How New Data Science Measures Advertising Success Across ‘Swim Lanes’
Typically, advertisers have measured the success of their online ad campaigns by the performance of each individual ad. How many clicks did it generate? How many conversions? Each advertising channel has operated under its own set of metrics, formulas and objectives.
In the world of advertising, such practices are no longer tolerable.
Failing to recognise the advances made in the field of data science will leave stodgy traditionalists in the dust. The rise of Big Data has given advertisers unprecedented insight into how the entirety of their advertising campaign exists within its environment and how their customers interact with their ads. Specifically, data science has shown that measuring the performance of individual advertisements in their own ‘swim lanes’ or silos provides an inaccurate picture of the success of a given ad.
For example, a Google search can lead to a YouTube video view, which can result in the clicking of a banner ad two days later. New developments in data science allow advertisers to measure the success of their ad campaigns across silos. Compared to traditional methods of measurement, these comprehensive analytics prove much more effective at developing an accurate picture of customer behaviour, showing advertisers which elements of their ad campaigns are responsible for overall success or failure and where to allocate their ad spending.
Advertising executives need to do two things in order to avoid falling behind: hire the right people and commit to the right process.
Online review giant Yelp sets an example for companies looking to hire the right people. An article in the Wall Street Journal features a data scientist—officially a Ph.D. in physics—who used one of his dissertation algorithms “to measure the effect on consumers when multiple small changes are made to online ads” on Yelp.com. Though the bridge between academia and online advertising might seem like a long one, the rise of data science in online advertising means that personnel with unique skill sets are actually in high demand. Cognitive scientists and breast cancer researchers have also recently made the move into commercial data science.
Such valuable labor does not come without cost. According to Ad Age, employees in the data science sector earn salaries of about 30% more than employees with similar experience in other areas of the advertising business. Is it worth it? Well, look at the data: despite high costs, the salaries for marketing analysts are expected to jump between 2.9 and 3 percent in 2015. That means demand is still increasing.
Data science specialists command such high rates because they have the unique ability to sort millions of data points into actionable reports. Simply put, there are no longer any single metrics for success. In order to decipher what advertising data means, ad teams need a code-breaker.
Getting the right people, then, is absolutely critical to success. But that won’t work unless the right process is instituted. Having someone brilliant to analyze data means nothing if the data collection process does not accurately reflect the environment of an advertiser’s potential customers. Advertising executives must understand the way that their audience behaves today in order to know what information to seek.
That’s not an easy task. Data hierarchy used to be much more clear-cut than it is now. As Matthew Keylock, Global Head of Data for dunnhumby, writes in AdExchanger, the customer must now be viewed as “a horizontal dimension across the enterprise” rather than a singular touch point. To properly mine data, research must account for the myriad ways in which a single customer can interact with a single business.
The ability to view ad campaigns across channels gives advertisers better insight into which parts of their advertising budget generate the best return on investment. At first, this seems counterintuitive: how does viewing the project as a whole allow advertisers to get a better look at each individual part?
The answer soon becomes apparent: it is impossible to understand a single result completely without understanding the environment that has produced it. In an article for the Harvard Business Review, MarketShare CEO Wes Nichols describes a client who switched from looking at individual swim-lane results to adopting a broader scope for evaluating ad performance. In this case, a change in process resulted in a nine percent increase in sales. Advertisers seeking similar results should look at their own process for collecting and evaluating data. Does it come from one narrow environment, or is it the result of a comprehensive, cross-channel data gathering process?
It’s that type of knowledge that makes data science in the online advertising industry more difficult—and more valuable—than ever before. Studying ad performance in isolation might be ineffective, but that does not mean it is impossible to correctly evaluate the performance of individual advertisements. Data scientists who study online advertising must merely account for the context of individual ads when evaluating their effectiveness.
In turn, this evaluation provides advertisers with valuable insight that teaches them where they should increase spending—and where they should stop throwing their money.
In a joint study between New York University and Google, NYU business professor Nikolay Archak provides a relevant example. He compares the effectiveness of two types of keywords—retailer-specific and brand-specific—and how their performance relates to website traffic and, somewhat transitively, revenue generated. According to the example, “Individual analysis would suggest that brand-specific keywords provide significantly worse return-on-investment.”
To many advertisers, such results would indicate a need to reallocate their investment towards retailer-specific keywords. “Yet it would not be wise,” Archak later continues, “for the retailer to significantly cut spending on the brand-specific keywords as it is likely to reduce inflow of users to the retailer-specific keywords as well.” In other words, digital advertising results cannot be viewed in their own vacuum or ‘swim lane.’ A proper understanding of how ads interact online is crucial to successful budgeting. Advertisers gain this understanding by implementing complex algorithms that account for these interactions—algorithms often developed and improved by data scientists.
Digital advertising requires the application of data science because its environment has reached a level of complexity only understood by a unique breed of specialists. It’s why sociologists are being interviewed for high-paying corporate positions. New personnel have arrived armed with the understanding of a complex customer environment and the knowledge that it’s not possible to correctly evaluate ad performance piece by piece any longer. They have become indispensable in an industry where advertising executives with comprehensive processes outshine those who are stuck in individual ‘swim lanes.’
Written by Miklos Zoltan Mattyasovszky.