Wednesday, November 26, 2008

Are Advertising Divisions Efficient or Inefficient?

In today's Wall Street Journal there is an interesting article ("Marketers Reach Out to Loyal Customers") about retail advertising. The basis of the article is that given the economic difficulties this holiday season - advertiser's are turning to statistical models and other technologies to advertise to customers.

My glass half-full reaction is this is great - firms are using their existing customer behavior to find ways to deliver ads based on customer's past purchasing patterns as opposed to the typical shotgun approach of hoping the ad message reaches their intended audience. Also firms like Sears are offering different discounts to customers based on predictions about the value those customers will bring to Sears in the long-term. As long as firms are not systematically discriminating against customers, this moves advertising in a strategic direction as another tool for firms to employ in the pursuit of maximizing profits.

But then my glass half-empty reaction is why have firms not already employed statistical models earlier - since the article says this is a new tool being used? I will offer three reasons off the top of my head - feel free to add others or comment on the three below.

The first is that using statistical models is not sexy. By this I mean that most advertising and marketing people's comparative advantage is in non-quantitative areas as opposed to quantitative analysis. Thus the ability to create ads that appeal to our emotions is perceived to be of more value in the hiring of marketing/advertising employees over individuals with the ability to apply statistical modeling techniques of customer sales data. Of course, I could be wrong, but my perception is that most marketing departments value qualitative skills over quantitative skills. Thus some marketers may view using scarce budgetary resources in employing statistical modeling techniques with skepticism.

A second reason is that advertising departments do not use statistical modeling because they are inefficient. In house marketing and advertising departments can take many different paths to create a profitable marketing campaign even if they ignore mining customer sales data in creating an advertising campaign. The question I have as an economist is, can the use of existing customer behavior make advertising/marketing even more profitable? If the answer is yes, then ignoring a tool means that advertisers are inefficient - even though it is still profitable - and thus an opportunity exists for some firms to gain a competitive advantage over their rivals.

The final reason is advertising departments only use statistical models when they are economically profitable. Since employing statistical analysis is expensive, it may only be profitable to use statistical analysis when there are substantial benefits to be gained such as during the Christmas holiday season. With greater demand the marginal benefits will be higher of advertising more efficiently and can possible cover the marginal costs of using this type of advertising tool. During periods of lower demand the marginal benefits may fall and not justify their use since it is possible that the marginal benefits of statistically analysis can be lower than their marginal costs.

An assumption in the reason above is that non-quantitative advertising methods are more efficient than quantitative methods, and as an economist I am highly skeptical that this is always true. Given that advertising to existing customers is much cheaper than to new customers, it seems that the marginal costs of using statistical analysis should be lower in some cases than the marginal cost of creating new advertising campaigns.

No comments: