Analysis of Advertising Dollars

An online retailer was getting ready for the holiday season and had placed ads with four different websites. They were looking to increase holiday-buying-season traffic to their site.

My task: determine the effectiveness of the advertising dollars.

First step: gather data

Analysis can be done when there is ample data available. In order to determine the effectiveness of the advertising dollars of my client, I had to first orient myself to the nature of their online traffic.

Online traffic analytics tool allows me to view a website’s overall traffic.

Figure A

In Figure A, We can see that the website’s traffic is divided into 3 sources. The one we are focusing on is “Referral”, which describes traffic originating from another website (that is not a search engine). This will include traffic from the client’s advertisers.

The second thing we want to look at is how much traffic is derived from each of our individual advertisers (Site A, Site B, Site C, Site D). In Figure B, we can see that over the entire data period, each of the individual advertisers contributed a decent amount of traffic (though they each only represent a fraction of the total traffic count).

Figure B

Second step: cost-benefit analysis

Why are Superbowl commercials going for $5 million per 30 seconds? Because the Superbowl consistently provides 100 million+ views or “impressions” (source). With a reach that wide, many large, marketing teams gladly pay for a Super Bowl ad spot because of the value (< 5 cents per impression).

To gauge the effectiveness of the client’s advertisers, a similar approach needs to be taken: we look at the amount of traffic received from the source and compare it to the cost of placing the ads.

Figure C

In Figure C, we start to compare traffic received against cost of ads. However, this graph is not particular useful, because there is no common base or denominator to compare them; we are comparing apples and oranges. To generate numbers we can actively use for analysis, we divide the advertisers’s fee by the amount of traffic received (“views,” in this case) to get a more useful number: the Cost-Per-Click Ratio.

Figure D

Now that we have calculated the Cost-Per-Click Ratio, we can compare the ratios of each advertiser. We do not need further deviation and variance analysis to see that one of these advertiser’s Cost-Per-Click Ratio is particularly high.

Last step: use findings to assist in decision-making

When it comes to cost ratios, we always want to keep them low. When it comes to ratios, they do not automate decisions for us, but rather, they provide useful data for us to make decisions.

Here are some basic ideas the data presents:

  1. Site C has a much higher Cost-Per-Click Ratio than the others, so it would be the most expendable. The reason behind this is if there is a limited budget, the least efficient should be the first to go.
  2. A general strategy would be that if there is a budget that needs to be best utilized, then it may make sense to remove the funds that are allocated to Site C and then use re-allocate them to more efficient advertisers, such as Site A.
  3. However, there are constraints to re-allocation, as we can’t be certain that the same Cost-Per-Click Ratio (a metric of cost efficiency) will continue if we were to double a budget for a particular advertiser. After all, if an advertiser has a reach of 200,000, showing more ads to that audience may not make them more likely to purchase than the first time the ad was shown. Or, if there is an increase, it may not be proportional to the increased in spending (doubling the ad spend may not result in doubled visitors and sales originating from that site).

Note: this analysis does not include an even more useful metric, Cost-Per-Conversion, due to the simplicity of this example.