voyager624 - Fotolia
Unilever, the multibillion-dollar multinational consumer goods company, does a lot of advertising.
Owning over 400 brands, including Dove, Lipton, Axe, and Ben and Jerry's, Unilever regularly processes massive amounts of consumer and social media data to help understand their customers and create personalized marketing campaigns.
To power its campaigns, Unilever turned to Google Cloud APIs, using natural language processing tools and AI for marketing.
Vision and Natural Language
"We have a broad repertoire of products that consumers love to eat, to use, and to clean with," explained Alex Owens, VP and global head of data and analytics at Unilever.
"My job is to get insights on people that use our products, to ensure that we can deliver the products that meet their needs, as well as get us the data that we have on consumers and that we collect with their consent, to power a real-time marketing capability," Owens said.
Due to its size and reach, Unilever has a special partnership with Google, Owens said. The company gets top-level support and enjoys early access to many Google products.
One that Unilever relies heavily on for its AI for marketing efforts is the Google Cloud Vision API, a product that, according to Google's webpage for it, provides easily modifiable pre-trained machine learning models that can analyze objects, images and text.
"What that does is it allows you to decode images that you might have on the likes of Instagram," Owens said. This essentially enables Unilever to gather massive amounts of metadata from social media posts to help power its AI for marketing campaigns.
For example, Owens continued, Unilever ran an advertising campaign on social media for the toothpaste brand Close-Up in Southeast Asia around Valentine's Day. Unilever wanted to target younger people, so, using the Cloud Vision API, the company analyzed hashtags and social media content from Southeast Asia.
"We realized by looking at the search analytics that the second most popular thing on Google around Valentine's Day is learning how to kiss. So, thankfully, we developed a campaign around the art of kissing," Owens said. Unilever released short videos on Instagram, Facebook and YouTube that millions saw.
Analyzing social media
Unilever also uses the Google Cloud Vision API, along with the Cloud Natural Language API, to help ensure that marketing efforts adhere to the cultural norms of the many "culturally diverse" areas Unilever advertises in, Owens said.
The Vision API can decode all the user-generated content across social media, Owens said, while the Natural Language API analyzes user comments.
According to Google, the Natural Language API can perform analysis and annotation on text, including sentiment analysis, which can help decode emotion; entity analysis, which can help discern what the text refers to; and syntactic analysis, which can help determine the makeup of a sentence.
Combined, the two APIs "allow you to basically summarize the themes as the sentiment of how people are talking about your advertising real time," Owens said. This enables Unilever to determine a campaign's effectiveness and make changes quickly.
In another example, Unilever drew on its vast data sources and used the APIs to promptly identify a defective product based on user comments and swiftly made changes to the product, Owens said.
Owens said while Google's tools work well for Unilever, and the relationship with the tech giant is important, he had some trouble initially getting the APIs to work as needed.
For example, it took a while to properly train the Vision API machine learning models, Owens explained.
"We played around a lot with the capability with Google to help refine it,”" Owens said. That's like any AI tool, he said, but still, it took time.
Unilever also faced challenges with the Google's translation technology. Years ago, the technology wasn't as accurate as it is now, Owens said. Internal teams had to tweak the translation and adapt the tools to meet their needs.
"It probably didn't cover the languages that maybe we would want, or at least have the coverage within those languages to translate what we needed," he explained.
Being such a large organization, Unilever has the financial resources and software talent to test new technologies and features before deploying them. When the technologies are released, they are already workable and scalable because they were tested. Unilever also pairs third-party technologies with Unilever's own platforms. All this ensures smoother rollouts of new features, Owens said.
Unilever also uses marketing tools from Adobe for Unilever's AI for marketing efforts, which Owens said pairs well with Google's tools.
With Google Cloud, "we support data-driven marketing to extract the insights that we need. And [to] ensure Unilever continues to be relevant to people that consumer our brands," Owens said.