Cannes Lions

Get Closer with Closeup - First use of Machine Learning Google Vision API

MINDSHARE ASIA PACIFIC, Singapore / UNILEVER / 2018

Presentation Image
Case Film

Overview

Entries

Credits

Overview

Description

Data insights

The brand leveraged cross-platform insights to win in moments that mattered most during Valentine’s Day. The brand leveraged data sources like - DoubleClick, Brand Watch, MOAT and Google Cloud Vision API

Innovation

CloseUp was the first-brand powered by Google’s AI solution Cloud Vision API. The brand used a bespoke app built using the API to look at open-source images from social media and discover visual insights on what’s trending in a specific country. The brand wanted to supercharge it’s creatives by being relevant across moments that are trending through the day. The Vision API gave insights that text-based analytics could not offer. i.e. the brand was able to read an image and break it down to landmarks, faces, emotions, objects & colors. It recognized Taj Mahal trending in India with location labels, parks & outdoors trending in Brazil, people in images indexing on emotions like “joy” when paired with flowers & desserts

Originality

The brand leveraged 10 Big Insights to build six-second real-time videos through Google’s Technology that allowed for customization at scale (opening sequence + closeup move #n ). The brand started with 4 base videos to 132 at campaign completion

Execution

• Data gathering: Data gathering was set up in our Loop Room where data from consumer insights and conversations, Google Trends, Google Cloud Vision were leveraged to inform creative design and video generation

Step 1: Overall campaign optimization across six markets and multiple ad groups - we shifted budgets towards what was delivering on campaign KPIs

Step 2: Through social listening corroborated with Google Vision API we uncovered insights like “single on Valentine’s” as huge conversation territories

Step 3: To win in these new, relevant moments, we built new creatives on the spot across markets at scale through Google’s Vogon Video platform

Step 4: In less than 8 hours since step 1 new videos were set live as part of the campaign, totaling 132 videos from 4 original assets

Data interpretation example:

Keyword "when is valentines day" has mass across platforms - gleaned from our programmatic reporting platform Ultra AI

People are talking about “single on valentines” - gleaned from CMI Unilever People Data center brandwatch dashboard

Analyzing images with valentines day #tags people are taking more “outdoor photos”, uncovering what “objects” drove high correlation with emotions such as “joy” - gleaned from first time use of Google Vision API

Outcome

Outstanding results across the six markets:

With millions of impressions across video and social channels, allowing us to reach an impressive 50% of the internet population in the six markets with the relevant dynamic brand message

AdRecall and Brand interest lift across the board +85% in Indonesia, +238% in Vietnam, +813% in Brazil, +56% in Philippines, +60% in South Africa, and +53% in India.

Creative resonated higher on 18-24 year olds highlighting the impact of a younger generation eager to embrace the Valentine’s celebration, help achieve brand KPI

Exceeded reach targets per market by more than 50%; 85% delivery on Mobile

Viewability rates were closely optimized and beat benchmarks with 70-90% completion rates. New assets created mid-campaign outperformed the first phase assets with higher completion rates

Lower CPM resulted in expanding reach for each market and 20% savings

The campaign was also supported by amplifying similar assets on open websites where dating apps were some of the top performing sites for video delivery relevant to consumer intent

The campaign helped spike organic searches for brand terms across Google and YouTube.

Similar Campaigns

12 items

Confessions

UNILEVER EGYPT, Cairo

Confessions

2020, UNILEVER

(opens in a new tab)