Measuring marketing effects

So you have decided to do something about that dwindling favorite product of yours. You have decreased the price to spurn its sales, or maybe you have started a nationwide ad-campaign. After a while you are asked to report on the success of that activity. After all, was it worth the while to spend all this money? Looking at the sales data, you will typically see something like in the graph on the left.

You are faced with a number of problems:

• the sales pattern is "noisy",
• though sales have gone up, they had also gone up in the past (just not as much as you had hoped for),

Intuitively, what you need is a "null-scenario" or a "baseline" to compare your actual sales with. Clearly that means something like "forecasting what your sales would have been, had you done nothing, based on the data before you actually intervened."

Now, does that sound complicated? From the point of view of time series analysis, the situation is simple. One applies the whole set of forecasting techniques (cf. our article on Forecasting ):

• extracting trends and seasonality,
• testing the influence of external factors,
• creating factors explaining your business logic,

• A variable describing the activity you want to measure.

So, forecasting techniques can give you a pretty good quantitative notion of the marketing effect. The point is to use results like these to guide future decisions about choosing marketing activities. The

• Price elasticity e : Price decrease by x% led to demand increase of e*x%.
• Advertising elasticity a : Increase of advertising budget by x% led to demand increase of a*x%.