Marketing ops · Production

Burberry harnesses Labelbox and Databricks to curate strategic marketing assets

The problem

Burberry works with high-volume unstructured image datasets to train object detection and classification models for global marketing campaigns, but manual labeling was not feasible at scale, and the open-source annotation tool they tried lacked proper data-source integration, required local storage, and was too slow.

First attempt

Burberry first tried an open-source image annotation tool that had serious drawbacks: it lacked integration with their data sources, required images to be stored locally on a data scientist's machine, and could only save progress in JSON files.

Workflow diagram · grounded in source
1
Image import from S3
integration
“The lightbulb moment was when we saw how easily we could import our images to Labelbox from our Amazon S3 bucket via API”
2
Image annotation for model training
ai_action
“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”
3
Stakeholder image upload to CRANE
trigger
“Marketing users can now upload candidate marketing images into our proprietary in-house-built Content Ranking Engine (CRANE) app”
4
Engagement prediction by channel
ai_action
“score them against pre-trained models that predict the engagement these images will get in upcoming campaigns by channel and by region”
5
Ranked insights delivered
output
“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”
6
Revenue tracking feedback loop
feedback_loop
“tracking the revenues generated by the emails they send. This information influences each batch of imagery we produce and leads to continuous improvement in our marketing programs”
Reported outcome

With Labelbox and Databricks, image annotation projects that previously took months now take hours, insight generation dropped from two months to a two-hour self-service process, 10 headcount were saved versus building in-house, time savings for generating insights improved by 70%, and Burberry has achieved four years of continually decreasing total cost of ownership.

Reported metrics
Headcount saved vs. building in-house10 head count saved
Time savings for generating insights from images70%
years of continually decreasing TCO with Databricks4 years
Image annotation project durationmonths to hours
Show all 5 reported metrics
headcount saved vs. building in-house10 head count saved
time savings for generating insights from images70%
years of continually decreasing TCO with Databricks4 years
image annotation project durationmonths to hours
insight generation timetwo months to two hours
Reported stack
LabelboxDatabricks Lakehouse PlatformAmazon S3CRANEDatabricks
Source
https://labelbox.com/customers/burberry-customer-story
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

With Labelbox and Databricks, image annotation projects that previously took months now take hours, insight generation dropped from two months to a two-hour self-service process, 10 headcount were saved versus buildin…

What tools did this team use?

Labelbox, Databricks Lakehouse Platform, Amazon S3, CRANE, Databricks.

What results were reported?

Headcount saved vs. building in-house: 10 head count saved; Time savings for generating insights from images: 70%; years of continually decreasing TCO with Databricks: 4 years; Image annotation project duration: months to hours (source-reported, not independently verified).

What failed first in this deployment?

Burberry first tried an open-source image annotation tool that had serious drawbacks: it lacked integration with their data sources, required images to be stored locally on a data scientist's machine, and could only s…

How is this marketing ops AI workflow structured?

Image import from S3 → Image annotation for model training → Stakeholder image upload to CRANE → Engagement prediction by channel → Ranked insights delivered → Revenue tracking feedback loop.