Why is this work relevant for Media?

Galvanising the often overlooked channel of programmatic display, this campaign for Oroton leveraged technology, data and machine learning to extract incremental value for the business by aligning media placement with context, audiences and creative message.

Background

Oroton is a well-known luxury brand in Australia. As a brand, they leverage programmatic display extensively as part of an always-on layer of activity. Display has proven effective in generating sales and assists other channels, such as search and social, in their acquisition efforts. A previous attribution study for Oroton showed that users who interacted with both social and display, saw a lift in conversion rate of 53%. Despite these impressive results, years of creative optimisation and steadily improving media metrics, last year performance began to plateau. People were not engaging more with our messages and we were losing the cut through we usually had after a seasonal creative refresh. In short, we were becoming less noticeable. The challenge: How can we increase performance of the digital acquisition program and particularly improve our engagement with message and cost per acquisition, with no additional budget?

Describe the creative idea / insights (30% of vote)

We conducted an analysis of our existing display banners to understand what elements drove performance increases. To ensure our findings were actionable, we looked beyond known drivers such as creative changes and application of best practices (simple visuals, consistent assets used across all channels, clear brand logo and call to action). This led us to uncover the improvement of viewability rate (from 60% in 2019 to 70% in 2021) as a key contributor to increases in performance. Looking deeper in the data, we also noticed that the time in-view for each message and format had an impact on performance format with a longer time in view recording better engagement (click-through rate was 30% higher on average). Our insight was clear: the longer we could hold the attention of the user, the more likely they were to engage with our brand (and our performance would improve as a result).

Describe the strategy (20% of vote)

We wanted main campaign parameters to remain the same: Similar tactics (prospecting and remarketing), same target audiences (Female interested in fashion apparel and accessories) and the same creative message. This way we would be able to establish a clear causality between new optimisation introduced and campaign performance. Our strategy for the campaign was to take into consideration not only if the creative had an opportunity to been seen (viewability) but what kind of attention the user paid to it. Using this data, we could then optimise the campaign toward placements and contexts where the creative message would have a greater attention time. To be able to optimise toward attention, we needed a technology which would use real eye tracking while also being privacy compliant and scalable across all our digital inventory. Understanding the attention paid to each placement and creative message would feed our optimisation algorithm.

Describe the execution (20% of vote)

We leveraged a technology which tracks user real eye tracking from an opt-in group of over 7,000 people. Along with the eye gaze data, it collects 40 other non-visual data signals. This data was fed into an AI platform over 100,000 times, training the model to understand when consumers were looking at ads without needing a camera. The model gets trained regularly to ensure accuracy over time. This technology was developed by PlaygroundXYZ and called Attention Intelligence platform. This solution enabled us to track and measure the Attention Time on our ads at both the publisher, placement, and creative level. From these insights, we made optimisations based on stronger Attention Time signals. To evaluate the effectiveness of this approach and remove potential seasonality bias, we set-up an A/B test, splitting the campaign in 2 identical implementations. One will only measure attention time, the other will optimise toward greater attention time.

List the results (30% of vote)

In the face of a challenging task – optimising performance with no additional budget – our attention led approach delivered in spades. We’ve delivered attention to the tune of 60% (vs the control group). Optimisations in Attention Time also led to a lift in engagement (clicks on banners) of up to 31% compared to the control group, with an overall lift of 17%. This engagement had a direct impact on our cost per acquisition. From a business impact point of view, it resulted in +25% lift in conversions and -20% Cost Per Acquisition when optimised toward higher attention time. The introduction of attention time as the main proxy of optimisation has also led to an increased effectiveness with -27% in Cost per Acquisition Month on Month.