Burberry uses thousands of images from multiple sources for its global marketing campaigns. The company must classify these valuable assets precisely so that it can use them effectively in its campaigns to drive the right action by the right audience. Burberry works with a high-volume of unstructured data sets to train their object detection and classification models. They quickly realized it wasn’t possible to identify and label all these images manually and needed their image recognition models to identify highly specific products and apparel SKUs.
Burberry implemented Labelbox’s data engine within its Databricks Lakehouse Platform environment. Image annotation projects that used to take months now take just hours, and marketing team members now have access to powerful content insights without asking for help from data scientists. Databricks’ Lakehouse Platform continues to give Burberry a total lower cost of ownership than the company’s original business case projected.
10 head count saved by implementing Labelbox rather than building in-house. 70% improvement in time savings for generating insights from images. 4 years of continually decreasing total cost of ownership (TCO) with Databricks.
Originally published from the Databricks website. Source: https://www.databricks.com/customers/burberry
Burberry is a British luxury brand, headquartered in London. As they market the company’s daring apparel, Burberry’s stakeholders search for the exact combinations of images that will capture the imagination, command attention and win sales. Seeking an efficient way to annotate its thousands of highly specific marketing assets, Burberry implemented Labelbox within its Databricks Lakehouse Platform environment. Image annotation projects that used to take months now take hours, and marketing team members now have access to powerful content insights without asking for help from data scientists. Meanwhile, Databricks Lakehouse Platform continues to give Burberry a total lower cost of ownership than the company’s original business case projected.
Curating data threatens to become a marketing bottleneck
With a catalog of bold, high-fashion apparel, Burberry uses thousands of images from multiple sources in its marketing campaigns. The company must classify these valuable assets precisely so that it can use them effectively in its campaigns to drive the right action by the right audience.
“We work with high-volume unstructured data sets to train our object detection and classification models,” explained Harry Collings, Data Science Manager at Burberry. “It’s simply not feasible to identify and label all those images manually. And because our image recognition models must identify highly specific products, we eventually realized we needed a dedicated solution that could work precisely while taking the manual effort out of the process.”
To that end, Burberry recently began looking for a solution that would provide the technology to improve the data for training its models quickly and easily. The company hoped to be able to produce labels for thousands of images and place them seamlessly into a model development pipeline where they could be conveniently reused. Burberry first tried using an open source tool designed for image annotation. The application worked, but with serious drawbacks.
“Our open source tool lacked a proper interface that could be read from our data sources,” Collings recalled. “All the images had to be stored locally on a data scientist’s machine, which was not ideal for a company of our size. To label, say, 1,000 images would take someone two hours — and they could only save their progress in a JSON file. We knew it was time to look for a more robust commercial solution.”
Labelbox and Databricks deliver insights that drive smarter marketing decisions
As Burberry evaluated potential image annotation solutions, the company favored platforms that would integrate well with its Databricks implementation. Burberry was already experiencing the benefits of having a single, unified analytics platform for all its stakeholders.
“When we first went live on Databricks Lakehouse Platform, our collaboration skyrocketed among data engineers, analysts and business users,” said Ben Halsall, Analytics Technology Director at Burberry. “Teams could easily interact with each other through notebooks and share ownership of jobs. With the lakehouse, we broke free from the limitations of traditional data warehouses. Suddenly, our analysts could take any data set — structured or unstructured — and start deriving business value from it right away.”
Burberry began evaluating Labelbox in part because of its tight integration with Databricks. The company soon discovered that Labelbox offered a far more customer-centric purchase process than its competitors.
“The Labelbox team immediately spoke to us in terms that were relevant to luxury retail like they had really done their homework,” said Halsall. “The fact that Labelbox understood our needs so well made us very eager to spin up a proof of concept. We were very impressed with the responsiveness and support they provided early on.”
Getting up and running with Labelbox was easy for Burberry — and the solution quickly proved it could meet Burberry’s needs. “The lightbulb moment was when we saw how easily we could import our images to Labelbox from our Amazon S3 bucket via API,” reported Collings.
Once Burberry had officially selected Labelbox, Halsall and his team connected Labelbox to the company’s core virtual private cloud and S3. They then built a smaller integration that delivers most of their master assets into S3. The team now treats images just like any other data set within its Databricks Lakehouse Platform. The solution was up and running within a month for DS team members — and following internal demonstrations, business and marketing users have also been onboarded expediently.
“Thanks to Labelbox and Databricks, stakeholders across our business can upload images and put them against pretrained models that help them predict what kind of engagement these images will get in upcoming campaigns, based on models of previous campaigns,” explained Collings. “Stakeholders can see a ranking of the images they’ve uploaded and gain insights that help them make better decisions for the current marketing campaign.”
Generating two months of insights in just two hours
Before Burberry implemented Labelbox, generating insights on images was a task that took two months and involved hands-on work by the data science team. Today, it’s a two-hour self-service process on Labelbox — and faster time to insight is translating to better decisions. Marketing users can now upload candidate marketing images into our proprietary in-house-built Content Ranking Engine (CRANE) app and score them against pre-trained models that predict the engagement these images will get in upcoming campaigns by channel and by region. Stakeholders can see a ranking of the images they’re considering using and gain insights that help them make better decisions on upcoming campaigns and to inform creative briefs.
“We’re not only helping our stakeholders choose the best images to use in campaigns but also tracking the revenues generated by the emails they send,” said Collings. “This information influences each batch of imagery we produce and leads to continuous improvement in our marketing programs.”
Without Labelbox, Burberry simply wouldn’t have had the in-house resources to build a product that could annotate a massive volume of images so efficiently while finding edge cases using Labelbox's Model product that boost model performance. This is accomplished by having an interface to find and fix label errors in the data that will most impact model results. “I would have had to ask for a 10-person team just to build a product that would let us view all our imagery, let alone generate insights on it,” Collings remarked. “In our day-to-day workflow, Labelbox saves us countless hours while delivering a tremendous amount of value.”
Meanwhile, Burberry’s Databricks Lakehouse Platform continues to exceed the company’s expectations. When Burberry originally built a business case for Databricks against its previous cloud platform, the company expected comparable performance. Since then, Databricks has released its Photon engine, cluster optimizations and other enhancements.
“When we look at performance over a five-year period, the relentless product road map of technology innovation in Databricks Lakehouse Platform has meant our cloud costs are actually decreasing, despite growing our user community,” said Halsall. “We achieved parity in year one and have reduced the total cost of ownership since then while still providing an improved experience for our engineers, analysts and data scientists. Databricks allows us to take on any analytical challenge with confidence.”
As Burberry continues to ramp up its marketing programs and ingest further data, the company will rely on Databricks Unity Catalog to help it keep data secure and comply with data privacy regulations. Burberry also appreciates that Databricks and Labelbox price their solutions aligned to business value.
“Databricks and Labelbox charge us based on consumption of resources rather than number of users,” concluded Collings. “That model aligns more closely with the value we’re getting from these solutions and gives us a better idea of what to expect as we grow. We’re extremely confident in the partners we’ve chosen to take us forward.”
“With Databricks Lakehouse Platform and Labelbox, we’re able to realize our full ambition of creatively led, data-driven content creation. Labelbox continually gives our team insights on how to optimize our campaigns. And Databricks helps us maintain an efficient, sector-leading analytics capability in the cloud.”
– Ben Halsall, Analytics Technology Director, Burberry