Click here to return to Stakeholder360 website or use your browser's 'back' button.
Haga click aquí para volver a la página web del Stakeholder 360 o utilice el botón de retorno de su navegador.
Presented at Professional Marketing Research Society annual conference, Toronto, June 2000
by Robert Boutilier, Ph.D.
President, Robert Boutilier & Associates
Vancouver, B.C., Canada
It's a great pleasure to address my colleagues today. I want to share with you some thoughts I've had on how management's expectations of marketing research are being changed by several forces that are historic in scale. I will first try to describe the evolution of those forces, and then look at what they imply for the evolution of marketing research practice.
Briefly, historic changes have created a situation where we are now expected to contribute to ongoing dialogues with consumers who are already customers of the company sponsoring the research. And these dialogues are ongoing, because both the customer and the company are already getting benefits from the relationship. This is a very different kind of expectation to have to meet. And what I want to talk about today is how it re-orders the priorities we place on some of our fundamental principles of scientific research. Specifically, it makes the principle of avoiding self-selection bias less important, and it makes more important the principle of using multiple techniques to make sure you get a well-rounded view of the problem.
A typical marketing research study involves a set of focus groups and a random sample outbound telephone survey. These are two different techniques, one qualitative, one quantitative. They give us insights into different aspects of a marketing challenge because each has its strengths and weaknesses but the strengths of one compensate for the weaknesses of the other. What I want to suggest to you today is that even within the categories of quantitative and qualitative, there is still much more to be gained by using a variety of techniques, including those that are loaded with self-selection bias.
Marketing research was born during the heyday of mass media. Just a few magazines had enormous audiences. The fragmentation in the magazine market had not yet happened. Likewise, radio commanded mass audiences. Later, television superseded radio as the mass audience medium.
Television sets became widespread during the 1950s when there was a rise in the affluence of masses of industrial workers. It was also a time when we were evolving from a society of producers to a society of consumers. People were still migrating to the cities from the farms. Many were being converted into "consumers" of household products. Time-saving products were developed and marketed to people who were still sewing clothing by hand and washing dishes in the kitchen sink. "New, improved" packaged goods were developed for those still brushing their teeth with baking soda, cleaning their windows with vinegar and water, and using home-made chicken soup as a cold remedy.
Marketing focusing discretionary dollars on particular brands to an extent that created massive economies of scale. When marketing succeeded, the demand was so great that high quality goods could be sold at low prices, with the profits coming from the volume of sales. If it weren't for marketing, no one brand would ever have enjoyed the high volumes that made the low prices possible.
In that type of social context, marketing researchers were called upon to sample the opinions of the general public, the mass of consumers.
The assumption was that people who hadn't heard of your product would buy it if you told them about it. New customer acquisition was the name of the game. Stirring up the motivation to buy was not the problem. The problem was just a lack of awareness. In that mass market context, segments were not very useful. They only served to narrow your product's appeal. And retention was not an issue because it was cheap to replace lost customers with new customers.
In the context of mass marketing, random sampling became an ideal because it allowed generalization of findings to the entire population of interest. When the population of interest was basically everyone, then there was no problem finding a large sample from it.
Random sampling, by definition, allows each and every member of the population a chance to be selected into the sample. The crucial question for this discussion revolved around who does the selecting. Offering everyone a chance to participate, and then letting them take the initiative to opt into the sample, might seem like satisfactory solution. This is known as "self-selection".
Unfortunately, evidence showed that there were crucial differences between those who did choose to participate and those who did not. For example, in the packaged goods categories, self-selection led to over-sampling of heavy product users and under-sampling of non-users. When the research goals were to find ways to expand the market among current non-users, this was a very negative sample feature. It was labelled "self-selection bias".
It was discovered that you could get large numbers of non-users to respond to surveys if the marketing researcher reserved the right to decide who would be offered a chance to participate. The list of those to be approached was randomly selected and their interest in the topic of the survey had nothing to do with being on the list.
Once approached, the prospective participant's interest in the topic could again lead to self-selection bias in the sample. However, marketing researchers discovered that they could obtain the co-operation of plenty of uninterested people by using a variety of persuasive techniques. For example, they would appeal to the prospect's respect for "scientific" studies. They would flatter the prospect's ego by emphasizing the importance of their opinions.
They would fall back on a sense of basic civility that mitigated against refusing any requests for assistance. U of T political scientist Neil Nevitte has documented a general decline in deference to authority over the past thirty years. After World War II, the population was full of former soldiers with little education but a strong belief in God and other higher authorities. Marketing researchers were seen a something akin to scientists, and when researchers asked for a moment of your time, most people were too polite to refuse. You only have to spend 15 minutes monitoring telephone interviewers today to know how that's changed.
In the earlier years of mass marketing, researchers approached prospective research participants door-to-door. When telephones reached a near universal penetration rate, outbound telephone calls became the preferred approach. Over the years, the marketing research profession encoded these lessons as basic operating standards. Researchers generally agreed that the outbound, telephone survey offered the most cost-effective prospects for minimizing self-selection bias.
Marketing researchers have successfully continued to find ways to minimize self-selection bias, even though it has become more difficult, and therefore more expensive. Not only do residential telephone subscribers have an array of call screening technologies to choose from, more importantly, they have reasons to use them. Electronic moats have been installed around the average household and as a result, the tried and true outbound telephone survey has become more expensive to conduct properly.
Why? First, the public has less respect for the marketing professions in general. Some observers feel that the telemarketing industry has hurt the image of all marketers. Even the minority of consumers who know the difference between telemarketing and marketing research, regard both in with the same low esteem and do not want to spend time conversing with either. Call screening technology lets them make that choice without giving the interviewer a chance to say a single word. Consumers are no longer flattered that someone wants to know their opinions. Today they are more likely to ask interviewers what kind of incentive is being offered in exchange for a completed interview.
Second, people no longer consider it impolite to let their answering machine answer for them. They have become more accustomed to carrying on conversations across a series of voicemail exchanged over a period of days. The days when dinners could be interrupted by a ringing telephone are fading fast.
Third, with the internet, home offices, and home fax machines, the number of hours spent using telephone services has increased in general. The result has been a traffic jam. Households with school-aged children suffer a shortage of telephone service most evenings because of heavy internet use. This makes it difficult to avoid busy signals or some other dead end like "call answer".
So random samples are harder to come by. They cost more now, and if it hadn't been for the price wars among long distance suppliers they would cost a lot more by now. But if they're worth having, we can still get them. And if we're still called upon to make generalizations to large populations, then they're still worth having. But that's big question. Are we still trying to generalize to the whole population?
Today, no single television network or program commands a mass audience. Consumers have hundreds of programming choices in an evening. TV is going along the same road followed by magazines. Each outlet caters to a smaller market with more customized, specialized content.
Just as TV superceded radio, the internet is about to supercede TV. However, the internet started out as a completely specialized, customized medium. It does not have a "mass market website", and maybe never will. On the content provider side, the costs of entry for competitors are too low, and on the consumer side, the literacy requirements are still too high to make websurfing accessible to all.
By the late 1960s, marketers began to realize that there were no "non-consumers" left to convert over to consumerism. The masses had been carved up into fiefdoms of customer loyalty. Any expansion of market share had to come out of a competitor's share. This made defending one's customers a higher priority.
By the late 1970s, this defensive orientation was picked up in earnest by the then fledgling services marketing community. Their rallying cry became, "It's cheaper to keep an existing customer than to acquire a new one." By the late 1980s, even hard, physical products were being differentiated by their service components (e.g., new car extended warranties, financing).
By the late 1990s, the quest for customer loyalty led us to one to one marketing, relationship marketing, and database marketing. They all share an emphasis on personal customization, as opposed to mass offerings. The aim has become to prevent defections by establishing a trust-based relationship with each and every customer.
Of course, this transition from the mass market to the individual market took decades, and that made it harder to notice. It started with the early segmentation systems which carved up the masses into five or six five or six lifestyle categories. Then segments got more specific to product categories. Then we got geo-demographic segments and there were sometimes two or three hundred of those. Then with data mining of in-house customer information systems, some segmentation systems get up into the thousands when channel preferences and the like were taken into account. That makes for segments that are anything but massive.
Although the changes in consumers have increased the cost of minimizing self-selection, they do nothing to undermine the goal itself. If we want random samples, we can get them. Getting a random sample from a segment that contains only 25,000 people, all scattered across Canada, is like finding a needle in a haystack, but with enough resources, we can do it. Question is, when the population gets that specific, are we still doing mass marketing? And if we're not still doing mass marketing, is it still as urgent to avoid self-selection bias?
Relationship marketing and one-on-one marketing approaches have changed the kinds of questions researchers are asked to answer. Now researchers are more interested in interviewing only those people who show an active interest in the product. There may be only 25,000 of them across Canada, but because they do show an active interest in the product, they take the initiative to identify themselves, either through purchases, or inquiries that get captured in databases.
These changes in the practice of marketing have made the quest for lower self-selection bias less relevant. We're moving from wanting to talk to everyone to wanting to participate in one-to-one personalized dialogues. So when a marketer wants to talk to Joe Canadian at 123 Maplewood Trail, we don't need a random sample. We just need to talk to Joe.
Instead of learning about Joe by treating him like the subject of a scientific investigation, now we learn about him by carrying on a long-term dialogue with him. This requires replacing the quest for a random sample with the quest for getting to know each customer as an individual. Instead of wanting a high resolution snapshot of the whole population, now we want an continuous video picture of each evolving relationship.
Moving from random sampling to self-selected participation in dialogues also involves a different set of ethical considerations. When we call up Joe Canadian and ask to speak to him, he is not anonymous. When Joe sends your company an e-mail or posts a message on a product discussion board, you really do want to merge his comments with the existing data you already have on him. In the mass marketing paradigm, this would be an unethical violation of confidentiality. Why? Because in that paradigm you encourage participation by promising anonymity. It would be breaking a promise to merge his answers back into a database with identifying information.
In the self-selected dialogue paradigm, you make different promises, and often they're promises that require a lot more care and effort. You promise to remember what Joe told you last time so he doesn't have to repeat himself over and over again. You promise to keep the data confidential. You promise to get his permission beforehand if you want to sell the data to another company. You promise to respect his channel preferences. You promise to respond to his questions or comments quickly. You promise not to pester him. Spam is the online equivalent of a telephone call at dinner time. But it occurs in a social environment where the codes of conduct are much less tolerant of unsolicited commercial interruptions.
So there are ethical standards in conducting a random sample mass market surveys, and there are different ethical standards in conducting an ongoing customer dialogue. The problems arise when you apply one set of standards to the other type of activity. If you promised anonymity in order to be ethical, and then merged the answers into a database with names, you would be breaking a promise. Likewise, if you promised to get permission before sending e-mail, then it would be unethical to send an unsolicited e-mail questionnaire even if you guaranteed anonymity. The essence of the ethical question is knowing what you've promised and sticking to it.
To some consumers, using the mail is a huge barrier to responding. There's the envelope to prepare, the stamp to buy and the trip to the mailbox. Other consumers see an IVR system as a barrier. They hate to have to fit their inquiry into one of the menu choices offered. Still others find websites daunting. Computer literacy is growing, but is far from universal.
But just as everyone has a channel they dislike, they also have a preferred channel. Making sure their preferred channel is one of the choices makes it more likely that they will respond.
This is just as true for marketing research as it is for direct marketing. Direct marketers need high response rates too, so they rarely force their target audience to use just one channel of communication. More and more, the consumer is encouraged to respond by whatever channel suits them best, be it store visit, mail, phone, fax, e-mail or website.
The financial services industry has found quite important differences among people who prefer to do their banking by different channels. The differences include education and literacy levels, income levels, and investment portfolio size. Channel preferences have been used as important variables in market segmentation. Using only one channel limits a survey to only one channel preference segment of the population. That's OK if you don't care about what the other segments think. But if you do, then you have to offer multiple response channels to get them responding.
The self-selection biases, that were previously viewed as shortcomings of many techniques, now appear as strengths for a growing array of marketing research problems. That's because only consumers who are interested are of interest. This makes it less likely that researchers will automatically gravitate towards an outbound telephone survey. With the same budget, they can conduct a multi-channel study using the newer electronic channels.
To sum it up, we're moving into a new historical period. My grandfather and grandmother were farmers who produced their own consumables. My parents became mass market consumers. I have become a niche market consumer. My teenaged kids are pushing into the one-to-one approach. They compile and burn their own unique music CDs out of downloaded MP3s. When they're interested in a product or a brand, they'll chat about the company with their friends on the ICQ. Then when they take the initiative to contact the company, usually through the website, they expect an ICQ-type response, a fast, frank exchange, e-mailed individually to them.
There'll never be a time when companies don't need to understand their customers. But how you go about doing that changes with every generation.