Rank priorities with MaxDiff and optimise reach with TURF

Rating scales have a weakness: when asked to rate twelve features from 1 to 5, many respondents rate almost everything 4 or 5. MaxDiff forces trade-offs and produces a clear ranking. TURF then answers a different but related question: which combination of products appeals to the most people?

MaxDiff (best-worst scaling)

Respondents see a series of small sets, typically four or five items each, and pick the most and least important item in each set. With 12 items, nine sets of four give each item three appearances per respondent. Because they must choose, respondents reveal relative preferences, and the results discriminate much better than ratings. MaxDiff is also free of scale-use bias: some people rate everything high and others everything low, but everyone chooses one best and one worst.

Reading the MaxDiff results

For 120 respondents choosing among 12 banking features, the multinomial logit model estimates a utility for each item and converts it into a share of preference: the probability that the item would be chosen as best from the full list. Strong security comes first (18.1%), followed by low monthly fee (14.7%), mobile app quality (13.3%) and no overdraft charges (12.2%); ethical investments is last (2.1%). Shares are ratio-scaled, so you can say that strong security is about 8.6 times as preferred as ethical investments, something rating scales cannot support.

The simple best-minus-worst counts (how often an item was chosen best minus worst, divided by how often it was shown) give nearly the same order, which confirms the result. The individual-level scores show heterogeneity: strong security is ranked first by 21% of respondents, low monthly fee by 14.5%. Use these scores for segmentation or to compare groups.

Designing a MaxDiff study

• Between 8 and 30 items, short and unambiguous, at the same level of abstraction.
• Four or five items per set; each item shown three to five times per respondent.
• A balanced design in which each item appears equally often and each pair appears together about equally often.
• Ask about importance, appeal or likelihood to purchase, and keep the question the same in every set.

TURF analysis

MaxDiff ranks items for the average customer. But when you launch a product line, you want to cover as many customers as possible, not to offer five variations of the favourite. TURF (total unduplicated reach and frequency) finds the combination of items that reaches the largest share of respondents at least once.

In the example, 300 respondents rated ten ice cream flavours, and a respondent counts as “reached” by a flavour they rate 4 or 5. Chocolate alone reaches 57.3%. The best line-up of three is Chocolate + Mango Sorbet + Coffee, reaching 88.3%. Simply picking the three most popular flavours (Chocolate, Vanilla, Cookie Dough) would reach only 80.7%, because the same people like all three. Mango sorbet is less popular overall, but it reaches people the others miss.

Reach and frequency

Reach is the share of respondents who like at least one item in the set; frequency is the average number of liked items per respondent (1.40 here). When two combinations have similar reach, prefer the one with higher frequency: more customers have more than one reason to buy. The incremental reach ladder shows diminishing returns: the fourth flavour adds 5.3 points, the fifth 3.3 and the sixth 1.7. If each extra flavour costs money to produce and stock, stop where the increment no longer justifies the cost.

Combining MaxDiff and TURF

TURF needs a yes/no definition of reach. With ratings, choose a threshold (top two boxes). With MaxDiff, use the individual utilities: an item reaches a respondent if it is among his or her top items, or above a utility threshold. The combination gives you a ranking of what matters (MaxDiff) and a portfolio that covers the market (TURF).

Reporting

A MaxDiff study (N = 120, 12 items) showed that strong security had the highest share of preference (18.1%), followed by low monthly fees (14.7%) and mobile app quality (13.3%). TURF analysis of ten flavours (N = 300) identified chocolate, mango sorbet and coffee as the optimal three-item line-up, reaching 88.3% of respondents with an average frequency of 1.40.