Google Mobile Facebook

Connecting the dots in the cross-screen era: How marketers can navigate mobile targeting


By Ronan Shields, Digital Editor


Promoted article

September 23, 2015 | 8 min read

Sponsored by:

What's this?

Sponsored content is created for and in partnership with an advertiser and produced by the Drum Studios team.

Find out more

Targeting users in a mobile environment is a constant challenge for marketers evaluating the pros and cons of the two main approaches to targeting.

The UK is now a ‘smartphone society’ according to Ofcom’s 2015 Communications Market Report, which recorded that 33 per cent of internet users see their smartphone as the most important device for accessing the internet.

Online audiences’ migration to mobile devices presents a myriad of problems for all parties in the media business as targeting ads towards mobile users is more challenging compared to a desktop environment.

This is due to the fact that cookies – the primary means of targeting ads toward internet users in the desktop era – don’t work as effectively on mobile devices. This is fuelled by the increasing number of ‘connected screens’ (think smartphones, tablets, and connected TVs) meaning it is now rare for internet users to begin researching a purchase, and then converting on one device.

Most advertisers are aware of the importance of making sure their ad campaigns take place across a number of screens, and then being able to attribute value back to each media placement. However, the difficulty lies in coordinating the messaging across screens in a manner that will tell the brand story, to the correct audience, and in a coherent manner.

Currently two models prevail in the industry, either through deterministic or probabilistic targeting (see box out).

Prevailing model?

Research from eMarketer claims that Facebook and Google dominate mobile advertising, with the pair commanding a joint market share of 68.5 per cent last year (21.7 per cent and 46.8 per cent respectively).

This figure suggests that advertisers are opting to target users employing the deterministic model.

First party benefits

Facebook, which relaunched its Atlas ad stack last year allowing advertisers to target users across third-party properties using the Facebook ID, with some suggesting its wealth of data could enable it to rival Google’s DoubleClick stack.

Steve Webb, head of Atlas, EMEA, claims the strength of its offering is derived from the insights gleaned from Facebook’s 1.49 billion users. He adds: “Through Atlas, marketers are able to target and measure their digital ad campaigns with people-based data rather than cookie-based data – providing high-quality inventory with attribution, targeting and audience reach.”

However, Webb rebuts assertions that Facebook, or Atlas, is a ‘walled garden’, highlighting the outfit’s partner enablement strategy, where it works closely with companies such as Marin Software and Kenshoo.

He adds: “The Atlas measurement tag can also be run in conjunction with most existing ad-serving tags, giving critical reach, frequency and conversion insights across devices and publishers.”

The Drum requested an interview with Google, but it declined to participate. However, speaking previously about the potential threat posed by Facebook’s Atlas, Darragh Daunt, Google DoubleClick’s head of platform sales, independent agencies, said: “It’s a question of graph size [ie how many logged-in users] and we would argue that our graph size across all screens is extremely formidable. If you think about [the fact that] Android penetration is 80 per cent in the market… and if you think of the people that use YouTube and Gmail and Google Now, hopefully, that is a formidable graph.”

Aside from targeting users across screens, a key element of the process is attribution (i.e. being able to assign value to every ad placement, and how it contributed towards a conversion, or sale), and key to this is the passing back of device IDs to the advertiser paying for the placement.

However, Facebook fell foul of advertisers when it proposed to stop doing so recently, so much so that it eventually performed a U-turn on this policy.

On this matter, Webb comments: “We encourage our advertisers to apply people-based measurement solutions so they can determine whom they’re reaching, not just where they’re reaching them. But we want to give advertisers the freedom to measure ads based on what is important to them. Facebook will therefore continue to give advertisers the option to receive device-level reporting from our mobile measurement partners for mobile app ads.”

Deterministic targeting

This model uses logged-in user data (or first party data) to identify users across screens via anonymised IDs provided by the specific service provider. The online advertising industry’s leading proponents of this model are Amazon, Apple, Facebook, Google, Twitter, etc.

All of which are commonly referred to as ‘walled garden’ ecosystems, as these IDs must be used within their respective ecosystems when it comes to targeting audiences. Accuracy rates are said to be typically over 90 per cent, but because the unique IDs are exclusive to each of the walled gardens, this can lead to the fragmentation of campaign execution.

Probabilistic targeting

This model uses a number of anonymised data points, such as device-type, location data and browsing history, etc, in order to recognise a user across screens. The online advertising industry is brimming with advocates of this model, some of the leading providers include Adbrain, Drawbridge, Moat and Tapad.

Each claims they can provide an open framework for advertisers to identify and reach users across multiple platforms and properties (ie they are not restricted to using this ID within a single ‘walled garden’). Accuracy rates vary – figures range from 60 per cent to 97 per cent. However, advertisers have to invest a great amount of effort in auditing the separate methodologies.

The case for open

However, as stipulated above, this is not to say that using such closed ecosystems, or walled gardens, is perfect. For instance, this can fragment the execution of a campaign, potentially leading to bad practice, such as bombarding a web user with the same ad, due to poor visibility on how many times a single web user has been served.

Caspar Schlickum, chief executive EMEA for GroupM ad tech hub Xaxis, says: “The idea of agnostically targeting users across devices is a good one. The probabilistic model is good for the market.”

He goes on to argue that a fundamental problem with walled gardens is that they benefit the provider ultimately, and that this can come at the expense of the consumer experience.

“We have to be able to frequency cap, so as we don’t annoy some poor consumer,” according to Schlickum. He adds that while Xaxis does work with companies like Google so as to prevent doing so, it is far from ideal.

Paul Gubbins, head of programmatic sales EMEA, for Millennial Media, explains how his company helps advertisers target users across screens, and then trace those ads back to a sale, or other action such as a website visit. This is performed by identifying household IP addresses, and then creating a device graph for mobile devices that are consistently present in that IP.

He adds: “The first steps to a successful cross-screen attribution strategy can be simply deployed by adding a mobile tracking pixel to the brand’s desktop site. If the object is to track visitation, this would be on the landing page.”

However, he does acknowledge the trouble facing advertisers is that the lack of a single standard for device mapping means no single vendor or technology can track every user on every device all the time.

As a result advertisers are faced with the task of deciding between the precision of a deterministic strategy, or the potential scale posed by probabilistic players.

Data leakage?

Although Gubbins warns how brands opting for the former model, may be giving away a key competitive advantage by inadvertently donating its customer insights to such third party providers (a phenomenon widely known as ‘data leakage’).

“Platforms that operate within a walled garden are happy to ingest first-party data but rarely enable the extraction of post-impression insights that allow chief marketing officers to inform other parts of their marketing strategies.”

This point is also echoed by Gareth Davies, chief exec of Adbrain, who adds: “The accuracy of first party data is good, but brands shouldn’t expose their data too much, and chief marketing officers should ultimately have full control. What they need is an independent solution.”

Hence brands must ask themselves: how much can a closed ecosystem open them up to competitors gaining their insights?

This feature was first published in the 16 September issue of The Drum.

Google Mobile Facebook

More from Google

View all


Industry insights

View all
Add your own content +