The inability to measure the return on your marketing dollars

In the United States, it is estimated that $260 Billion a year is spent on advertising. This is nearly half of all ad dollars spent in the world.  Assuming a population of 328 million Americans, it breaks down to $793 per man, women, and child in the country. This is a staggering sum when you think that each day those same advertisers would have to make $2.18 in a greater margin to justify the money spent. 

In the end, the ability to understand the value of marketing is so abstract as to be nearly impossible to measure or understand. Throughout history, we can point at success stories that have changed the course of a brand or the capital value of the underlying company.  The "Intel Inside" campaign which began in ~2009 is an example of an advertising campaign that changed the course of the whole company. Not many remember the prior campaign regarding the “Intel Experience”, but it was the “Intel Inside” campaign that brought attention to the processor inside and fame to Intel. Although the campaign was abandoned in ~ 2013, the impact on the consumers buying preference remains clear today. 

Brand based marketing strategies have the greatest abstraction and require the greatest faith. Examples of these are rampant in sports marketing campaigns. Campaigns that trade on celebrity such as having the NIKE swish on Tiger Woods ball cap, or advertising Viagra on a NASCAR. These campaigns are common but seem detached from prompting action from the consumer.  The ability to measure the impact of advertising campaigns have raged on for decades as marketing professionals try to convince the Chief Financial Officers of their company that watching a NASCAR will assist in greater sales of Viagra.  As the ability to quantify even the causal impact of a campaign cannot be substantiated the argument quickly turns to rhetoric. 

The puzzle of advertising is how can so much be spent with such an imprecise measurement of its value? Although media companies throughout time have tried to give a measurable algorithm taking the form of everything from circulation audits to digital analytics, in the end, they were only able to measure outreach and the overall impact. This has left much of the decision making to that of intuition, a mere belief or faith. Although outreach can be measured out to the individual level along with purchases, tying the two together to measure cause and effect has been elusive. All too often the intuition that believes that all it would take is a small conversation to justify the money creates an emotional response that is based on a "feelings". This is more often the case with digital campaigns were impressions are a fraction of a cent and can be justified with even the smallest return.

Contrasting this with extremely large and expensive campaigns that reflect a "hail Mary" approach can be witnessed with ads taking place during the NFL Super Bowl.   The cost of a 30-second spot will exceed $5 million with a cost per impression being at ~ $0.024. This boom or bust mentality reaches ~103,000,000 viewers and would need to return more than $0.05 per view to break even.  Trying to calculate the Standard Deviation of sales on the individual level is complicated by the relatively noisy economic conditions and the subtle nuances of the campaign.   So even as the size of the campaign grows to more than a million viewers, the impact of the campaign may still be elusive.  

As it is believed that most advertisers are driven by their calculations to determine ROI, the ability to calculate the profits generated through advertising as a percentage of cost seems to support the broadest audience. Our goal was to eliminate from our calculations indicators used by many as distractions and focus specifically on profits. It was clear that determining the ROI with any accuracy would require a sample size that was greater than 7 figures. It was recognized that -100% ROI or the ad didn't perform to the ad was profitable with a > 0 ROI would be crucial.  This number would be key in comparing it to the internal cost of capital and convince the CFO that the campaign was a success. 

Once the ROI is determined for a campaign a much bigger challenge in determining the optimal amount to spend. The approach that Google campaigns take is to convince you that a small spend will contribute to your success. And using the modest advertising spend, they attempt to convince you that any positive return can be attributed to the campaign and thus can justify an even greater spend.  Trying to determine the optimal amount to spend during a campaign is a question that stymies’ many as there is the belief that there is in essence a “perpetual motion machine” at which point the ROI is sufficient to continually fund the campaign. 

One of the major issues in determining the shortcoming of a campaign are the specific attributes of the campaign. Consider this simple example, an ad cost $0.05 per delivery. The one-time campaign has a marginal profit conversion of $20. This would mean that 1 in 4000 people would need to be sold for the ad to break even. But suppose the purchase probability was to have an increased baseline of 10%. The effect would be 400 times larger for the same ad. 

Other examples might include campaigns with a finite audience. Many clients face a situation where their product a so niche that the ability to use broad-based marketing is not practical. This was witnessed recently when we were promoting an event in Salt Lake City, Utah. The relatively small concentration of Embedded Engineers in the total population of the city required us to achieve a 10x higher rate of return to achieve the desired audience size. In this case or focused shifted from a cost per thousand to a need for a campaign with a greater impact.  

Using the empirical sales volatility observed and the magnitude of impact that variables have on the equation, the ability to create an unbiased answer requires a precision that is not available. The observational methods, the potential for omitted control variables or biases will generate results that cannot be relied on. Adding to this lack of precision is the need to understand the implications of frequency. It is assumed in calculations that there is an endless supply of independent consumers to add to the campaign.   In reviewing the research of others who had conducted 25 field experiments with campaigns with 500,000 to more than 1,000,000 unique users it was concluded that the median standard of an error on ROI was a stagger 51%. What this said was that supposing the ad was profitable (>0 ROI), 9 of the 25 campaigns lacked sufficient power to reject -100% ROI.  In further calculations, many campaigns could not reject the null. Only 3 of the 25 campaigns could reliably distinguish between wildly successful (+50%) from those that broke even (0% ROI). It was determined that the campaigns would have to be 100 times larger than the median campaign to dispel this concern. 

The implications of not being able to quantify the ROI have caused ad agencies to rely heavily on their reputations and provides a false perception of larger publishers who can carve out a monopoly on “reliable” feedback. To create a simple model that would consider and optimize a range of activities that resulted in the use of various media and the interactions, the study focused on a campaign as a set of advertisements delivered to a select set of consumers through a single medium, over a specific and relatively short period.  In pursuit of our goal of determining the rate of return for a specific expenditure and delivery of ads and ultimately the optimum rate of spend, it was assumed a campaign is executed through a single publishing channel.

In trying to determine the amount to spend in advertising to each consumer, it is difficult to take into consideration the impact that variables have on the algorithm. Considerations such as seasonality, time of the day, customer preferences, length of the campaign are a few of the many concerns associated with a campaign that has a significant impact on determining the optimum campaign spend. Our experience, after 30 years of managing campaigns, is that most advertisers evaluate their spending at the campaign level. They will determine their next spending from their intuition of the first.

Measuring the Campaign Cost and Return

The true value of a campaign for example of a display ad in a publication, the cost can be measured per reader at an average cost $0.04 - $0.08 per impression. Per reader costs for digital campaigns are less expensive, with the average ranging from a fraction of a cent to high of $0.03. Given the total volume of ads a consumer receives across the many mediums, the ability to capture a person's attention during a single campaign is less than 3%.  

In moving on to quantify sales volatility we adopted three components: the average magnitude (mean sales), heterogeneity (variance of per-person means), and the rarity of purchases (stochasticity in purchasing). Throughout the embedded market these components will vary widely from company to company. Companies varied based on the average sale price of their product, the commodity nature of their product, and the design lock achieved. Those that were targeting large and lucrative design wins within the defense market are found to have long-term clients with a sustainable relationship that sometimes lasts decades. Those with lower-cost solutions that are easily displaced, with one-time buys, bracket the volatilities faced. 

There are times when external factors influence measurements such as gross margins and volume. These can be macroeconomics, pandemics, and seasonality to name a few. As a result, creating a control group that is not part of the campaign, but which is made up of the same quantity of individuals with the same demographic will enable a campaign to be more accurately measured.  

Estimating an ROI

Although ROI for advertising spends is impacted by several variables, being able to use an understanding of their application will enhance your ability to spend money wisely. To understand that these variables will create a level of uncertainty does not obviate the Marketing Professional from applying their best in determining the course of action. 

In contrasting one campaign with another, the calculus of audience size in contrast to their interest level will substantially contribute to changing the baseline of calculating the ROI.   If the audience returns a 10% greater likelihood in sales than another, the calculation of costs at the campaign level will be substantially lessened by the quality of the audience. Digital campaigns are subject to the largest variance as the rate of delivery, open, and follow-through are contrasted by the magnitude of the list and the cost of delivery.

The subjective nature of reaching the right audience with a frequency complementary to their buying habits and having the required impact to prompt action is the goal of marketing professionals. To believe that there is a formula that will create a "perpetual marketing machine" where every dollar spent will result in more than a dollar in margin reflects a limited understanding. And yet to throw up your hands and believe that the biggest list at the lowest price is the goal is to ignore the true understanding of marketing.