By Dani Arussi, Vice President Paid Search
The direct marketer’s perennial question is how to best allocate the annual marketing budget across a bewildering array of channels and tactics to yield the maximum return on investment. The bigger the company, the bigger the challenge. With some of the very largest companies spending billions of dollars each year on marketing, the challenge can be almost insurmountable.
Enterprise marketing typically presents a unique set of challenges which arise because of the complexity of a larger business. Typically, the Enterprise will invest in a number of marketing channels such as TV, print, radio, direct mail and online. Within each channel, there might also be a number of sub-channels. For example, online might consist of display, paid search, email, affiliate, social media and search engine optimization. Within sub-channels, it is also possible to find a number of different distribution networks such as Google, Yahoo and MSN, each of which have both search and contextual networks.
Solving the Headache of Budget Allocation
The problem of budget allocation can be solved with the concept of dollar cost averaging and the incremental cost of sale. Stock traders are familiar with dollar cost averaging and might adopt it as a strategy to lower the overall average price paid for a stock by buying in increments as the price falls. Conversely, in marketing, it often costs more per sale to generate each incremental sale. For example, in paid search, one might bid higher to receive better placement and consequently greater volume. In direct mail marketing, one might send more mail pieces to a less targeted audience resulting in higher cost and lower response rates.
Although the overall average cost per sale with each additional incremental sale increases marginally, the incremental cost of sale might be significantly higher than the average. For example, one might generate 10 sales at an average cost per sale of $10 for a total investment of $100. However, if the 11th sale costs $20 to generate, the overall average cost per sale will only be $10.91 (100+20/11) while the incremental cost of that last sale is much higher at $20. If another marketing channel has an incremental cost of sale lower than $20, all other things being equal, it would be more efficient to spend the extra dollars there instead.
Data Needed to Make It Work
This concept can be used to balance the budget across all marketing channels, sub-channels and distributions. Determining the correct allocation is predicated on explicit knowledge about each tactic’s behavior; how much each sale costs and the incremental cost of sale which are typically represented by a yield curve that defines the relationship between sales volume and cost per sale. In order to measure each tactic’s yield curve, historic tracking data is required to connect each individual sale to its corresponding marketing tactic. In the real world, especially the offline world, those connections can be very difficult if not impossible to make. Wait. Hope is not lost.
For offline channels, rudimentary knowledge and ‘guesstimates’ can be useful, at least directionally. Or perhaps some sort of sampling process can be adopted to measure a small fraction of the overall response activity manually, the data for which can then be extrapolated. In addition to this, a partial solution across a subset of channels and tactics can be a more realistic goal and still yield dramatic efficiency gains.
Tracking, Tracking, and More Tracking
Unlike offline channels, online channels lend themselves very well to tracking, where a click can often be tied to a unique computer (by its IP address) and thus an online sale can be connected to the originating click. This accountability factor alone is, in part, responsible for the meteoric rise of online marketing and its increasing share of total marketing activities.
Online marketing can also create tracking complexity whereby it will often generate offline sales, some of which can be accounted by telephone tracking. Furthermore, a sale is often the culmination of a number of marketing touch points and for offline marketing activities, the effect of this is almost untraceable. Online however can offer some insights in to multiple search and display activity through the use of cookies, code on the customer’s computer that can link the various activities together.
Once armed with tracking data providing the relationship between cost per sale and the number of sales, the mathematical solution adopted depends on the complexity of the underlying relationship. Generally speaking, the tools adopted come from the field of operations research. In the case where the relationships are linear or a smooth power function, as they often are for paid search, a complete solution can be attained by using a non-linear programming technique such as reduced gradient descent. One such tool, known as the ‘Solver’, is built in to Microsoft Excel and although not installed by default, it can be installed as an ‘add-in’ free of charge (See Excel help on the Solver for further info).
Is All This Worth It?
Although it can be simple in theory, solving the marketing budget allocation problem is not always easy in practice. However, the typical increase in efficiency might typically be in the range of 5% to 15%. For a company spending upwards of $1M a year on media, that can translate to significant gains. It wouldn’t be unusual for the typical enterprise level company to achieve efficiency gains to the tune of millions of dollars. That kind of reward certainly warrants the investment of significant expertise, time and effort.
Short answer: You bet it is.