InterSurvey's statement on methodology
(Some graphics removed.  Visit their site for full presentation.)

(Also, see here for an email exchange about these methods)

Unlike Ordinary Web Surveys,

InterSurvey's unique approach to consumer research is built on the solid statistical foundation of probability sampling.

Sample Quality

"It all depends on who you ask."

No matter how good your questions are, if you're not asking the right people, your answers aren't going to be worth much.

For any type of broad-scale consumer research, the "right people" means a statistically representative subset of your target population. Usually, respondents are selected through some sort of random procedure which draws from the entire population, or "universe" of the audience.

For example, if you want to know what people all across the United States think about your product, you can't just ask the folks in Oshkosh. You have to ask -- or "sample" -- a representative cross-section of people from every part of the country.

Obvious? You'd be surprised how much consumer research is only done in "Oshkosh."

Why You Have to Get It Right

If there's anything worse than not knowing, it's thinking you know when you don't.

That's the pitfall of any research that isn't grounded in sound statistical science: drawing conclusions for an entire population based on data from an incomplete or non-representative sample. For example, finding out about the weather in California by asking the people in Oshkosh.

The Web -- with its instant access to consumers and whiz-bang multimedia capabilities -- has become a highly attractive medium for conducting consumer research and opinion polling. The bad news is that there's no scientific basis for drawing inferences from current Internet samples.

The fact is, most Americans don't own computers. Even fewer have access to the Internet, and of those who do, only a tiny percentage takes part in online surveys. No sampling technique can be representative of population members who have no chance of being selected. So these aren't probability samples at all; they're volunteer, "convenience" samples for which probability sampling theory simply does not apply.

The good news is that InterSurvey has a solution. We draw a random sample of households using a scientifically-accepted computerized technique called Random Digit Dialing (RDD), and then provide those households with Internet access.

That way, we can offer all the speed, convenience, flexibility and economy of online delivery, without the sampling biases of ordinary Web survey methods. Most important, you can be sure you're getting it right.

What Traditional Methods Can't Do

What Traditional Research Can and Can't Do

Research using existing survey methodologies often involves a tough tradeoff between sampling reliability and interviewing logistics.  

Telephone surveys. Some techniques, such as Random Digit Dialing (RDD) telephone surveys, provide random samples that are guaranteed to be representative of the target population. Scientifically, they're rock solid.

But...the telephone allows only limited interview content, poor audio and no video. It's also getting harder to convince people to participate, because consumers often equate teleresearch with annoying telemarketing.

In-person surveys. If you're face-to-face with the respondent, there's not much you can't ask. Audiovisual material, graphics, verbal questions - anything goes.

But...when the interviewer has to make "house calls" (at $200 or more per interview), costs can quickly become prohibitive, especially when sampling across a wide geographic area. In fact, the Federal Government is just about the only organization still doing large-scale, in-person interviews.

Focus groups and mall intercepts. These get the researcher face-to-face with respondents. Because they allow visual content, and even live product demos, they can be tremendously informative.

But ... the sample size is small, geographically-limited and far from random.

Mail surveys. It's not hard to build a cost-effective, scientifically sound mail survey. And it's a relatively easy matter to administer it.

But ... it's called "snail mail" for a reason. This is by far the slowest market research you can do. And response rates to mail surveys, even with incentives, are often too low to provide good reliability.

Volunteer Web surveys. Definitely a step in the right direction, volunteer Web surveys are fast and inexpensive. Some of them take advantage of the multimedia capabilities of the Internet, too. Even better, Web surveys are easy to administer.

But ... as they're done today, Web surveys are so statistically biased that their results are virtually useless to serious market researchers. Self-selected surveys of current Internet users reach only about one percent of the American public - predominantly single, young, affluent, well-educated males - and miss the majority of consumer households.

Why We Can Do It

And They Can't

InterSurvey research takes the best of traditional methodology -- probability sampling -- and applies it to the cutting edge technology of the Web.  

Other Internet-based research relies on scientifically invalid "self-sampling." What we do is considerably more complicated and effective.

First, InterSurvey selects a representative sample of all households using Random Digit Dialing, a scientifically proven method of random selection. Selected households who agree to enroll are equipped at no cost to them with the necessary hardware to receive our Internet surveys. This is all done on a standardized platform that allows the delivery of high quality audio, video and other forms of multimedia.

By using probability sampling to select panel households, InterSurvey can provide accurate and valid information about any target population. This gives you access to all of your customers, online or not. It also gives you the certainty that your research data will be based on nationally representative samples, or demographically targeted samples, of your customers.