The Pseudo-Science of How Brands Grow - Part 2

I pointed out How Brands Grow never proved its "laws" in Part 1 because Byron Sharp failed to use any time-series data (as growth is a time function), relied on flawed omnibus data, and reduced brand growth to simply self-reported omnibus market share. In Part 2, I deconstruct the remainder of the chapters to point out the flaws in his data and reasoning.

Chapter 5 

Byron proclaims: "Common marketing ideology states that differentiated brands should sell to different groups of people, or should sell for different occassions or users.". However, his "vastly more comprehensive studies" showed that competing brands sell to the same sort of people.

Face smack. That's because people consume multiple brands. No brand ever sold to a single segment.  In reducing brand to merely product, Byron fails to recognize the symbolic aspect of brand. This symbolic aspect is why brands try to stand for something; not simply to attract certain types of customers (e.g. Gain appeals to ethnic consumers), but embed heuristic (e.g. Apple products are well-designed) and even aspirational (e.g. driving a Mercedes means I have made it) value in the consideration process.

To support his claim, Byron cites that there are no difference in profiles of regular and diet soft-drink users because the market is roughly 2/3rds regular soda and 1/3 diet soda market – which is same for both genders. He provides table 5.9 where he "deliberately chose this style of analysis to show even tiny potential differences. But as you can see there isn't much to show."

But there actually is a difference when one compares between 34 year-olds (74% regular soda, 24% diet soda) and 55-75 year-olds (59% regular soda, 38% diet soda).  Like his predecessor Andrew Ehrenberg, Byron Sharp's modus operandi is to show only carefully selected data that supports his polemic.

Chapter 6 

Byron introduces the Duplication of Purchase law where all brands share customer base with other brands in line with size of other brands (the canny will recognize it as the Double Jeopardy law from Chapter 2 restated). He argues it would not hold if several brands successfully targeted exclusive customer bases or particular types of people. Furthermore, he claims "the story of customer overlap is at odds with the picture provided by perceptual maps" (see example of a perceptual map below).

Although there many different types of perceptual maps (MDS, PREFMAP, CA etc.), they are all derived from aggregate consumer perceptions of brands. A perceptual map has no relation to purchase behaviour or market share or consumer segments. Only a fool would believe perceptual maps "implies there are very significant market partitions" or there is "more market segmentation than really exists".

Feeling confident in pooh-poohing something he obviously knows nothing about, Byron then presents a table of purchase behavior, and notes that "Take Care" and "Feel Good" products share customers to a far higher degree than the Duplication Law would predict because they are marketed as new healthy brands for women - which essentially nullifies his Chapter 5 assertion there is no truth in "differentiated brands should sell to different groups of people".

Chapter 7 

In which the train starts to go off the rails when Byron cites examples of respondents knowing the name brand (Coke, McDonalds) being more likely to say it tastes better in taste tests. People are predisposed to their favourite brand.

Then the train goes completely off the rails when he contradicts his previous chapters (loyalty doesn't exist, stupid to market to loyal customers etc.) with "We observe this sort of loyalty in all categories … buyers restrict their purchases to a personal repertoire of brands. People buy far fewer different brands than they could." 

Byron presents Figure 7.1 that shows number of TV channels people have access to and how many they actually watch. Because the line flattens about a dozen or so each week regardless of whether they have 40, 80 or 200 channels, he argues this is "obvious loyalty". Not quite, finite viewing hours limit the number of channels one can consume.

In an attempt to get the train back on the rails, Byron gives the example of Harley Davidson ("poster child for emotion-based brand loyalty") where "stereotypically fanatical" Harley bikers represent less than 10% of owners, and only 3.5% of sales revenue. The fact that the largest segment of owners (40%) don't fit that profile (i.e. motorcycle parked most of time, like wearing helmet, don't know many other people that wear motorcycles) is not a case against loyalty (once again, Byron conflates purchase frequency with loyalty) but a case for the symbolic power of the Harley Davidson brand that that 40% want to be a part of.

Chapter 8 

Byron pooh-poohs the use of multivariate methods in brand perceptual data (probably due to his ignorance as evidenced in Chapter 6), accusing it of heightening small differences, arguing that the raw data looks similar for most brands.

It is precisely issues with descriptive data (he himself even mentioned sampling issues, and that people tend to score some attributes and brands higher than others) that people employ multivariate methods. To use an analogy: an amateur geologist will look at a field and see a uniform expanse of dirt, a professional geologist will use techniques to reveal subtleties underlying the surface that could point to veins of gold ore.

He makes yet another fallacy (of extremes), that there aren't any exotic and unique associations, which would be the case if brands are truly differentiated. Who gave him his PhD? He would have failed my undergrad class.

Byron makes another fallacious claim the NBD-Dirichlet model (which is the model his institute shills) as "successfully predicting a wide variety of brand performance metrics", therefore the real world conforms to this model. Model input assumptions aside (e.g. assume brands compete undifferentiated, frequent & loyal buyers underpredicted), the data the NBD-Dirichlet model is built upon and typically used to validate is self-reported omnibus survey data - Garbage In Garbage Out.

He claims "Apple's level of perceived differentiation is higher than other computer brands but it is not high". Again, he arbitrarily sets the threshold (because less than a quarter perceive Apple as different or unique) and presents Table 8.4 that showed the users of Apple who rate it as Different (15% vs. 9% average for the category), Unique (25% vs. 10% average for the category), and Either Different Or Unique (23% vs. 14% average for the category).

Check your work Byron, your data is wrong! An OR set is constrained by the larger of its constituents. So Apple's perceived differentiation is at least 25% if not higher (and much higher than the average 14%). 

Final Wipe

In Part 1, I mentioned how Byron not only misuses tautology, but is guilty of it himself. Tautology is an argument that claims to be true by repeating the same concept or assertion (in different phrasing or terminology). The laws Byron proposes: double jeopardy, pareto 60, buyer moderation, natural monopoly, user base seldom varies, duplication of purchase, usage drives attitude, attitudes & beliefs reflect behavioral loyalty, law of prototypically - are essentially tautologies. They are not so much laws of marketing, but founded upon the same systematic bias of brand attribution in omnibus studies.

The paradox is by reducing growth to (omnibus) market share, brands will not grow dramatically (nor necessarily be profitable) if they take Byron Sharp's advice, even even under the Dirichlet model, because the customer base is finite and limited by purchase cycles.

If you are one of those who have bought into this book, remember these 5 syllables the next time you are inclined to believe an international bestseller can't be wrong ... Fifty Shades of Grey.