The Horizontal Segmentation Principle
There is no perfect product—only perfect products for different people
The Horizontal Segmentation Principle overturns the assumption that there is one perfect product for everyone. Howard Moskowitz's breakthrough insight was that when you look at messy consumer data, you should not search for the single most popular answer—you should look for clusters of preference. There is no perfect Pepsi; there are perfect Pepsis.
Moskowitz proved this with Prego spaghetti sauce. He created 45 variations, tested them on thousands of people, and found Americans clustered into three groups: those who like plain, those who like spicy, and those who like extra chunky. The third finding was revolutionary—one-third of Americans craved extra-chunky sauce, yet no one was making it. Prego launched an extra-chunky line and made $600 million over the next decade.
The deeper lesson is philosophical: searching for universals—one perfect product, one correct answer—does people a massive disservice. The same coffee rated 60 when one-size-fits-all becomes 78 when matched to preference clusters. The difference between 60 and 78 is the difference between coffee that makes you wince and coffee that makes you deliriously happy. Embracing variability is the path to true happiness.
- There is no perfect product—there are only perfect products for different clusters of people
- The mind knows not what the tongue wants—people cannot always articulate their true preferences
- Horizontal segmentation replaces hierarchical thinking about quality
- Pursuing universals does people a massive disservice
- In embracing the diversity of human beings, we find a surer way to true happiness
- Stop Looking for the Single Best AnswerAbandon the assumption that there is one optimal version of your product. When Moskowitz tested 45 spaghetti sauces, he did not look for the most popular variety—he looked for clusters. If your consumer data is messy and does not produce a clean bell curve, that is not an error—it is evidence that you are dealing with multiple distinct preference groups who need different solutions.Pro tipMoskowitz's line 'the mind knows not what the tongue wants' captures the key insight: people cannot reliably tell you what they want in focus groups. You must test and observe.WarningDo not interpret messy data as bad data. The mess IS the insight—it reveals multiple clusters hiding inside an average.
- Identify Natural Preference ClustersTest a wide range of product variations and analyze the data for clusters rather than peaks. Moskowitz varied spaghetti sauce across every conceivable dimension—sweetness, garlic, tartness, visible solids—then looked for groups who consistently preferred similar profiles. The clusters emerge from the data, not from assumptions about what segments should exist.Pro tipFocus especially on clusters that are large but unserved. One-third of Americans wanted extra-chunky sauce and nobody was making it. The biggest opportunity is the large invisible cluster.WarningDo not force segmentation where natural clusters do not exist. The clusters must emerge from real preference data, not from marketing theory.
- Create Distinct Products for Each ClusterOnce clusters are identified, develop distinct products optimized for each group rather than one compromise product for the average. When coffee is matched to individual cluster preferences, satisfaction scores jump from 60 to 78—the difference between wincing and delight. The cost of additional product lines is almost always justified by the satisfaction increase.Pro tipMoskowitz's success with Prego led the entire industry to segment: 7 vinegars became common, 14 mustards, 71 olive oils. Even Ragu eventually hired Moskowitz and now offers 36 varieties in 6 categories.WarningSegmentation creates operational complexity. Ensure the satisfaction gains justify the added complexity of manufacturing, distributing, and marketing multiple variants.
Campbell's Prego was losing to Ragu despite having better tomato paste and spice mix. Moskowitz created 45 variations, tested across multiple cities, and discovered three clusters: plain, spicy, and extra chunky. No one was making extra-chunky sauce despite one-third of Americans craving it. Prego launched the line immediately.
When asked what coffee they like, everyone says 'dark, rich, hearty roast.' But only 25-27 percent actually prefer it. Most people like milky, weak coffee but will never admit it. This proves people cannot reliably articulate their own preferences—observation and testing reveal truths that self-report conceals.
Howard Moskowitz, a psychophysicist who graduated from Harvard, was hired by Pepsi to find the optimal sweetness for Diet Pepsi. He tested every level from 8 to 12 percent, plotted the results, and got a mess—no clean bell curve. Rather than picking the middle like everyone else, he was bedeviled for years by the question: why does the data not make sense? The answer came like a bolt of lightning in a diner: they were looking for the perfect Pepsi when they should have been looking for the perfect Pepsis. Nobody would listen. He traveled the country giving talks and was dismissed as crazy. His vindication came through Vlasic Pickles (creating 'zesty') and then Campbell's Prego, where the extra-chunky discovery changed the entire food industry.