Between highly publicized shifts in media viewership, increasing media competition and fragmentation, and higher simultaneous media usage, marketers are more confused than ever about their media decisions. Chief executives are also being held to a higher standard of accountability for their choices, and passing these pressures down the food chain to chief marketing officers and other senior marketing executives. Many are being asked to implement more while spending less.
Marketers are increasingly relying on their agencies for analytic expertise to better understand their customer base and maximize the value from their media campaigns. The demand has pushed agencies to beef up their analytics expertise from what in years past was a handful of statisticians, to what is now a dedicated and marketable core competency highlighted in every client pitch.
Analytics departments, staffed with a dozen or more people with MBAs and Ph.D.s. in math, statistics, econometrics, physics, and artificial intelligence, no longer simply translate data from the client's last campaign. Instead, they are considered the "go-to" department for understanding the market before concepts for creative are developed and campaigns are planned.
At Carat Interactive, for example, "At the beginning of each creative engagement...we'll typically get invited into the meeting and say, 'Here's what a high-value customer looks like,' and 'Here's what we should be communicating to get the best impact,'" says Barry Peters, vice president of data intelligence, who was charged with turning Carat's expertise into a full-fledged data intelligence center a year ago.
New measurement opportunities
While many agencies analyze the effectiveness of both on- and offline media, they are finding informative new measurement capabilities online. One area especially affected by this trend is the way in which return on investment (ROI) is defined. "We're looking at more sophisticated models where we measure the number of page views that somebody reads when they get to a Web site and how that translates into conversion," says Gerard Broussard, senior partner and director of media analytics at mOne Worldwide.
"For a pharmaceutical client, for instance," he adds, "a lead generation study showed that redistributing advertising impressions away from people who were getting too many, and toward people who were getting too few, improved response by 15 percent strictly through reallocation, without spending any additional money."
Overall, "there are four categories clients want to understand today," explains Dave Hauser, vice president and director of analytics at MPG: "What the ROI is, when they can expect to get it, the build and decay effect, [or] how much they should build into it, and at what point advertising starts to [lose its effectiveness]. And budget level...what's the minimum necessary spending. Everything generally falls into those four camps, including what media they spend their money on, outside marketing issues, and price elasticity."
The bottom line, of course, is still about moving the needle. For instance, Peters says, "In the old world, if the data said that product 'A' was driving a higher conversion rate than product 'B,' we would simply serve more banners of 'A.' Now, with backend site-side analytics, we can see that, yes, 'A' is driving a higher conversion rate, but [it] is driving more returning customers to the site, while product 'B' is driving new customers. So let's promote product 'A' through direct mail and e-mail to existing customers, and use product 'B' online as an acquisition tool."
Even brand marketing is becoming more trackable and definable online. "When we first started working with vendors like Dynamic Logic, the first thing we would measure was whether or not online advertising was impacting the brand," explains Terry Cohen, vice president/director of analytics, research, and measurement at Digitas. "We then expanded to looking at what features and attributes drive the brand. So where it used to be 'yes' or 'no,' now it's, 'Is it rich media?' [and] 'Is it ad sequences?' It's more about how do I allocate my funding to really drive the brands."
Rich media measurement
Rich media is taking online brand measurement to new levels. Says Cohen, "We're able to set up tracking that will tell us 30 seconds in, are they still online? Sixty seconds in, are they still online? And what's the associated level of engagement with the final outcome? "Then, we can set up tests that will show, 'What is the impact on the brand if they see two rich, followed by one traditional, or two traditional followed by one rich?'" "We don't need to do as much testing," adds Kathleen Sheridan, senior vice president and leader of the digital marketing practice at Digitas. For example, "For one client, we developed a variety of different search landing pages. But rather than test them all, we tested a few and used the results to do predictive modeling on which of the remaining test versions would work best."
These capabilities have highlighted some flaws in traditional online methods, such as ad server data, which emphasize the need for human expertise to analyze the data that are automatically collected. "Let's say you saw an ad on Yahoo! and again two days later on MSN, but on the third day you finally clicked on CNET," explains Gordy Abel, director of marketing at Carat Interactive. "Ad-serving technology would give the credit to CNET." Through in-depth analysis, "we're able to see that it took all three [to convert you]. So now that we've got that finding, we'll do a test and serve a single ad on CNET and see if that garners a click."
Adds AKQA's Bensen, who's based in San Francisco: "When New York started to catch up with what we were doing on the West Coast, for about a year the TV mentality clients kept saying 'I need to know the GRPs [gross rating points]. But nobody remembered how small the online GRP was, and if we'd used the GRP definition, it would never have been acceptable. So we're still educating people on looking at the measurements that work for online."
Going forward, one of the main issues in Web analytics is understanding the distinction between background and foreground media. "The concept itself is not new," says Forrester's Nail. "We've known forever that people get up from the TV and go to the bathroom or talk to someone else in the room."
What has changed, Nail says, is that "now, multi-tasking and fragmentation have reached a point where a major prime-time network TV show is lucky to get an 8 rating as opposed to years ago, when anything less than a 20 rating would be canceled. So you're getting a small[er] audience to start with, plus the audience isn't really getting the message, because people are skipping through ads. So marketers are saying there has to be a better way."
"I think there's a need to understand the role of foreground and background media and the synergies between different channels," adds Coleen Kuehn, executive vice president/chief catalyst strategist, MPG. "Putting them all in a single currency and understanding their effects and how they work with each other is key."
A New Business Model
When Beth Fleming got laid off from Grey Advertising in San Francisco, it didn't take her long to get back on her feet. Her phone began ringing that same day with calls from agencies that needed experienced analytics personnel. Since the offers were for project work rather than full-time jobs, Fleming decided to freelance.
Today, she does freelance analytics consulting for AKQA, Publicis & Hal Riney, ClearInk, Kadium, and Butler, Shine & Stern, and works on accounts that include Hewlett-Packard, Visa, Palm, TIAA-CREF, and Philips.
"I believe my business model works because most agencies do not have enough business to keep a senior-level analyst on staff full-time," Fleming says. "Day-to-day reporting is often handled by junior employees with one to two years of work experience. You can always throw bodies against a workload, but it takes someone with more experience to problem-solve larger process issues."
As an example, "many agencies get stuck in a current process, such as copying and pasting an updated media plan from a new Excel sheet to another that houses the formulas. This is time-consuming and is prone to manual errors," she says. "For one agency, I pulled everything directly from the database and automated all of the calculations, saving about two days off report production time and eliminating manual errors."
Christine Bensen, director of media and analytics at AKQA, hired Fleming to take over some responsibilities when Bensen's job description expanded to include media. "She's been working with my analysts to understand what types of reports they require, and in some cases, refining those reports to be even more beneficial," Bensen says. However, not everyone in the business is as enthused about the freelance model. "It's an interesting business model," says Dave Hauser, vice president/director of analytics at MPG. "But doing projections on an ad hoc basis is not in the best interest of the client. You need to build a long-term relationship."
Still, Fleming wouldn't have it any other way. "Another benefit is getting away from the politics and incessant meetings that can negatively impact the work," she says. "When you feel internal pressure to have every campaign be a success, it's difficult to maintain integrity. I can produce better work as a freelancer because I can concentrate on the truth of the numbers without as much concern for someone 'killing the messenger.'"
Marketing Mix Modeling Is the Wave of the Future
The most promising thing I've seen [in analytics today] is marketing mix modeling," says Jim Nail, principal analyst of Forrester Research. "It's not new, but it's gotten more sophisticated. Ten years ago, most companies were looking at the marketing activities of a single brand. Today, they're looking across an entire portfolio of brands. For example, if you're Tide and you've got Tide with bleach and Tide for cold water and Tide with Downy, what happens if we advertise the mother brand only? How does it affect the other brands?"
Marketing mix modeling can often run into competing objectives, cautions MPG's Hauser. "When there are sister brands, each brand manager wants to sell more of his own brand. And if we optimize sales for one it might sacrifice the health of the other." The retail banking division of JPMorganChase can relate to that problem. Once a fan of media mix modeling - the science of analyzing a company or product's entire media portfolio to understand how the media interact with each other and affect final sales or conversions - the division stopped using the technique more than a year ago due to what Mike Eichorst, vice president of database mining and predictive modeling, says was an inability to "trust" the calculations.
The problem was, outside consultants who supplied the models usually wouldn't explain their results for fear of giving away their strategy. That made it hard for marketers to prove to brand managers why the models made sense. "They didn't want to lift up the hood and show us how the whole thing worked," Eichorst says. "So when all the managers responsible for how the money was spent came to us and said, 'Do you vouch for this?,' we couldn't say 'yes,' because we couldn't understand it."
Meanwhile, "the guy buying TV doesn't want to give up his money and the guy buying print doesn't want to give up his, and it becomes an internal debate. People felt uncomfortable following the recommendations of the model without an understanding of the intuitive underlying drivers." Furthermore, Eichorst says, even if the model's recommendations were followed, "there was no way to validate that a lift [in sales] was a direct result of the model, because there were always other factors involved."
In essence, experiences like Eichorst's drove Marketing Management Analytics Inc. (MMA) - an independent operating unit of Carat North America - to launch Avista, a hosted analytics solution that puts clients in control and allows them to do their own "what if" analyses. Avista, which competes with a product from Veridiem, assists with optimization, benchmarking, and portfolio management. It automatically defines ROI. "Some of the challenges of mix modeling are that by the time you get the results, they're often late," explains Doug Brooks, director of product development, marketing and user experience at MMA. "This [tool] gives you the power of having information throughout the year. It helps move companies from measurement of ROI to management of ROI on a continuous basis, and they make better decisions throughout the year as a result." Avista was scheduled to go live with two clients on July 1.
Best Practices in Online Media Measurement
- Compare page-view counts against conversion rates to understand the threshold in page views needed to get an online visitor to convert.
- Separate your advertising response results by existing and new customers, then optimize your media for each group separately.
- Study response by number of impressions and gauge how many impressions is enough vs. too much. You may be able to achieve higher ROI by adjusting your ad allocation without having to increase the budget.
- Look beyond traditional brand awareness impact to uncover specific advertising features and attributes that drive the brand. For instance, is it rich media? Or is the sequence the ads are being shown in?
- Shorten landing-page testing time by taking the results from your first few tests and using them to build a predictive model of which remaining test versions would work best.
- When studying ad server data, make sure you look at all the underlying data in addition to the reports. For example, if a consumer saw the same brand advertised on three different sites but clicked on the ad on the third site, the ad server technology would assign the credit to the third site, when in fact the clickthrough could be a result of the customer's seeing all three.
- Understand the distinction between background and foreground media and use it to your marketing advantage.