Thirteen: Insights and Customer Data
Companies have sought insight from the customers to make more informed decisions, but the practices in doing so have been deplorable. Some companies force the data to fit their preconceptions, others surrender themselves entirely to it without applying much thought or logic to what it means, many proceed with boldness and confidence on incomplete, skewed, or otherwise misleading data. And the results have been tragic.
Torture the Numbers
A saying among statisticians is that if you torture the numbers, you can get them to confess to anything you want to hear. This is very often the practice of firms who use methods that are biased to collect data that conforms to their opinion, subject well-collected data to biased methods of analysis to the same end, or simply ignore any evidence contrary to their preconceptions.
A particularly glaring example is in the way that many firms approach the female buyer: a group of male executives have highly skewed and stereotypical perspectives on what they think that women think of their brand and product, and are receptive to data that supports that preconception and hostile to any suggestion to the contrary.
The natural consequence is that the executives feel comfortable at being assured they are right - but when their marketing plan is put into action, it not only fails to attract female customers, it actually offends them.
(EN: The author doesn't suggest how to avoid this, and it can be difficult to talk an executive into setting his ego aside and consider without prejudice what the data is clearly indicating. It's largely a matter of culture, and the best advice is likely to get out of any organization where rank trumps reality, because it's being steered for the reef.)
Data Worshipping
At the opposite extreme are firms that cast aside all judgment and follow the numbers wherever they lead, and seem incapable of making decisions based on common sense, without numbers to guide them.
This is particularly problematic in the computer age, in which firms amass huge amounts of data and rely on statistical analysis to identify correlations - and then obey the correlations in defiance of all common sense.
Another issue is in placing the burden of data on the customer. The author takes as example a software company he consulted with, who had a very long and complex application for clients to purchase their products. Very few prospects completed the form, and most did not purchase. On closer inspection, it was found that very little of this data had any relevance to the customer's needs.
A final problem with data-worship is that not all data is good. A customer responding to a survey is responding to a survey, speculating about what he might desire at a future date, unable to accurately articulate what really motivates him, and even hiding highly-motivational factors of which is he embarrassed. Which is to say that there is a difference between giving a customer what they ask for and giving them what they really want or need.
(EN: A list bit that the author misses is that data is known and historical - it tells the story of the past and assumes the future will follow. There is no historical fact to indicate how a customer will react to some change that impacts the future - it requires speculation.)
Superficial Research
Another common mistake occurs when decision-makers don't really care about data, but feel they need to do some research to justify their decisions. They tend to do this on the cheap, send out some hapless intern with a clipboard to waylay people on the street to get their opinions.
(EN: I've often seen this in secondary research as well - when someone made a decision and was looking for support, they use Google to find a study that seems to agree with them, and often do not read deeply enough to recognize that it's a bad fit. I've even seen people, and often even present studies that have been debunked. Consider the number of people who subscribe to "subliminal advertising" and quote the "scientific study" done in move theaters - even though James Vicary, who had no credentials as a scientist, admitted it was a complete hoax.)
Superficial research is worse than none at all - it gives support to decisions and the confidence to take bold action on what are essentially "ghost stories" about consumer behavior.
Research is Not a Crystal Ball
The future is unknown. All research determines, at best, what is true in the recent past, and the "most horrific blunders" often occur when there is a demand to know what people will do in the future, particularly in reaction to something that does not exist in the present.
Asking a subject what they might do in future is inherently inviting them to speculate, confabulate, and make up stories about a hypothetical situation. Even the most sincere respondent responding to the most unbiased survey instrument doesn't really know what he is actually going to do in a real situation
Even when research is not intentionally skewed, it can be unintentionally skewed. There isa difference in the responses you will get if you ask a person "what do you think" about a product and "what do you like" about it.
Surveys also make people think about things that they don't normally think about. Ask a subject why they choose a given brand of mustard and they will give an elaborate and reasonable response. In the wild, they grab what's cheapest, of one whose label is impressive, or whatever is at eye-level, or whatnot.
The act of study also takes the consumer out of the context of a decision. Consider taste-tests of products, which rests on the assumption that customers will buy a product that tastes better. But unless samples are given away in stores, the consumer won't know what a new product tastes like until they have already bought it - though it could be said that their evaluation of taste might lead them to repurchase more often.
The author mentions the "New Coke" fiasco in that regard. Allegedly, the company taste-tested the new formula against the old and customer in the lab situation said they preferred it - and found it revolting when it was placed in the stores (EN: There's great speculation that the "study" was a diversion when the firm intended to switch formulas, reintroducing "classic" coke with corn syrup rather than sugar - but even if that is so, it's often observed that Pepsi's sip-tests often find it to be preferred, but there have not been many converts.)
In one sense, asking the customers for their opinion seems a better course that relying on insider opinions - and it may be a more democratic method of deciding among options, but it is not necessarily accurate. Often, customers do not know how they will react and are speculating.
On the other hand, surveys are very useful in discovering the known, and investigating the past experience of customers. That is, few customers would tell you that a clean bathroom ranks highly among their criteria for deciding where to dine, but many customers who have been to a restaurant will complain about a dirty bathroom - and it may be the reason they do not return.
Disappointing Demographics
Marketers used to delight in demographic information because it is objective and measurable. Unfortunately, it really doesn't provide much in the way of useful information.
As an illustration, the authors suggest a demographic profile of a "college educated single female between the ages of 33 and 36 who makes between $65-75K and lives in midtown Manhattan. This seems very specific - but it does not speak at all to the needs of the customer, just their superficial descriptions. They ten create hypothetical profiles of four women who all meet the same criteria, but who have entirely different goals and need different information to evaluate purchasing options.
Ultimately, demographics are a more sophisticated way of stereotyping customers - assuming that all people of a certain race, gender, age, income bracket, and so on are all exactly alike in their needs and interests, or similar enough that they can all be treated just the same. Is it any wonder that this data has failed to produce consistent and reliable results?
(EN: I'm inclined to agree, but also to observe that serotypes would not exist but for prototypes. The company that uses demographic data is likely a bit better and more focused than one that uses nothing at all - it is not as focused or accurate as it might ideally be, but it is not entirely invalid.)
In Search of Relevant Behavioral Data
Of all information available about customers, behavioral data holds the most promise for creating reliable predictive models of customer behavior - as it is actions taken in the past that are better predictors of the future than demographics.
Consider the patterns that emerge from purchasing behavior: consumers tend to have regularly occurring needs. They buy a gift for a child's birthday every year, they get their chimneys cleaned the same month each year, they buy a new car every four years, they go grocery shopping on the same day each week, etc. Knowing this, you can approach a customer at a time when they are likely to be considering a purchase.
In these instances, customers are receptive to and even grateful for advertising, as it calls attention to something that will provide them with a benefit. If a customer who was planning to service their car on the weekend gets a coupon for an oil-change on Thursday, they are delighted.
But get it wrong, and you merely seem foolish: consider when a person who lives in an apartment receives a promotion from their bank for a mortgage refinance. It's doubly ignorant of the bank because not only should they be aware the customer writes a rent check every month, but his address contains an apartment number.
Amazon has long struggled with purchasing behavior in recommending items based on what the user just purchased. Sometimes, the logic is too simplistic: a customer who buys a children's book as a gift for a relative will continue to receive promotions for children's books for a long time afterward, even though it is atypical to the kind of books they usually buy. In other instances, the logic is entirely absent, such that the customer is confounded as to why the item would be recommended.
As such, getting it right can be very effective - but getting it wrong damages your credibility: it doesn't take too many inappropriate recommendations to convince a customer that it's not worth their attention.
How People Buy and what to Do about it
Data of any kind is only of value if it can provide an accurate indication of how and why people buy. In a way, demographic and psychographic data are meant to imply this - because a person belongs to a group he can be expected to behave in a manner similar to other members of that group, but it is often very indirect, hence inaccurate.
Further, data is only of value to a business if it is actionable. There is a great deal of research done to gather information that is never put to user - and in many instances it is fairly obvious that the data being collected could have no practical use. As a result, organizations have amassed huge amounts of useless information, and are now subjecting it to statistical analysis to discover if there is any meaning to it.
(EN: This came around again, and has not yet gone away, with the notion of "big data" - and already there are accusations of the two extremes. Some firms are torturing the data to elicit a confession; others are prostrating themselves before data, even when it yields bizarre results. In all, it's like looking for an answer when they haven't asked a question.)
The answer to the problem is not better information, but more meaningful questions - those that inquire into the behavior and motivation of customers, and whose answers will provide an indication of actions that can be taken.
Persuasion Architecture : A New Framework
In working with various clients, the authors found themselves taking bits and pieces from different disciplines and methodologies, evolving their patchwork approach over time, and feel it has become successful enough to advocate to others under the name "Persuasion Architecture."
(EN: There's various other happy-talk about the success of the system, but nothing substantial.)