Marketers aiming their brands at large segments need to find a common thread. What I mean is this: Find the core connection that motivates people with similar demographic, behavioral, or psychographic characteristics. Indeed, marketers rely on these “threads” to guide their messages, promotions, and branding.

Many believe that commonality is a natural dynamic of human behavior, fashioned by our peers, personalities, families, schools, and community leaders. Moreover, it holds the secret to retaining your loyal customers and preventing customer churn. It’s one thing to know there’s a significant trigger that can fire up the emotions of many consumers to favor your product, but finding out what that may be is quite another. That’s where stratified random sampling comes in.

What is stratified random sampling, and what is a stratified sample?

Picture this: Atlanta city and districts are home to millions of people. Among them — a relatively small sliver — are Hispanic females with a college education, ages 25 to 35, unmarried but earning more than $60,000 annually. Hypothetically, it may be a huge target market for a Latin American apparel brand if the company can get this group’s attention and interest.

At the last count, there were over 300,000 people matching the demographics as described above. Of course, this is not an actual situation, but it is a realistic one. It typifies the type of customer profiles that thousands of businesses throughout the USA look at every day to carve out viable markets.

Surveying every potential consumer in a defined group numbering in the hundreds of thousands (or even many fewer) isn’t feasible. On the other hand, statisticians point out that if you can understand the minds and feelings of, say, 2,000 of them, you’ll get a picture of how the mass segment will react given the same need situation. The problem is that the target market as described above divides further into subgroups called strata, with significant inter-strata differences. Surveys following on the heels of subdivision identification could quickly reveal the common threads that stir each subgroup’s members to buy a brand.

The mindset behind it is that there’s no one, big amorphous marketplace. Rather, it branches out into separate lanes, each one leading to a highly defined target.

So, stratified random sampling covers how to:

• Divide a larger target market into meaningful customer groups (strata);
• Identify and recruit sample respondents without biases to represent each group by applying statistical tools to ensure the stratified sample sizes are reasonably accurate, and randomly drawing respondents (strata respondents, if you will) to conduct research.

Connecting a survey’s findings to everyone in the subgroup is the first practical step toward influencing buying behavior. So, it’s no wonder that the professionals alternatively refer to stratified random sampling as “probability sampling.”

What does probability sampling do for you?

Probability sampling removes the impossible task of asking every segment member what their reaction would be to X or Y happening. History shows us that a relatively small percentage of respondents can answer for the broad group.

Not only does random stratified sampling create accurate insights, but it also offers a massive affordability advantage. Aiming one’s interviews at 2,000 to get a bead on a market size that’s 150 times bigger is a no-brainer.

When you get into the weeds with stratified random sampling, you apply “random quota sampling.” (Note: Defining a segment and determining a stratified sample size is still not enough — if the survey questions are off point and reveal little, failure will be the net result.)

Back to our Hispanic example above, all 300,000 consumers should have an equal chance of joining the research panel. Next, you calculate the statistical probability of the results being accurate. You have to do this to answer a fundamental question: “Is the sample size big enough to represent the entire segment?” It’s not good enough to decide on size with “say, 2,000” as we did above.

Crucial steps when conducting stratified random sampling

Describing stratified random sampling can be confusing. To keep it simple, we’ll stick with the example above.

Step 1: Know your target market and the relevant subgroups within the market that may create different behavior and motivational patterns. So, in our example, the data revealed a target market of 300,000 Hispanic females living in the Atlanta region who also reflect:

• A college education
• Ages 25 to 35
• Single
• Earning more than $60,000 annually

Step 2: Decide on the subgroups (strata) within the target market that you want to know about more precisely. In this case of stratified random sampling, the demographic research also informs us that Hispanic females with a college education living in Atlanta, earning more than $60,000 annually, can be subdivided into:

• Living in the city: 50,000
• Living in the Atlanta districts: 250,000

That was still too wide, so digging deeper, management decides that they wanted to break it down into those who (1) only shop online, (2) only shop offline, and (3) shop online and offline (i.e., three new constructs.) Therefore, each of the three attaches to the two groups decided on above, creating six strata.

Those living in the city which show participants who:

• Only shop online: 5,000
• Shop online and offline: 30,000
• Only shop offline: 15,000

And those living in the Atlanta districts which show participants who:

• Only shop online: 50,000
• Shop online and offline: 170,000
• Only shop offline: 30,000

Now, suppose the company wants to insert another stratification — those with undergraduate and advanced degrees. Again, the split applies to all six above, multiplying the strata to 12. Introduce a fourth pair of constructs, and the stratified random sampling program goes to 24 subdivisions, and so on. I’m sure you get the idea. In short, as the strata increase, they fractionalize the original target size of 300,000 into a spread of smaller groups.

Two things about stratified random sampling should hit home at this point:

• The number of subdivisions can get so large that it either becomes meaningless or uncontrollable.
• The cost of surveying them becomes unaffordable.

Also, keep in mind that selected respondents can only fit into one stratum exclusively. So, a respondent in the Atlanta district can’t qualify for an Atlanta city subdivision or vice versa. The stratified random sampling process shouldn’t focus on a number much more extensive than six strata. Therefore, selecting the characteristic focus is vitally important.

Step 3: Decide on what your sample sizes for each stratum must be. You can see in the example above that strata are seldom conveniently equal. Therefore, it’s likely that sample sizes will be disproportionate. You can only go with the same size samples if the strata sizes are more or less the same magnitude. It’s not the case here.

At this point you’ll probably want a statistician’s input or the services of a company such as Sogolytics to decide on the two most essential items as it relates to your stratified random sampling accuracy: What the sample ratios should be to represent the target population.

Then, the size of each disproportionate sample as determined from (1) above. In other words, the number of respondents in each that will deliver an acceptable margin of error, thus representing a reasonable success probability.

Step 4: Create a random sample for each stratum. Your stratified random sampling program obtains random samples for each stratum by applying proven methods like simple random selection or systematic sampling. If done correctly, the randomization ensures that biases can’t distort the representation or your results.

Step 5: Conduct your survey — ask the same questions to each group. That’s all that’s left to do once your stratified random sampling is complete. For example:

• Q1: Would you recommend ABC brand to your friends and family?
• Q2: Why have you answered Question 1 the way you have?

At this point, anyone can see that each stratum’s respondents may answer the same two questions entirely differently. Alternatively, answer Question 1 the same, but with vastly different reasons.

• Perhaps city-dwelling online respondents have had an outstanding customer experience, whereas offline, not so much. Or vice-versa.
• Or maybe district-dwelling respondent feedback in all three categories varies significantly from one another and the city subgroups.

The possible variances are endless — that’s why you survey stratified samples. Once completed, you’ll have feedback from six distinct groups with an idea on how emphatically your brand connects with the demographics. It may lead to you dispelling segments entirely to concentrate everything on those that respond most positively.

Conclusion

Stratified random sampling is a process that establishes priorities and defines target segments precisely. It’s the fuel that feeds market segmentation exercises and keeps the wheels rolling on proficient marketing plans. Everything we’ve covered here converges on knowing your market profiles intimately so that you can construct differentiated brand offers.

If you’d like assistance with putting together effective surveys, don’t hesitate to reach out to the Sogolytics team for help.