Cracking the research code

From our earliest days in school, we're taught to categorise, shifting information into sets and groups to make it easier to work with. Often, these sets are used so often that they assume a kind of natural order, becoming seemingly inviolable.

The marketing world has its own presets, such as the 16-24 and 50+ age groups. People of a certain age, or income level, are assumed to consume products and media in similar ways. These accepted groupings not only make the masses of data in AMPS easier to work with, but also give the industry one consistent language, making it easier to compare data.

At times, however, you need to work outside of the presets, creating your own groups to suit your market.

“Coding your own filters can be beneficial when analysing certain products or media,” says Botha. “Many brands, for instance, don't fit neatly into the preset age definitions. Once you've done a detailed analysis of the product or media consumption for a brand, you may find that you'd rather speak to people aged 30 -40 than use the 25 - 34 and 35 - 49 groups.

“Doing your own coding lets you shift the breaks within your target market, giving you the flexibility to redefine an age or income group, or even apply new definitions for heavy, medium and light media and product consumption.”

A number of people, however, think that this flexibility can only be achieved after hours of laborious coding. In fact, Botha says it couldn't be simpler.

Creating your own codes

“Let's say you're targeting men who are heavy consumers of chocolate, and heavy drinkers of beer. Instead of manually putting in these filters each time you access data for this market, you can code your own group, which can be saved and used in all your future data runs,” she says.

Botha explains the process, which can be used to harness together any of the filters in the SAARF databases. “To set up a code for these chocolate-eating, beer-drinking men, you'd simply go into ‘Demographics' and select men, moving this filter across into the coding grid. You'd then click on ‘and' and select choc bars heavy ‘and' heavy consumers of beer. You'd click on each filter and pull it into the coding grid. Then go to ‘Codes', ‘Save own codes', and give your grouping a heading - ‘Heavy choc/beer men' - and then save. That's all there is to it. You can now use this code in any future analyses.”

According to Botha, you should be able to code a complex run within a couple of minutes.

“If it's taking you longer, however, you're probably not using the TNT programme to its full potential. I find many people don't use all the features which have been built in, often because they don't even realise the functionality is there,” she says.

She recommends calling the helpdesk and checking that you're taking the easiest route to set up your codes, or that you're on the right track when you have a complex code in the works.

Code violations

Of course, when something is this simple, there's bound to be a caveat or two. Botha warns that users must be aware of the implications of the codes they're putting together, or they could get the wrong information out. “Be careful of mixing a filter with a household weight with a filter with a personal weight,” she says. “In some cases, it can be done, but in others it's less advisable.”

She gives these examples. You could combine household income, which is weighted up to the number of households in the country, with age or gender, variables which are weighted up to the number of people in the country. Essentially, you're asking how many people of a certain age come from households with a certain income. “In this instance, mixing a household and personal weight wouldn't be a problem.”

But not all household weights can be mixed with personal weights, certainly not without complicating the final outcome, such as if you tried to create a code using households with dishwashers, and the number of people who buy dishwashing liquid. “Would you quote your findings as number of households, or number of people? This code would produce confusing results,” she says.

Another caveat is to watch out for unstable data caused by too small a sample size. When you create too detailed a group, you could find that AMPS is drawing the results from a small number of respondents. If you have fewer than 50 respondents, you move into unstable territory and TNT+ automatically flags results with ‘**' where the respondents are less than 50, and with ‘*' where the respondents are less than 100.

“Again, when in doubt, give SAARF or the Telmar helpdesk a call,” she says. “And a final word of advice: read the questionnaire to get a precise fix on what was actually asked. This will help you determine whether combining certain variables is a good idea or not.”