Wednesday, February 19, 2014

Big Data zeros in on ad inefficiency

John Wanamaker, the innovative Philadelphia merchant who pioneered the modern department store in the Gilded Age, was such a fan of newspapers that he is credited with buying the first full-page ad.  

But even Wanamaker knew that the most efficient form of advertising available in the 1890s wasn’t terribly efficient at all. “Half the money I spend on advertising is wasted,” he is famously reported to have said. “The trouble is, I don’t know which half.”

Nowadays, Wanamaker could find out, with considerable precision, by hiring Applied Predictive Technologies (APT), a Virginia-based company that mines all manner of data to determine not only the optimum ways to buy advertising but also where to locate bank branches and which under-performing entrees to nix from restaurant menus. 

In fact, as we will see in a moment, a modern-day retailer did hire APT to scrutinize the efficacy of its ad expenditures.  While publishers won’t be thrilled with the results, the study is a valuable lesson for media companies in the power of Big Data to either support – or subvert – their businesses. 

First, the background: A privately held company, APT received a hefty $100 million in equity capital last year from Goldman Sachs, making it one of the biggest players in the world of predictive analytics, the practice of sifting mounds of Big Data for patterns that help companies make money, save money or, ideally, do both at the same time.

With customers like Walmart, Lowe’s, Office Depot, PetSmart, CVS, Target, Walgreen’s and many other global merchants, APT asserts that it captures and crunches 20% of data generated in the “U.S. retail economy.” It bounces this rich transactional data against everything from weather records to Twitter streams to help companies “measure the profit impact of pricing, marketing, merchandising, operations and capital initiatives.”  

Given the roughly $14 billion that national and local brands spend annually on newspaper advertising in the United States, it was only a matter of time before one of them asked APT to answer the question that vexed John Wanamaker: Which advertising dollars are being wasted?

In a white paper published here, APT reports that it ran the numbers for an unidentified national “big-box” retailer with a $100 million advertising budget. “On average,” APT stated, “newspaper advertising did not pay back,” unless the merchant had saturated a market with a large number of stores.  “In markets with low presence, the cost per store far exceeded the incremental marginal dollars created by the [ad] circular,” said APT. “Removing underperforming markets eliminated an additional $5 million in waste from the marketing budget, while having minimal impact on revenue.”

Because that “waste” represents publisher revenues, the problem for local media companies is obvious. If this sort of analysis catches on widely – and it’s likely that it will – then it will play havoc with the business models of local publishers and broadcasters. Armed with better data than ever about the efficiency of their ads, marketers are bound to either bargain for lower rates or cut spending. Or both. 

The consolation for local media companies at the moment is that only the largest merchants have the sophistication, resources and motivation to employ services like APT. But this technology – like all technology – will get faster, better, cheaper and become widely available in the fullness of time. As the awareness and adoption of predictive analytics inevitably ramps up, here is how local media will be threatened: 

Companies like Google, Apple, Facebook and Amazon – and dozens of others – are investing heavily in capturing as much data as they can from mobile devices, digital media consumption, social activities and online purchases. The current and future advertising, subscription and commerce businesses being pursued by the tech powerhouses depend on obtaining the maximum amount of actionable intelligence from as many individuals as possible, including who they are, how much money they make, where they live, who they know, that they are reading, where they are going, what they have bought, which videos they have uploaded, what they are shopping for and – most precious of all – how to generate more cash by influencing their future behavior. 

One way the big tech companies can capture more data is by offering cheap or free analytical services to the Main Street businesses that are the core clientele for every local media company. The businesses get cool new marketing tools and the techies get tons of additional data. 

Local media companies can defend against this threat – and build strong businesses for themselves in the future – by getting ahead of the tech behemoths. In other words, they need to establish themselves as savvy digital marketing sherpas before the Big Data guys get there. 

Because we are in the early days of Big Data, there is time for local media companies to get up to speed. But they have little time to lose.  

© 2014 Editor & Publisher

2 Comments:

Blogger As I see it said...

It sounds like these people get their "big data" primarily from heavy internet users, so I wouldn't be surprised to hear that they may have missed some equally big "non-internet" data from traditional sources like customer surveys, matching product sales in brick and mortar stores with specific advertising campaigns, etc. Can you enlighten us about any internet bias for these services?

2:22 PM  
Blogger Bill Seliger said...

I'd be very surprised if they were doing this with only online sales (I can't think of very many multichannel merchants that are driving a significant percentage of online sales; most newspaper circulars seem to be driving in-store sales). My guess is that most of the data is point-of-sale.

I read the white paper at the jump and it was interesting - this is their methodology "Test and Learn for Ads™ directly addresses this obstacle by measuring differences in sales patterns in stores and markets receiving support in real time against a portfolio of tailored holdout stores or markets that remain dark for a given campaign. As opposed to static point-in-time models, advertisers maintain dynamic, ongoing testing programs supported by Test & Learn for Ads™ and can accurately and effectively determine the incremental returns generated by various channels and campaigns, reading through broader macroeconomic changes and marketplace noise. "

It sounds as if they are A/B testing markets and even products (which is not new for multi-channel merchants - they've been doing that for years) and correlating aggregate sales or even sales of specific products at specific stores.

It would be relatively easy to perform this analysis yourself in R for free (but of course you wouldn't get a fancy SAAS webpage on which to view your data).

7:16 PM  

Post a Comment

<< Home