Using Structural Equation Model (SEM) to identify KPIs

Effectiveness of Advertising
"Using Structural Equation Model (SEM) to identify KPIs"
Article No. 6 (by Hideaki Koizumi)

In my previous article, I mentioned that the most important thing when measuring the effectiveness of advertising is to determine the key factor that most contributes to the communication's final objective, sales growth for example. We call this most effective key factor 'KPI' (Key Performance Indicator), and SPI focuses its efforts on identifying it among many factors. Every category/product has a different KPI, which can not be found simply by putting data into some model formula. Recently, SPI has used 'Structural Equation Modeling' (SEM) as a way to clarify KPI in various media analyses we have done. By the way, 'structural equation' means the real relationship between cause and effect. There is a similar analysis called 'regression analysis' which is also a 'relationship between cause and effect' type analysis. Heavy regression analysis can determine how much each factor, such as advertising, distribution rate and the price of the goods, contribute to actual sales. This analysis works well for 'low involvement products', such as confections. In other words, the results show a clear influence by advertising. But the problem with heavy regression analysis is that while the advertising effect over the short-term can be clarified, it is hard to tell the effect on a long-term basis. Some people believe that heavy regression analysis does not work on 'high involvement products', such as automobiles, and that advertising has low effect on these types of products. In such cases, SEM will be a more appropriate approach. By conducting this analysis, we can see the process (or structure) between TV/NP advertising and actual sales. 'DAGMAR Theory' assumes the following process:unknown -> acknowledgment -> understanding -> confidence -> action. But this doesn't always fit for all the categories/products. The other things we can determine from SEM analysis are; how much effect is generated by each communication activities, and which activities lead to sales in the end. In practice, the 'path coefficients' on the 'path drawing' (please refer to the diagram below) show how much influence each has on each intermediate attribute. By this type of drawing, the real factor, which may look ineffective over the short-term, may actually in the long-term influence the intermediate attribute that leads to sales, can be clarified. By having a clear mechanism and structure of sales such as this, we can find multiple KPIs which exist for one brand or one category. It should be clear that not just one factor, but a mix of factors (awareness, likeability, etc.), influence sales. To sum up, with SEM analysis, you are able to clarify which factors influence sales and how much influence each factor has on sales. In my next article, for those of you who are not good at quantitative analysis, I will introduce another approach to identify KPIs in a more logical way.

Please contact us with questions or for more detailed information.
spiindex@spi-consultants.net