SensoGram 2: When Counting Heads Is Not Enough Consumer research often relies on the counting of heads, or polling.  In a product test, for example, we might determine how many people prefer product A and how many prefer product B.  But this approach has shortcomings, as the following example will demonstrate. Suppose we have 100 consumers evaluate two products, A and B, and state which one they prefer.  In a second test, 100 consumers evaluate products C and D and likewise make a preference judgment.  The results come out as follows: - . Test 1 Test 2 . A            B C         D Number preferring 75          25 75       25 The natural reaction is to declare products A and C as winners, and B and D as losers.  Suppose now that, in addition to each paired-preference test, we also do a SensoMetrics evaluation in which 100 consumers score the products on a 0-100 scale of liking (0=horrible, 100=excellent).  Now we find the following: -   With a score of 70, A is a very good product; with scores of 55 apiece, B and C are both sound products; with a score of only 40, D is an inferior product.  So, the SensoMetrics testing tells us that B, the loser in the first paired-comparison test, is actually as good as C, the winner in the second paired-comparison test.   Why is this?  The apparent discrepancy occurs because, in a paired-preference test, the response to one product depends on the product with which it is compared.  In the paired testing, Product A won because it was better than B; product C won because it was better than D.  But this revealed nothing about the absolute acceptability of the products. The SensoMetrics scale, on the other hand, tells us about the absolute acceptability of the products.  The score for one product is not affected by other products in the test.  Obviously, if you need to compare products over time, it makes much more sense to measure them on an absolute, context-free scale. This is an important point:  the ultimate success of a product depends on its absolute appeal to consumers, not merely on how it stacks up against its competitor.  Far better to be the marketer of product A than product C, even though they did equally well in the paired-comparison testing! When it comes to measuring, monitoring and predicting product performance, counting heads is not enough.