Under Armour Audit

Observed Behavior: On the Men’s Shorts PLP, selecting “Loose” fit and size “L” returns several products tagged as “Compression” fit in size L. The fit filter appears to be ignored when combined with a size filter.

Technical Root Cause: The PLP filtering uses a compound query where the fit parameter is applied to the product level but size is applied at the variant level. The join logic returns all variants of a product if any variant matches the size, regardless of fit.

Business Impact: Users browsing for loose-fit training shorts see compression tights in their results, creating a confusing and untrustworthy browse experience. This is especially damaging for UA because fit specificity is a core purchase criteria for their customer base.

Remediation Path: Refactor the filter query to apply both fit and size at the variant level, returning only products where a matching variant satisfies all active filters simultaneously. Add integration tests covering multi-filter combinations.

Observed Behavior: On footwear PDPs, the “Add to Bag” button is active on page load before the user selects a size. Tapping it without selecting a size scrolls the page to the size selector but shows no error message. The user is left to infer that size selection is required.

Technical Root Cause: The CTA component renders in an enabled state by default. An onClick handler performs validation and programmatically scrolls to the size section, but no visual error (red border, inline message) is triggered.

Business Impact: Users who don’t understand why the page scrolled up may try tapping the button again, creating a frustrating loop. On mobile, the auto-scroll may not bring the size selector fully into view, compounding the confusion.

Remediation Path: Disable the “Add to Bag” button until a size is selected, with a subtle label: “Select a size.” Alternatively, show a clear inline error message on the size selector row: “Please select a size to continue.” Add a brief shake animation to the size selector for visual emphasis.

Observed Behavior: A product on the Outlet PLP shows “$45.99 (was $80)” — a 43% discount. Clicking through to the PDP shows “$49.99 (was $80)” — a 38% discount. The prices differ by $4 between the two pages.

Technical Root Cause: The PLP uses a cached pricing feed updated every 6 hours, while the PDP fetches pricing from the live commerce API. When a price change occurs mid-day, the PLP and PDP display different values until the cache refreshes.

Business Impact: Users who click through expecting $45.99 and see $49.99 feel deceived. Even a small upward surprise in price at the PDP stage is a known trust breaker that suppresses add-to-cart rate.

Remediation Path: Reduce the PLP pricing cache TTL to 30 minutes, or implement a real-time price validation check when the user clicks through from PLP to PDP. If the prices differ, update the PLP tile retroactively via a client-side correction.

Observed Behavior: On mobile PDPs, hovering (long-pressing) the “UA HOVR” technology badge opens a tooltip describing the technology. The tooltip text is truncated after two lines with no “Read more” affordance, and it clips behind the sticky header.

Technical Root Cause: The tooltip component uses max-height: 48px; overflow: hidden with no expand mechanism. Its absolute positioning does not account for the sticky header’s height, and no collision-detection repositioning is implemented.

Business Impact: UA HOVR, ColdGear, and HeatGear technology descriptions are key differentiators that justify premium pricing. When they’re unreadable, the product’s value proposition is undermined.

Remediation Path: Replace the tooltip with a bottom-sheet modal on mobile that displays the full technology description, a product-in-action image, and a “Shop all [Technology]” link. Use a tooltip on desktop where hover works naturally.

Observed Behavior: The PLP filter panel offers options like “UA Storm,” “UA HOVR,” “Iso-Chill,” “ColdGear Infrared,” and “RUSH” without any explanation of what these terms mean. Users who are not existing UA customers cannot distinguish between them.

Technical Root Cause: Filter labels are pulled directly from the PIM taxonomy, which uses internal product-line names. No human-readable descriptions or helper text are provided in the filter UI.

Business Impact: New customers exploring the brand for the first time cannot use these filters effectively. They either ignore them (missing relevant products) or select the wrong technology (finding irrelevant products). This is a critical barrier for customer acquisition.

Remediation Path: Add a one-line description below each technology filter: “UA Storm — Water-resistant outer layer” and “Iso-Chill — Cooling fabric for hot conditions.” Alternatively, offer a “What’s right for me?” quiz link at the top of the filter panel that recommends technology based on activity and climate.

Observed Behavior: On desktop, footwear PLPs show a “Compare” checkbox on product tiles. On mobile, the checkbox is hidden. Additionally, on desktop, selecting 3+ products for comparison opens a panel where only the product image and price are shown — no technology attributes, weight, or drop measurements are included.

Technical Root Cause: The comparison component was built for desktop only with a display: none breakpoint at mobile widths. The comparison data model only queries name, price, and image from the product API, omitting the technology attributes stored in a separate API endpoint.

Business Impact: Under Armour’s competitive advantage is in product technology differentiation. Without a functional comparison tool, users default to comparing on price alone — where UA loses to Nike and Adidas.

Remediation Path: Rebuild the comparison feature to pull technology attributes (cushioning, weight, drop, material) from the extended product API. On mobile, implement it as a swipeable side-by-side card layout. This is table-stakes for performance footwear e-commerce.

Observed Behavior: Searching “cold weather running jacket” returns zero results. Searching “ColdGear” returns results. Users who search by need (activity + condition) rather than by UA technology name get no results or irrelevant results.

Technical Root Cause: The on-site search uses keyword matching against product titles and descriptions. Products titled “UA ColdGear Infrared Shield 2.0 Jacket” are not indexed with synonyms like “cold weather” or “winter running.”

Business Impact: New customers who search by need — the natural behavior for someone who hasn’t yet learned UA’s terminology — are told “no results found,” which is a dead-end exit point.

Remediation Path: Implement synonym mapping in the search configuration: “cold weather” → ColdGear, “cooling” → Iso-Chill, “rain” → UA Storm. Alternatively, integrate a semantic search layer (e.g., Algolia AI Search) that understands intent-based queries.

Observed Behavior: On mobile PDPs, scrolling down to the reviews section is partially obstructed by the 64px-tall sticky “Add to Bag” bar at the bottom. The last review in the visible viewport is always cut off, and users must scroll past it to read the full text.

Technical Root Cause: The reviews section has no bottom padding to compensate for the sticky bar height. The bar uses position: fixed; bottom: 0 with no corresponding padding-bottom on the page content.

Business Impact: Reviews are a critical conversion driver, especially for technical apparel where users want to verify claims like “keeps you warm at 20°F.” Obscured reviews mean fewer users read them, reducing their conversion impact.

Remediation Path: Add padding-bottom: 80px to the page content container when the sticky bar is visible. Alternatively, hide the sticky bar when the user has scrolled into the reviews section and show it again when they scroll back up to the product area.

Observed Behavior: The cart page shows “You’ll earn 450 UA Rewards points on this purchase.” The user has no way to understand what 450 points is worth in dollars without navigating to a separate Rewards FAQ page.

Technical Root Cause: The rewards display component renders the raw point value from the loyalty API without applying the redemption conversion rate (100 points = $1). The FAQ page has the rate but it’s not exposed to the cart component.

Business Impact: Points displayed as abstract numbers have no motivational impact. Users who don’t understand the value don’t factor it into their purchase decision, making the loyalty program a cost center rather than a conversion driver.

Remediation Path: Display points with a dollar equivalent: “Earn 450 points ($4.50 reward) on this purchase.” On the account page, show the total balance as both points and dollar value. Make the conversion rate visible everywhere points are mentioned.

Observed Behavior: Under Armour runs “Buy a Top + Bottom, Save 20%” promotions, but the discount is only applied and visible after both qualifying items are in the cart. On the PDP, there is no indication that the product qualifies for a bundle deal.

Technical Root Cause: The bundle promotion logic is handled entirely by the cart pricing engine. The PDP template has no integration with the promotions API to display qualifying bundle offers.

Business Impact: Users who buy a single item at full price never discover they could have saved 20% by adding a matching item. The bundle promotion fails to drive its intended AOV lift because it’s invisible at the point of product consideration.

Remediation Path: On qualifying PDPs, show a “Complete the Outfit & Save 20%” module with suggested complementary items. On the cart page, if a user has one qualifying item, surface a “Add a [matching item] and save 20%” suggestion above the checkout button.

Observed Behavior: Items added to cart by guest users expire after 30 minutes of inactivity. The user returns to find an empty cart with no explanation of what happened or ability to recover their selections.

Technical Root Cause: Guest cart sessions are stored in a server-side cache with a 30-minute TTL. When the cache entry expires, the cart state is lost. No client-side backup or recovery mechanism exists.

Business Impact: Users who are comparison-shopping across tabs or who step away from their phone return to an empty cart and must re-find and re-add all items. This is a high-friction experience that directly causes purchase abandonment.

Remediation Path: Extend the guest cart TTL to 24 hours. Mirror the cart state in localStorage as a backup. If the server-side cart is empty but localStorage has items, auto-restore the cart and show a “We saved your bag” message. For logged-in users, persist the cart indefinitely.

Observed Behavior: Under Armour offers a 15% student discount via SheerID. Clicking “Verify Student Status” in the checkout redirects the user to SheerID’s external site. After verification, the user is redirected back to the homepage — not to their cart or checkout.

Technical Root Cause: The SheerID integration uses a generic redirect URL configured in the SheerID dashboard. The return URL is set to https://underarmour.com/ rather than a deep-link back to the user’s cart or checkout session.

Business Impact: Verified students must navigate back to their cart, re-enter any promo codes, and restart checkout. Many don’t complete this journey — the verification creates a net-negative conversion impact despite offering a discount.

Remediation Path: Set the SheerID return URL to the cart page with the discount auto-applied: https://underarmour.com/cart?promo=STUDENT15. Pass the cart session ID as a URL parameter to restore state. Test the full round-trip flow end-to-end.

Observed Behavior: UA technology pages (e.g., /t/hovr, /t/coldgear) exist in the sitemap but contain only a hero image, a one-sentence tagline, and a grid of product tiles. No descriptive content, no comparison to other technologies, and minimal internal linking from other pages.

Technical Root Cause: The technology pages were created as minimal product-listing views during a re-platform. Content briefs for educational copy were deprioritized in the launch backlog.

Business Impact: “UA HOVR review,” “ColdGear vs HeatGear,” and “best moisture wicking shirt” are high-intent queries with significant search volume. Without substantive content, UA’s pages cannot rank against editorial competitors (RunRepeat, WearTesters) or Nike’s equivalent technology pages.

Remediation Path: Add 500-800 words of educational content to each technology page: what the technology does, what activities it’s designed for, how it compares to alternatives, and athlete endorsements. Cross-link between related technology pages and from blog content.

Observed Behavior: Each color variant of a product generates a unique URL (e.g., /shoes/hovr-phantom-3/color-black, /shoes/hovr-phantom-3/color-white). All variants share identical page title, meta description, and body copy — only the hero image differs.

Technical Root Cause: The PDP router generates unique URLs per color variant. The SEO meta template uses product-level fields (title, description) that are shared across all variants without color-specific differentiation.

Business Impact: Google indexes multiple near-identical pages per product, splitting ranking authority. The color variants compete with each other instead of the base product consolidating all signals.

Remediation Path: Set all color variant URLs to canonicalize to the base product URL (e.g., /shoes/hovr-phantom-3). Handle color selection via client-side state (hash or no URL change). If color variants must have unique URLs, differentiate the <title> and meta description to include the color name.

Observed Behavior: PDPs include visual breadcrumbs (Home > Men > Shoes > Running) but the corresponding JSON-LD uses ItemList schema instead of the correct BreadcrumbList type.

Technical Root Cause: The developer implementing structured data used the wrong schema.org type. The ItemList schema is valid JSON-LD but is not recognized by Google as breadcrumb markup.

Business Impact: Google doesn’t render breadcrumb rich snippets for UA product pages, missing an opportunity for enhanced SERP visibility. Competitors with correct breadcrumb markup display more navigable, clickable search results.

Remediation Path: Replace ItemList with BreadcrumbList in the JSON-LD template. Each breadcrumb item should use ListItem with position, name, and item (URL) properties. Validate with Google’s Rich Results Test.

Observed Behavior: The homepage and several category pages autoplay a full-width hero video on mobile. The video file is 8.3MB (1080p MP4) and begins downloading immediately on page load, even before the user scrolls.

Technical Root Cause: The video element uses autoplay and preload="auto" without any responsive source switching. The same 1080p source is served to all devices regardless of viewport width or connection speed.

Business Impact: The 8.3MB video payload on mobile pushes LCP beyond 4 seconds on 4G connections. This directly harms Core Web Vitals scores and Google ranking for mobile queries, which account for 70%+ of UA’s organic traffic.

Remediation Path: Serve a compressed 720p version (target: 2MB) for mobile viewports. Use preload="none" and trigger playback only after LCP completes or on user scroll. Provide a static poster image as the LCP element. Use <source> with media queries to serve appropriate resolutions.