Do you wonder which combination of MaxDiff items best reaches or motivates your respondents? Then, TURF optimization may be the right approach. TURF is a general optimization technique originally applied to other survey research data, but it also can be useful for MaxDiff studies.

TURF stands for "total unduplicated reach and frequency." It is an optimization approach for finding a subset (portfolio) of items that "reaches" the maximum number of respondents possible.

The classic TURF illustration involves which flavors of ice cream to stock at a grocery store. The grocer may decide that he/she has limited space and can only include up to 8 flavors of ice cream (out of 30 possible flavors). The grocer wants to maximize the chance that shoppers will find a flavor that they like enough to buy. The problem isn't as simple as including the portfolio including the 8 most preferred flavors on average across the sample. Niche flavors that appeal to segments of the population (and that can increase total reach) would be overlooked.

For the ice cream example above, TURF examines all possible portfolios (subsets) of 8 flavors of ice cream (out of 30 total flavors), and for each set counts how many respondents are "reached." The top few portfolios of 8 flavors that maximize "reach" are listed in the output with the percent of respondents reached shown next to each portfolio.

One challenge with TURF is that many solutions typically yield essentially equal reach. However, this could be viewed as an opportunity rather than a problem. You can bring other information to bear on the decision (such as expert opinion) to help decide which portfolio is best to solve the business problem. For example, if the grocer knew that one particular flavor (that appears in many of the top sets) tends to spoil more quickly than others, such solutions would be avoided in favor of other similar-reach solutions.

The "Frequency" part of TURF is how many times respondents are reached by the items in the portfolio. When two portfolios feature the same Reach, you probably should favor the portfolio that has the higher Frequency. (Frequency only applies to the First Choice and Threshold reach methods.) For example, consider 100 total respondents in your data set. If all 100 respondents are each reached two times by items within the portfolio, the Reach is 100% and the Frequency is 200.

## Items per Portfolio

Specify how many items should be included in each portfolio (the optimal subset of items). For example, if you are trying to figure out which 8 out of 30 ice cream flavors would best reach customers, specify 8 as the portfolio size. You might try different portfolio sizes to see how much additional reach is provided by a 4-item portfolio over a 3-item portfolio, etc.

## Portfolios

Specify how many of the top portfolios to report in the output. For example, if you only want to view the top 12 portfolios found, then specify 12.

## Filters

Filters may be used for excluding certain items or combinations of items from searched portfolios, or for excluding certain respondents from the analysis.

Items: to exclude specific items from searched portfolios, indicate those exclusions here

Respondent Groups: to exclude respondent groups from the analysis, indicate those here

Prohibitions: to exclude multiple items from occurring together within a searched portfolio (e.g. Vanilla may not occur with Chocolate in an optimal portfolio), specify two or more items here

## Reach Method

The MaxDiff Analyzer provides three different options for assessing "reach" in TURF:

Weighted by Probability (Standard MaxDiff Scores): This approach assumes that reach isn't an "all or nothing" proposition for a respondent: respondents can be fractionally reached. It uses the logit equation to compute the likelihood that the respondent is reached. More details.

First Choice: A respondent is counted as "reached" if the portfolio contains his/her top item (the item with the highest utility score). This option also reports a "Frequency." The "Frequency" is the number of top items in the portfolio, counted across people (for MaxDiff data, the frequency only differs from Reach in the exceptionally rare case that a respondent has two items exactly tied as the top utility item). Note: if you are excluding items from the analysis, then first choice considers whether the portfolio includes the respondent’s highest utility score item among only those items included in the analysis.

Threshold: A respondent is counted as "reached" if the portfolio contains an item with a larger Probability of Choice than specified in the Threshold field. If your MaxDiff questionnaire displayed 4 items per set, then the average item in your MaxDiff study would have a 25% Probability of Choice. So, specifying a reach threshold of 80% would mean that a respondent is reached if the portfolio includes an item that has a 80% Probability of Choice (when placed in a set involving three other items of average preference). Type 80 to mean 80% (not 0.80). The "Frequency" is the number of items exceeding the threshold within the portfolio, counted across people.

## Recommendations

Different reach methods (Weighted by Probability, First Choice, Threshold) often will lead to different optimal portfolio results. For MaxDiff data, we tend to prefer the Weighted by Probability approach, as it leverages relatively more information from what is often sparse data. And, small changes to the data (such as adding a few respondents) have less impact on the rank-order of the optimal portfolio outcomes compared to the First Choice and Threshold approaches.

The First Choice approach makes less efficient use of the granular preference information provided by MaxDiff and is the most "steppy" of the reach methods, where small changes to the data have relatively more impact on the rank-order of the outcomes.

The Threshold approach represents a middling ground between First Choice and Weighted by Probability, but the challenge for MaxDiff data is what probability threshold is the best value to use.