Advertising Principles - Evidence-based principles

How to Analyze Media Investments

Persuasive Advertising focuses on how to develop persuasive ads, but not on how to deliver those ads to customers. Here are five methods that can be used to decide a media budget as well as how to allocate that spending across various media. I start with the weakest method and work towards the strongest. Ideally, one should use all five methods, but put more weight on the findings from the stronger methods.

1. Last year's advertising expenditures for the brand. This method assumes that last year’s advertising was optimal. The task is to determine how expected changes in the future would affect changes in the ad budget.

  1. Ask two or more heterogeneous experts, working independently and anonymously, to examine last year's advertising-to-sales ratio. Afterwards, they should provide a written list of the arguments favoring increases or decreases in advertising expenditures. They would then be asked to make subjective estimates of whether to increase or decrease advertising expenditures and by what percentage.
  2. The judgmental estimates from the experts would then be averaged, presumably using equal weights.
  3. Obtain advertising-to-sales (A/S) ratios for as many prior years as possible for this product.
    The Delphi procedure might be useful with steps “a” and “b” above.

2. Typical expenditures for the firms in your industry. The typical expenditure method assumes that one’s competitors are rational and that it is not wise to depart substantially from typical behavior unless there are strong reasons to do so. This is usually a reasonable assumption although advertising warfare and hysteria (such as with the advertising during the dot.com bubble) create problems.

  1. Ask experts, working individually, to write down arguments in favor of, or against, an advertising-to-sales ratio that differs from the product category average. For example, the development of important, new products would increase A/S ratio. It would also be higher (lower) if it had a Glossary Link campaign that followed (did not follow) many of the advertising persuasion principles, or if it did well (poorly) relative to other ads in Glossary Link copy tests.
  2. Ask the experts to provide estimates of how much the A/S ratio should depart, in percentage terms, from the typical expenditure.
  3. The estimates from the experts would then be averaged, presumably using equal weights. To the extent that there is uncertainty (e.g., as shown by widely varying estimates), be conservative and avoid large deviations from the typical A/S ratio for the product class.
  4. Consider using the Delphi procedure for steps “a,” “b,” and “c.”
  5. Obtain historical information on the A/S ratio for the product in question and adjust it by the estimates in “c” above.

3. The Glossary Link elasticity method and Wright’s Rule. To apply the elasticity method, you can start by assuming that an advertising campaign will have a typical elasticity of about +0.1. This is based on the Sethuraman and Tellis (1991), Tellis (2009) and related papers summarizing over 260 estimates of elasticities. Better yet, find the elasticity for your situation considering the type of product type and the media. You can get some idea for this from Table 1 in the Setheraman and Tellis paper. For example, the elasticity for durable goods is about 0.23 while that for nondurables is about 0.03. The elasticity for TV commercials is about 0.03 while that for print ads is about 0.17.

The elasticity would be higher if your campaign followed the prescribed principles, and lower if it violated principles. In addition, elasticities are higher in the introductory phase of a product. They are lower at the brand level than at the industry level if the competitive brands tend to match their competitors advertising expenditures (Assmus, Farley & Lehman 1984).

Once you have an estimate of the elasticity, apply Wright’s Rule (Wright 2009) set the advertising budget at a level equal to the elasticity times the gross margin on each item of the product being advertised times the forecasted sales.

For products that have been advertised for some time, the elasticity method can also be used to examine modest changes in proposed advertising expenditures. For example, to assess the impact of an increase of 10% on advertising expenditures, use the elasticity to assess the changes in the number of additional units that would be sold, then calculate the change in the gross margin that would be achieved.

4. Decomposition. Take a complex problem, break it down by parts, estimate each part, and then combine them. It is especially useful when the breakdown allows one to draw upon different sources of knowledge and when each of the components is subject to less uncertainty than is the global estimate of elasticity. For evidence on the effectiveness of this method, see MacGregor (2001).

Determine elements of the process by which an ad can lead to behavioral change. The decomposition procedure will guide you through this process. Here is a possible format for a print ad. There are a number of other ways one might use to structure the decomposition such as:

Number of people in the target market exposed to the ad = T Percentage of above who

  1. See the ad
  2. Read the ad (% of “a”)
  3. Understand the ad (% of “b”)
  4. Intend to purchase because of ad (% of “c”)
  5. Actually purchase (% of “d”)
  6. Repeat purchases from new buyers (% of “e”)

    ROI = [{(T)*(a)*(b)*(c)*(d)*(e)*(f)*unit margin} minus the cost of campaign] divided by the cost of the campaign

The inputs can come from data or from expert judgments. For example, to estimate “understanding” of an ad, one might draw upon data from Beltramini and Brown (1994) who estimated correct comprehension of 80% for print ads and 73% for TV commercials.

When relying on judgmental estimates, obtain independent estimates from at least five people, each of whom has knowledge about the effect of advertising. Ask each expert to provide an estimate of the expected effect of the ads, and also provide 95% confidence intervals about the estimates. Use equal weights to combine estimates of the expected effect and of the confidence interval.

*See the Glossary for an explanation of elasticities.

5. Experimentation. Field experiments are ideal, given that it is feasible and economical to obtain information on response to the ads.

Experimentation can be easily done for direct response advertising. For example, if the advertising will be done by direct mail, try a sample of 10,000 names and test an alternative version of an ad, say 5,000 of each version. Examine the profit made from the resulting sales and divide by the variable costs of the test ads (ignoring the cost of developing the ads and any fixed costs) to determine the return on investment. If profitable, then send the most effective ad to the rest of the mailing list. If they are close in effectiveness, send different ads to each subsample to spread the risk.


© Copyright J. Scott Armstrong and Kesten C. Green. All rights are reserved.

Contact us with any suggestions here