Analytics & Attribution in the Multiscreenverse

notes for a panel at Admonsters Screens

These are the notes I’ve made for a panel that I’m on today at the Admonsters Screens conference. There’s not really a flow here (I’m hoping that will emerge during the panel), but some of the links are good.


People have never confined their activities neatly to one channel, and it’s almost impossible to imagine any smartphone or tablet user watching TV without that device open on their laps. It’s only the niceties of social etiquette that keeps me from using mine at the cinema. So – even though this is how we all talk – it’s ridiculous to think about “online audiences”, “TV audiences” and “mobile audiences”.

  • 90% of people move between devices to accomplish a goal (browsing, shopping, managing finances, planning a trip).
  • 77% of viewers watching TV with another device in hand (49% with a smartphone, 34% with a laptop)

(Google, ‘The New Multi-screen World: Understanding Cross-Platform Consumer Behavior’, August 2012)

Caveats & Terms

Most of what I’m interested in is how FMCG advertisers navigate the multi-screen environment. This means that it’s a lot harder to talk about “attribution” – which is really the domain of direct advertising and performance marketing. So I’m probably misusing my terms slightly.

However, I think that there’s more commonality than might immediately strike the eye; and I hope that some cross-fertilisation of ideas might actually be productive.


Rather than simply buying seconds and using spots to promote social media campaigns, visits to Facebook pages or rallies to Tweet a branded hashtag (brandtag), think about it as a way to tell a story that can live beyond the spot or beyond the campaign.

(Brian Solis, ‘The future of TV is more than social, it’s a multi-screen experience that needs design’, May 2012)

The good news for advertisers seems to be that the more easily distracted people are – well, the easier they are to distract. Here are some highlights from a recent combined observational study/survey:

  • 73% of all qualitative study participants agreed that having other devices with them while watching TV shows made them less likely to fast forward through ads
  • Viewers 21–36% more likely to cite auditory over visual attributes of various ad executions as “attention-grabbing” (i.e. if their eyes are occupied, grab their ears.)
  • 45% of multi-screeners in the quantitative survey reported that they’re more likely to remember brands if they see ads on more than one screen versus just TV alone.

(Bravo / Latitude, ‘Deconstructing the Multi-Screener’, November 2012)

 Example: Coca Cola Polar Bowl (2012)

(see: The Coca-Cola Polar Bowl – Game Day Moments)

  • Coca Cola estimated that 60% of the Super Bowl’s 111 million viewers would be using a second screen during the game.

“We needed to make sure our idea tapped into that and took advantage of every single screen on game day,” said Jennifer Healan, Coca-Cola’s group director-integrated marketing content. “But the idea needed to be additive to the experience.”

Experiment judged a “resounding success”, with 9 million consumers across various platforms checking in on what the polar bears were up to.

(Ad Age, Coca-Cola Polar Bowl Engaged 9 Million People, May 2012)


Avinash Kaushik identifies 3 kinds of Multi-Channel Attribution challenges:

  • Multi-Channel Attribution, Online to Store
  • Multi-Channel Attribution, Across Digital Channels
  • Multi-Channel Attribution, Across Multiple Screens

(Avinash Kaushik, ‘Multi-Channel Attribution: Definitions, Models and a Reality Check’, April 2012)

There’s a tendency to get obsessed by the data, but the data we have may only represent 5% of the picture. Not everything can be tracked with a cookie, click or logged in profile, but toys tools like Google’s flow visualisation can fool us into thinking that we have the whole picture.

In fact, flow visualisation highlights a key conceptual problem: there’s an instinctive tendency for marketers to think in linear terms (the user starts here, travels through a neat funnel, and ends up at the purchase.) We’re in danger of trying to adapt to the multi-screen environment by just replacing our old linear model with multiple linear models. One for search, one for social, one for TV, one for email. And yet, the whole concept of simultaneous multi-screening makes a nonsense of this (yes, it was already nonsense, but we could avoid thinking about it.)

What might a solution look like?

Recognising that we need to move away from linear models helps us understand where a solution might lie: we need less counting, and more statistical models.**

Our maths needs to get better. One profitable route explored by our analytics team has been Structural Equation Modelling – which allows them to model multiple simultaneous and multidirectional relationships between discrete pools of marketing activity.

Please tell me what you think.