Chip, AI Fitting & Shopping Assistant
Chip is TaylorMade's AI assistant that turns product discovery, club fitting, shopping, and support into one conversational experience.
AI Fitting
In-chat Shopping
Search, reimagined
I led the UX audit and redesign strategy, grounding every recommendation in real behavioral data, to improve the fitting funnel, make the conversation shoppable, and personalize it around each golfer's account and game.
Proven impact, untapped potential
When a golfer engages Chip, nearly every downstream metric improves.
Golfers who engage Chip on club PDPs convert at 3.53% vs 1.72% for those who don't.
7.51% add-to-cart for Chip users vs 2.05% for users who never engage.
Chip sessions carry a $441.76 AOV vs $346.53 for Traditional Search.
Source: Quantum Metric · Jan 6 – Apr 15
Chip's gaps
- 1Reach is thin. Chip touches 0.67% of site sessions vs 1.98% for traditional search.
- 2Mobile carries the volume but underperforms. ~70% of Chip sessions are mobile, yet desktop converts higher (5.99% vs 4.18%) with a bigger basket ($534 vs $387).
- 3Most fittings don't finish. Only 37% of AI fittings are completed, most golfers leave before they get a recommendation.
- 4Chip creates intent but hands off the purchase. 69.94% of sessions reach a PDP, but only 7.51% add to cart, the conversation hands shopping back to the site at the decision point.
Five areas to optimize within Chip
I grouped the gaps into five areas Chip should own, each tied to a desired outcome. Every idea that follows maps to at least one.
| Pillar | Purpose | Desired outcome |
|---|---|---|
| Discovery | Find the right product faster with guided recommendations. | Less browsing, less overwhelm, more confidence. |
| Fitting | Get customized product recommendations. | Accessible fitting, confident purchases, higher AOV. |
| Shopping | Move from recommendation to add-to-cart to checkout seamlessly. | Faster purchases, higher conversion, increased AOV. |
| Support | Get instant answers to order, stock, and account questions. | Less friction, faster resolutions. |
| Personalization | Use customer and game data to make every interaction relevant. | Better recommendations, stronger loyalty, repeat purchases. |
Optimize the fitting funnel and make the conversation shoppable & personalized.
Competitors using AI shopping assistants well
- Amazon Rufus shoppable answers that keep the purchase inside the conversation and continue shopping.
- Nike account-driven personalization and recommendations.
- Sephora guided product discovery that turns complex catalogs into confident easy choices.
Needs & Pain Points
Three core user behaviors emerged from the fitting and shopping flow data.
1. Guided Fitting Inputs (AI Fitting)
Fitting · Dev effort: Low
Add tappable answer options, an "Unsure" path, and a progress indicator alongside the open-text fitting questions, 1–2 questions per step.
Only 37% of AI fittings finish today, most golfers leave before they reach a recommendation.
More completed fittings ending in a saved, shoppable recommendation and a confident purchase.
Increase fitting completion by 15 points, from the current 37%.
Previous Design
Optimized Design
2. Shopping Modules
Shopping · Dev effort: High
Product cards, carousels, quick-add, configure CTAs, and save, all inside the chat.
69.94% of sessions reach a PDP but only 7.51% add to cart, there's no shoppable moment at the point of decision.
More add-to-carts and purchases from Chip, and more qualified traffic into club configuration (PDPs).
Launch with an 8–9% in-chat add-to-cart rate as a baseline, against 7.51% today.
Previous Design
Optimized Design
3. Player Profile (Saved to Account)
Fitting / Personalization · Dev effort: Mid
Save hand, height, gender, handicap, and sizes so future fittings pre-fill instead of restarting.
Cuts repetitive input, the top friction point for returning golfers, and builds a reusable player profile.
Faster repeat fittings and recommendations tailored to the golfer's game and size.
Profile completion, and an increase in fitting completion.
Previous Design
Optimized Design
4. Familiarize Users With Chip
Discovery · Dev effort: Medium-High
Use the patterns, language, and interactions golfers already know from search, product carousels and search-style prompts, so Chip feels familiar instead of new.
Chip reaches 0.67% of sessions vs 1.98% for search, but carries a 27% higher AOV.
Captures more discovery volume with product-first results.
Increase Chip's sessions and engagement.
Optimized Design
Use Of Search Terms.
Optimized Design
Product carousel
5. Quick Start (Personalized Prompts)
Discovery / Personalization · Dev effort: Lower
Show 3–5 quick-start prompts: a default state with category suggestions and a tailored state based on the page, personalized by history and behavior.
Cold-start is a major drop point, golfers who don't know what to ask leave without interacting.
More sessions start with a productive action.
Grow Quick Start click rate; reduce no-interaction exits.
Previous Design
Optimized Design
Default Personalization prompts
Optimized Design
PDP Specific Tailored prompts
6. Consistent Visual System
Discovery / Shopping · Dev effort: Very High
Give every product answer the same shape: a short intro, product cards with key specs and shopping CTAs, next steps, and CTAs to shop the category or switch to the fitting tab.
Text-heavy replies bury the recommendation; comparison drives confident decisions.
Faster comprehension and easier comparison.
Increase conversion rate once the user is on a PDP.
Previous Design
Previous Design
Optimized Design
Product Recommendations
Optimized Design
Product Comparison
7. Use Existing Account & Fitting Modules
Support / Personalization · Dev effort: High
Surface order status, loyalty tier/points, and past fittings for signed-in golfers.
Loyalty is the highest-intent audience, leverage what we already know about our members.
Relevant answers on orders and account info; loyalty status motivates the purchase.
Returning-customer CVR & AOV lift; increased account engagement.
Previous Design
Optimized Design
8. Follow-Up (Personalized Prompts & Questions)
Discovery · Dev effort: Low
End every response with promoted follow-up and suggested-topic chips.
Conversations dead-end today, killing momentum toward purchase. Low effort, high return.
Longer sessions with more turns ending in an action.
Grow follow-up click-through rate; increase sessions reaching a PDP or a Fitting.
Final Design
9. Feedback Module
Support · Dev effort: Low
Thumbs up or down, an optional open-form comment, and a clear confirmation.
Only thumbs and no confirmation exist today; feedback creates a response-quality signal.
Continuously improving response quality based on how users engage with Chip.
Feedback submission rate, and continuous response-quality improvement.
Prompt
Confirmation
10. New-User Onboarding
Discovery / Personalization · Dev effort: Lower
A short, skippable first-run intro to Chip's two modes, Chip and AI Fitting, then an invite to save a player profile.
New users underuse Chip because they don't know what it does and fall back on search.
Higher first-session engagement and adoption of Chip.
Onboarding completion rate, first-session engagement time, and profile completion rate.
Step 1
Step 2
Step 3
Step 4