There are several great enterprise data platforms that can put you on your way to algorithmic attribution, but many marketers don’t have the budgets to support that investment. So how do you determine which channels are performing best for you without relying on simple but unsatisfactory attribution methods like first exposure, last click or arbitrary weighting?
Here’s a relatively easy way to measure the incremental value of each marketing channel to determine which channels are performing best so you can optimize your marketing mix.
First, pick a set of geographically similar markets — one for each channel that you’re using plus one to act as a control cell. Make them as closely matched as you can in terms of size and demography — so don’t mix big markets like Chicago with smaller markets like Waco. You also want to stay away from markets that have competing media — for example, Princeton, N.J., is exposed to both New York and Philadelphia media. Data from the Statistical Abstract of the United States and Census.gov can help you select markets that work for you.
Second, create a test matrix where one of your markets serves as a control, and the balance of your markets eliminate one of the channels you’re evaluating. For example, in the matrix below all channels are used in the control cell, and one channel is eliminated from each of the test cells.
Conduct your test long enough to get a statistically reliable number of responses. With 250 to 300 in each column and each row, you can be 90 percent confident that your results won’t vary by more than 10 percent in a rollout scenario.
Third, examine your cost per response by market in a matrix like the one below. (These numbers are for illustration only and are not meant to reflect actual costs or responses from these channels.)
Cells that have a higher cost per response from the control indicate that the channel you eliminated from that geo area is valuable to you because it was lowering the average cost per response in that cell. In the example below, the geo areas where email, search and social were eliminated had a higher cost per response overall, indicating that these channels were important parts of your media mix. Cells with a lower cost per response from the control indicate that the channels eliminated from those geo areas were increasing your overall cost per response. In the example below, direct mail, display and mobile all had higher costs per response than the control cell which included all the channels.
You can do the same analysis on revenue and profit if you are engaged in catalog or e-commerce. The difference in profit between the control profit and the profit in each equally matched geo cell provides the incremental value, whether positive or negative, of the channel that was omitted in that cell.
This experiment has its limitations. Your markets will not be perfectly matched and external factors can affect your results. However, it will provide valuable insight about the interplay among the different elements of your media mix.
Finally, remember that eliminating different channels from your media mix will also have an effect on your response or sales volume. To understand how to best manage volume within your allowable cost per response or cost per order, check out this former Here’s What Counts post.