Automated Maximum Difference Scaling (MaxDiff)

---

MaxDiff stands for Maximum Difference Scaling. It is a sophisticated research methodology used to determine the true preference or importance of a list of items (such as product features, brand claims, or ad headlines) by asking respondents to make trade-offs.

MaxDiff identifies consumer preferences by asking respondents to choose the best and worst options from a set of product features, advertising claims, or brand attributes, resulting in a ranked list of priorities — forcing tradeoff decisions rather than allowing respondents to select "all that apply" (often resulting in the "everything is important" fallacy). 

MaxDiff can be used for a variety of new product development and marketing efforts,  such as uncovering which product features drive purchase decisions and which advertising slogans resonate most with consumers. 

Brands can also leverage a MaxDiff to determine which offerings to include in a loyalty program, which features to keep when redesigning an existing product, or which brand imagery is most compelling to include in a future campaign. 

By forcing respondents to choose their absolute favorites, you gain a clear, prioritized roadmap for your creative and product strategy. 

MaxDiff measures each attribute on the same scale for direct comparison, while Conjoint analysis uses an additive model that compares the utility of attributes within a set rather than directly between attributes.

MaxDiff finds the rank order and relative preference of individual items.

TURF (Total Unduplicated Reach & Frequency) finds the optimal mix of items to reach the most unique people.

Think of it this way: If you're an ice cream shop, MaxDiff tells you that Vanilla and Chocolate are the two most popular flavors. However, if you only have space for two tubs, a TURF Analysismight reveal that you should actually carry Vanilla and Mint Chip — because the people who like Chocolate already like Vanilla, but the Mint Chip fans won't buy either.

The main difference lies in how the data is processed and how much detail you get about individual respondents. While count-based is a simple "tally," HB (Hierarchical Bayes) is a sophisticated statistical "prediction."

Count based MaxDiff analysis measures preferences through simple subtraction: (times chosen Best) - (times chosen Worst). 

MaxDiff HB analysis uses advanced regression, with the group's data "informing" individual scores (provides a unique utility score for every individual person.)