How Under Armour uses data, AI and 3D in product creation and consumer insight
Under Armour has described social listening, connected-shoe data and 3D printing in the past, and now cites AI mainly as a risk. What is documented, and what is not.
KEY TAKEAWAYS Summary by the editors
- At a 2018 analytics conference, Under Armour's vice president of enterprise transformation described using natural language processing in SAP HANA on social media data to inform a 3D-printed shoe, instead of asking consumers directly.
- The same talk said connected shoes revealed that customers often used cross-training shoes for running, an example of product-usage data feeding product decisions.
- Under Armour ranked its data by trust: SAP ERP data highest, then data it pays to acquire, and social media data lowest, which is used with caution.
- Under Armour's annual report for the year ended 31 March 2026 says the company is upgrading its end-to-end planning technology system, and describes AI-enabled shopping tools and AI governance as risks.
- The filings reviewed do not describe AI in design or 3D product creation, so current use of generative design or 3D tools at Under Armour is not publicly documented in them.
Under Armour's public record on data and product creation is split between an older, detailed talk and newer, cautious filings. In 2018 an executive explained how social listening and connected-product data informed a 3D-printed shoe. In 2025 and 2026 annual reports, AI appears as a competitive and governance risk, alongside an upgrade of the company's planning technology. No source reviewed documents current AI or 3D design workflows.
This case study reports both layers and flags the gap between them, so readers do not assume that a 2018 example describes today's operations.
How did Under Armour use data to inform a 3D-printed shoe?
ASUG reported on a 2018 BI and Analytics conference session in which David Roberts, then vice president of enterprise transformation at Under Armour, described a shoe with a 3D-printed midsole sold directly to consumers, made after an online order. He contrasted this with the conventional model, in which most shoes sold in the United States are made in Asia on a design-to-sale cycle of about six months that he said had been the same for decades.
For design inputs, the team analysed social media with natural language processing in SAP HANA rather than seeking feedback directly from consumers. Roberts also said Under Armour had acquired several fitness tracking app companies and connected them to its shoes to gather internet-of-things data, and that one insight was that customers often use cross-training shoes for running.
How much did Under Armour trust each data source?
A useful detail in the talk is the data hierarchy. According to ASUG, Roberts ranked SAP ERP data as most trusted, then data the company pays to acquire, with social media data least trusted and used with caution. He said, "Technology gets you to the table, but you have to decide how to break the data down."
| Source | Use described | Trust level stated |
|---|---|---|
| SAP ERP data | Core operations and reporting | Highest |
| Acquired data | Added to internal sources | Second |
| Connected shoe and app data | Insight into how products are used | Not ranked |
| Social media | Natural language processing to inform design | Lowest, used with caution |
The ASUG article also gives cost figures for 3D-printed versus factory-made shoes that are internally inconsistent in its own text, so they are not repeated here. The point for planners is the sourcing model, on-demand production after an order, not the unit economics quoted.
What do Under Armour's recent annual reports say about AI?
The Form 10-K for the year ended 31 March 2025 says Under Armour's inventory systems, including its global operating and financial reporting system, are designed to improve forecasting and supply planning. It lists the use of data analytics, artificial intelligence, machine learning and the digital consumer experience among areas where competitors may have an advantage.
The filing for the year ended 31 March 2026 refers to upgrading the end-to-end planning technology system, to the need to adapt to the use of AI-enabled shopping tools, and to the need to optimise available consumer data. It also cites controls over the use of AI and machine learning technologies, and the European Union AI Act, among compliance risks. On product creation it says only that products are developed by internal product development teams.
What is and is not documented about AI and 3D in design?
- Documented: 2018 use of natural language processing on social data and connected-product data to inform a shoe.
- Documented: a planning technology upgrade and AI named as a competitive and governance matter in recent filings.
- Not documented in these sources: generative design, AI trend forecasting, 3D apparel sampling or digital product creation at present scale.
- Not documented: measured effects on lead time, sample counts or hit rate.
Lessons for product teams. First, signal quality matters more than signal volume. Under Armour openly ranked social data as the least trustworthy input and used it with caution. Second, product-usage data can reveal behaviour that surveys miss, such as shoes used for a different activity than intended. Third, planning technology upgrades, even without AI, are a prerequisite for any later use of machine learning on forecasts.
Why do dated case studies need careful handling?
Under Armour's 2018 example is still widely quoted because it is specific, with named tools and a clear sequence from data to product. But five or more years is long in analytics. Natural language processing methods, social data access and 3D tools have all changed, and Under Armour's own business has been restructured since. Quoting the example as current practice would be misleading.
Its value today is as a pattern. The company combined three kinds of signal: public conversation, device data from connected products and internal transaction data. It then applied different levels of trust to each. That structure, a hierarchy of evidence, is as relevant to a generative AI tool summarising customer feedback as it was to the original analysis.
How could a product team apply the same approach with current tools?
- List the signals. Social listening, reviews, returns reasons, product usage data and sales history.
- Rank them. Decide in advance which sources can override which, as Under Armour did informally.
- Test before relying. Compare insights with sales outcomes from past seasons to see which sources predicted well.
- Keep a human decision point. Designers and merchandisers should review any AI-generated insight before it shapes a brief.
Any such programme also needs rules about consumer data. Social media analysis and connected product data fall under privacy law in many markets, and the annual reports reviewed list data protection and AI regulation among the company's compliance risks.
What does on-demand production change for planning?
A made-to-order product reverses the usual planning logic. Instead of forecasting demand months ahead and committing to quantities, the brand waits for an order and produces afterwards. That removes forecast error for the item, but it shifts the challenge to production capacity, lead time and cost per unit. The 2018 talk presented the shoe as an alternative to the conventional cycle, not as a replacement for it, and the company's current filings still describe forecasting and supply planning as central.
For most apparel brands, on-demand production remains a niche because garments are labour intensive and materials vary. The relevant lesson is the principle that data from early customers can steer what is made next, which brands can apply through small-batch drops, pre-orders and fast replenishment even without 3D printing.
Brands considering similar steps should record outcomes, such as sell-through of limited runs against forecast, so that later AI tools have a reliable history to learn from.
Frequently asked questions
Does Under Armour use AI in product design?
The sources reviewed do not document AI in Under Armour's design process. A 2018 talk described natural language processing on social media data to inform a shoe, but recent annual reports mention AI only as a risk and competitive factor.
What is Under Armour's 3D-printed shoe?
It is a shoe with a 3D-printed midsole that, as described at a 2018 conference, was sold direct to consumers and made after an online order. It was presented as an alternative to the conventional design-to-sale cycle of around six months.
How does Under Armour use consumer data?
Its annual report says its ability to respond to consumer preferences depends partly on optimising available consumer data. In 2018 it described social media analysis and data from connected shoes and fitness apps.
Is Under Armour upgrading its planning systems?
Its filing for the year ended 31 March 2026 refers to upgrading its end-to-end planning technology system. The filing does not name the vendor or say whether AI is part of the upgrade.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
