The promise of digital marketing and what we're able to measure has evolved greatly in the last decade. We spoke to Sasha Grujicic, chief strategy officer at Dentsu Aegis Network, about what he's learned, his approach to measurement, and how Dentsu tackles planning in our new(est) digital age.

Published
November 2016
Topics

In the early days of digital, marketers had all kinds of ideas about what was possible and what would work for their clients. A lot has changed since then—including our expectations. Here are Grujicic's thoughts on where we are and where it’s all headed.

You've said that figuring out how digital works and how to measure it is a hard problem to solve. Why do you think that is?

Personally speaking, I think the easy jobs of digital have all been taken care of: communications, content, networks, and sales. These are all societal (economic and social) facets that digital is naturally suited to create solutions for.

The hard task comes when you try to digitize the rest of it: stores, supply chains, labour, and customer service. These are harder aspects to digitize, but the potential outcomes for those that do it are enormous.

Once we are able to digitize these, we will be able to much more accurately measure and optimize the holistic costs of selling goods and services to people. This will broaden our understanding of what digital is—and what it can do—to close the loop of measurement to generate a true sense of an ROI. In short, we will see the whole picture of how a business operates and goes to market through the lens of digital and data. We can then measure and optimize our investments accordingly.

"I think a lot of the digital metrics we see today are indicators but are wildly open to different interpretations."

Until then, we need to use approximations like media mix modelling, attribution modelling, and machine learning. We can do a lot right now but often it's our gaps in understanding the inner workings of the non-digitized economy that lead us astray. This is why the problem is hard.

What digital metrics do you pay attention to and which ones do you think matter?

I think it really depends on the problem you're trying to solve. If it's sales, then it's sales. If it's opinion, then it's opinion measures. I think a lot of the digital metrics we see today are indicators, but they are wildly open to different interpretations. Just because I watched something, it doesn’t mean that I'm always indicating to you that I'm interested in purchasing something. I could have actually watched it for any number of reasons: I was bored, I like the music, or I want to buy something similar.

What matters is not the single measure but how that measure lives in context. Where we're getting quite good as an industry is starting to stitch together a context where the likelihood of that indicator—me being interested in buying something—is much more likely than not. I'm most excited and scared about the contextual layer of metrics that we can now overlay on top of specific metrics. I mean, what a time in our history to be a data analyst!

You're now advocating moving from channel/screen-based planning to audience-based planning. How does this work at the Dentsu Aegis Network?

In short, the media selection is secondary; it's the audience that matters. We build our attitudinal and behavioural segments through first- and third-party data in our proprietary system, CCS (Consumer Connection System). We then quantify the size and reach we can achieve within that digital universe.

"In one case the ROI from high-impact digital and online video was 7X traditional broadcast...The implications from that study resulted in a large shift into digital as it warranted much more investment."

Next, we augment that reach with broadcast media bought through our connected buying methodology (i.e., not demographic-based buying) to ensure we don't lose the segmentation-level nuances we have just developed through our planning process. Once we're in market, we calibrate the plan to maximize its impact.

Tell us about the media mix modelling study you recently ran with Google for one of your Canadian clients.

[[callout-headline-1]]

We ran a few recently and both came out with some very interesting findings. In one case the ROI from high-impact digital and online video was 7X traditional broadcast.1 This was a result of the solid performance of these digital channels, the underspend in the medium, and the saturation of that traditional buy. The implications from that study resulted in a large shift into digital as it warranted much more investment.

How might studies like these change Canadian marketers' perceptions of what works and how they make decisions?

We need to legitimize the economic impacts behind the investment. We coasted too long on the consumer data we see from usage to justify investing at the levels we do. Like I said at the YouTube Pulse event, we messed up the first go-round.

We thought that we could translate the monumental effects of the internet at a one-to-one ratio to our client's businesses—like every product was going to start a revolution or movement. However, the Arab Spring and selling airline tickets are two very different things. We just got really excited about the prospects of what the digital revolution could do for our clients, and we wanted them to win in this new context.

"The #1 thing I would recommend is to 'listen' to the data before defining what you're going to evaluate for."

To be clear, it was never malicious. I wanted it to happen for them. I wanted our clients to experience a popular uprising that helped them grow their business. It just didn't happen that way. This is why we need to study the effects of digital on different industries and use those as a baseline in the way we invest going forward.

Last question: What’s the #1 thing you think advertisers can do to rethink or get more out of their marketing measurement?

Only one?! Well that’s not fair.

I guess the #1 thing I would recommend is to "listen" to the data before defining what you're going to evaluate for. Take the data you have access to, develop some baseline assumptions, try to incorporate some contextual data points (weather, economic data, news data, etc.), and look for patterns. Vary your statistical techniques and try to discover new things about your business. From there you'll be able to generate some working assumptions and insights before you go live with an investment. This will give you a much better read on what to expect from your campaigns and how to evaluate them.

[[inline-image-1]]

Source
1 Google/Data2Decisions, "MMM Study," August 2016, Canada.