My Expertise
I personally designed and wrote a program to calculate TURF analyses and in the past I use it to run complete custom jobs.
Some of my unique capabilities include:
- Detailed data on the top 1000 results
- Standard results by reach, total volume, and max volume
- Complete control of changing calculation settings and criteria for custom results
- Variable forcing and subgroup analysis customization options
- Two other personally unique analyses included with each TURF report: the variable relation matrix and variable depth calculations (seen above in the sample report in the in Example Analysis section)
Recently, graduate school has been taking my personal and professional time and I am not as flexible as I have been in the past for professional projects. I am certainly still interested in discussing potential projects, so please send me an email if there is anything I can help with.
Frequently Asked Questions
I am trying to figure out a TURF analysis equation or a TURF analysis function and was hoping you could help me.
While I did design my own program from scratch to run the calculations, the way I went about it is complex and difficult to describe. I had to use a lot of programming and quite a few equations to get at the answer. If anyone knows of a simple function or a set of array formulas to easily solve a TURF, I would really like to hear about it.
How can I contact you?
Please email me at will@wrightw.com. I try to respond to all emails within one business day.
I have some additional questions about a TURF calculation or how it works.
I am happy to discuss TURF with anyone even if you do not need to have one run, so feel free to contact me.