November 2008

Creating demographic
variables using survey data
Here’s what to do when the data you
need about niche markets isn’t available. By Daniel Lemire

The Canadian census provides extensive data on more than 1,800 variables such as the age pyramid, income, ethnicity, employment, religion, households and dwellings. Most of the time, we are able to define which of these variables are the most correlated to the consumption of any given product to allow us to build a forecasting model. Moreover, if the answers aren’t immediately apparent in the census data, we can often obtain data on some 4,000 behavioural variables that are available in the market. Indeed, it is possible to purchase data from certain specialized suppliers of various geodemographic variables extrapolated from the data collected by the Print Measurement Bureau (PMB), NadBank, certain financial databases or the Family Expenditures (Canada) survey (FAMEX).

“To begin with, review your survey data. In all likelihood, your company already conducts an annual customer opinion survey.”

Obtaining specialized data creates special challenges
However, it sometimes happens that a specific project requires specialized data. Indeed, certain niche markets may be insufficiently quantified in terms of the availability of data. For example, if you were selling clothes for small women, it would be impossible to find any helpful information on this market. Of course, women’s clothing consumption statistics are easy to find, but the breakdown between small, regular and larger women is simply not available. If you were manufacturing recreational vehicles and you were interested in the personal watercraft market, reliable, easily available data on this market just doesn’t exist. Finally, if you operate a video club network and wish to increase your presence in the Blu-ray Disk market, you would currently be unable to improve your knowledge about this market with the currently available demographic data. Fortunately, there are solutions to address these situations; in particular, three approaches that merit consideration.

To begin with, review your survey data. In all likelihood, your company already conducts an annual customer opinion survey. In the overwhelming majority of cases, certain control questions are asked (age, income, gender, family status, size of household, etc.). These control variables could be used as a bridge between your sample of respondents and the Canadian census to create a new geodemographic variable (say for example, to determine the breakdown in the spending habits of “petite” women in Canada).

You can also use the data in your customer database to calculate ratios, and then apply the results to the entire population—to the extent that your customers are a representative sample of the Canadian population. (For example, this could be used to calculate the degree of consumption of Blu-ray Disks per income bracket.)

Finally, it is even possible to collaborate with your competitors to exchange information anonymously. For example, all recreational vehicle manufacturers in North America send their monthly sales figures to a third party who takes care of compiling the total amount of sales per FSA in Canada and per ZIP Code in the USA.

This same third party then calculates each manufacturer’s market share and delivers the performance results back to each one individually without disclosing other competitors’ performance. As a result, each manufacturer knows how it’s performed, as well as identifying the high penetration zones wherein one could build and fashion a customer profile.

Calculating the “Size of the Pie”
Regardless of the source of available data, the methodology for extrapolating it from a sample to the entire Canadian population remains the same and can be broken down into four steps. First, one must define the size of the overall market, if it is, in fact, a segment of a larger market that is possible to quantify and validate. For example, based upon the FAMEX survey, one can define the men’s clothing market as a $15 billion market. If you’re interested in the clothing market for “big” men (6’ 4" and taller), one must necessarily end up with a market that is worth less than the overall amount. Given that it’s unlikely for men to be this tall before being an adult, one can take away the children’s and adolescents’ clothing market (let’s say $4 billion) to end up with an $11 billion market. Next, Health Canada data may tell us something about the number of “big” men per age group and per province. These numbers, multiplied by the average clothing spending of Canadian men would give us a market in the area of $2 billion. This is what I call the “Size of the Pie.”

One must then determine what demographic variables explain how the pie breaks down across the country. The more men age, the less they tend to be big. The more well-to-do they are, the bigger they tend to be. The more they work in liberal professions, white-collar jobs, or trades related to agricultural work, the more they will tend to be big as well. These variables, which correlate significantly with the relative presence of big men, can now be used to calculate weighting ratios that can be applied to each of the Dissemination Areas (DA). The census tells us precisely how many men 18 years old and older live in each DA. The younger the age pyramid, the more weight we will give to the probability of finding “big” man in any given DA, (e.g., the higher the income, the greater the probability of finding “big” men, etc.).

Dispersion matrix
Armed with these ratios, we will finally be able to calculate the dispersion matrix that would enable us to estimate the number of “big” men in each of the 54,000 neighbourhoods in Canada and thereby recreate a market in the area of $2 billion.

Statistics, databases and geodemography are some of the tools available to you to help you increase your knowledge about micro markets and returns on any potential investments in advertising or direct marketing.

Be imaginative and make good use of the information contained in your own surveys, databases and the publications in your industry to create instructively relevant variables for your research. When in doubt, don’t hesitate to consult a statistician, a demographer or a geomatics professional.

Daniel Lemire is president of Indicia Inc., a fact-based marketing organization located in Montreal. He can be reached at 514.722.3699 or by e-mail at daniel.lemire@indicia.ca.