AdvancedAnalytics
Trending project
AdvancedAnalytics projects with the most verified people and teams
Featured project
Arla Video Intelligence - A Recipe for Relevance
Digital marketing is under disruption. As an industry, we are forced to reinvent the way we have executed digital marketing for more than 25 years. Consumers' demand for relevance and personalized experiences is greater than ever and at the same time data privacy regulations are stricter than ever and we have probably only seen the tip of the iceberg in this area. Very soon the tracking cookie, as we know it, is put out of commission and with the cookie buried the whole ad industry is facing radical transformation at a pace where the brand aspiration to deliver relevance at scale to consumers seems like nothing more than an ancient ad utopia. The data-rich companies with impressive CRM solutions and loads of deterministic data are still keeping up the faith. But what do you do when you are not an airline or a bank with thousands and thousands of permissions? When you are a classic FMCG company like Arla, selling Milk and Butter? With no first-party data to provide real-time consumer insights. Whatsoever. Completely reliant on third party data that is now out the window. In Arla, we decided to take up the challenge and show the industry the way forward. Our longstanding ambition did not change: We want to provide relevant messaging to every consumer in any given touchpoint across all our markets and, needless to say, in a privacy-compliant manner. However, to be able to deliver relevance at scale, we continuously need to understand consumers' needs, triggers, and moods. But with no behavioral data from cookies to lean on and activate against how do we then interpret consumers' state of mind to serve them the most relevant message in any given situation? Facing a cookieless world we needed to come up with new ways of identifying consumer signals to serve the right message. Luckily, marketing automation technologies have come to a more mature state in the lifecycle, so naturally, we looked towards machine learning and advanced analytics tools to play a prominent role in our future setup. Basically, we were looking for the perfect recipe for relevance in the new Digital Era. Background Digital marketing is under disruption. As in industry we are forced to reinvent the way we have executed digital marketing for more than 25 years. Consumers demand for relevance and personalized experiences is greater than ever and at the same time data privacy regulations are stricter than ever and we have probably only seen the tip of the iceberg in this area. Very soon the tracking cookie as we know it is put out of commission and with the cookie buried the whole ad industry is facing radical transformation at a pace where the brand aspiration to deliver relevance at scale to consumers seems like nothing more than an ancient ad utopia. The data rich companies with impressive CRM solutions and loads of deterministic data are still keeping up the faith. But what do you do when you are not an airline or a bank with thousands and thousands of permissions? When you are a classic FMCG company like Arla, selling Milk and Butter? With no first party data to provide real time consumer insights. Whatsoever. Completely reliant on third party data that is now out the window. In Arla we decided to take up the challenge and show the industry the way forward. Our longstanding ambition did not change: We want to provide relevant messaging to every consumer in any given touchpoint across all our markets and, needless to say, in a privacy-compliant manner. However, to be able to deliver relevance at scale, we continuously need to understand consumers needs, triggers and moods. But with no behavioural data from cookies to lean on and activate against how do we then interpret consumers state of mind to serve them the most relevant message in any given situation? Facing a cookieless world we needed to come up with new ways of identifying consumer signals to serve the right message. Luckily, marketing automation technologies have come to a more mature state in the lifecycle, so naturally we looked towards machine learning and advanced analytics tools to play the prominent role in our future setup. Basically, we were looking for the perfect recipe for relevance in the new Digital Era. Solution To fulfill our ambition we applied a Netflix data mindset to our content. With 34 brands across more than 50 markets we have loads of content and campaign data. Like Netflix uses their data to make that perfect new series, we wanted to use our data and create our own recipe for relevance – an Arla developed predictive model for food inspiration telling us how to make the most inspiring food content, what consumers to serve it to and when. With a fully connected Martech stack across markets, including an Arla owned DSP, we utilized our partnership with Google and then applied advanced machine learning to find patterns across all assets produced within the last year. We constructed a data model, that automatically provides us with a frame to frame information about all the videos we produce. The model is successfully able to analyze and identify objects like “human” and specific types of “food”, but also advanced and abstract elements like certain “flavors” and “emotions” with a nanosecond precision in the videos. This gave us priceless insights into the content elements and cues in the most engaging videos, but also balancing other parameters impacting the performance of an asset, like the media channel, time of the week, weather, seasonality, etc. However, we still needed the most important part of the equation linking the right assets to the right audience at the right time in real-time media activation. Together with Google, we identified 840 global unified audiences. These audiences are not cookie-dependent, and are continuously updated by Google giving us a very robust pool of audiences. With the video detection data in place, we then combined the video data with the audience data from Google to understand the connection between the audiences and the video content. Now we have a fully automated Video Intelligence solution using advanced video detection machine learning that incorporates more than 2.000 variables (elements, gestures, sentiments, etc.) and combines that with performance data from 840 audiences. The Video Intelligence solution makes us able to predict how any new video will perform in a given audience and provides us with very valuable and detailed insights on both content and consumers. The best part of this? The results are impressive. Result The Arla Video Intelligence has not only delivered more relevance, but it has also delivered more scale and reach. In 2019 we gained a total of 15,7 million extra video views and 105 million extra impressions from our campaigns just by delivering more relevant content to the consumers – in Denmark alone. In actual performance numbers, we have seen a 24% increase in the amount of video viewed when using the Video Intelligence solution. Because of the relevant match between audience and video content, we got rewarded by algorithms with a better impression price which resulted in a 19% better cost per impression. That corresponds to 8,6 Million DKK saved or reinvested in more effective marketing! To get a more nuanced perspective on the relevance improvement delivered by the Arla Video Intelligence solution, we decided to also have a look on the consumer journey after watching our videos. It showed that consumers spent an average of 14% more time on our website after being presented to the video content amounting to 2.700 hours spent extra on Arla.dk. Facing a cookieless world we truly believe, that we have cracked the future code to relevance at scale, and we are scaling Arla Video Intelligence to UK, Sweden and the Middle East during 2020. In UK, Sweden and the Middle East combined we look at up to 40 Million DKK in media savings or possible reinvestment if we deliver same performance optimization as in DK. That’s huge numbers. It still feels like we have just started the journey with improving and optimizing our efforts through marketing automation, machine learning and advanced analytics. In the Video Intelligence solution there are many other data sources to bring into play in the near future via data partnerships with eg. retailers, brands or companies outside the category that can improve our results even more. Furthermore, our brand marketers and insight teams can benefit immensely from the data we capture to bring even more and better insights into the marketing and commercial planning cycle. This being our first real stab at marketing automation we are really excited about what the future holds, because we see so many possibilities to put automation into play that will benefit Arla in the years to come.
Highlighted brands
