INside Performance Marketing
Top 50 Industry Players from 2017
Machine Learning Algorithm Lituus to Offer “Unparalleled Accuracy” for Advertisers

Machine Learning Algorithm Lituus to Offer “Unparalleled Accuracy” for Advertisers

PerformanceIN

Lituus combines machine learning in the planning and implementation stage of mobile advertising campaigns.

Mobile data platform Ogury has launched its new machine learning algorithm, Lituus in a bid to enhance its targeting capabilities for brands and marketers.

The new solution employs machine learning in the planning and implementation of mobile advertising campaigns, delivering audience targeting of unparalleled accuracy at scale.

“Ogury’s new targeting solution offers higher performance at a bigger scale and with increased transparency,” said Christophe Thibault, chief algorithm officer at Ogury.

‘Unparalleled accuracy’

According to Lituus, the algorithm works in three parts - learning, optimising and sharing.

The first part starts by learning criteria at the start of an advertising campaign - such as the brand website data - to gain insights from users’ qualified traffic.

Lituus comes into play by learning from the qualified traffic of the chosen websites and identifying both the discriminant and similar attributes amongst users. It then builds a targeting matrix composed of hundreds of different criteria.

Optimising involves the algorithm inputting and learning throughout the duration of the campaign to constantly exclude the lower performing criteria and upweight impressions from the higher performing criteria.

It also takes into account conditions such as optimal time of day, publishers, device models, connectivity, and localisation. This is all achieved using Ogury’s device level first-party behavioural data, which provides insight into users’ apps installed, apps usage and mobile web browsing.

At the end of the campaign, a ‘targeting performance’ section is included within the campaign report, offering a comprehensive view of the best and worst performing combinations of targeting criteria, websites visited, apps owned and apps used by the audience of the campaigns. This provides the customer with a transparent data set that is based on actual learnings from the campaign, as opposed to the ‘black box’ data collection approach that is currently widely used in the industry. These learnings can then be applied to the next campaign and for strategic audience planning.

“In the first applications of this new machine learning approach we have observed up to 50% drops in user bounce rate in cost per click campaigns, and up to 16% increases in video completion rates in cost per view campaigns over human targeting alone.” added Thibault.

Continue the conversation

Got a question or comment – tweet PI @performancein or comment on Twitter, Facebook or LinkedIN.

PI Newsdesk

PI Newsdesk

The PI Newsdesk brings you quick, day-to-day updates from the performance marketing industry, helping you keep an eye on the competition, whether that's product launches, key partner wins and more.

Read more from PI

You may also like…