Shoppers left the product page because it couldn’t answer what a store counter would.
Aurumgemé, the online fine-jewellery label of a large Indian gold retailer, was losing 68% of visitors on its product page. As the only researcher on the agency team, I ran the research end to end and drew the first wireframes. After the redesign went live, the client reported that conversion had doubled from 3% to 6%.

SituationA trusted jewellery name, losing people on one page
Aurumgemé sells gold and diamond jewellery for everyday wear and gifting, to shoppers from 8 to 80, through its own stores, online marketplaces, a website and an app. The parent brand was one of the most trusted names in Indian jewellery. Online, that trust wasn’t carrying over.
The product display page (PDP) was where shoppers decided whether to buy, and most of them left from it. The client knew where people were dropping off, but not why.
Where shoppers dropped off
TaskFind out why, and give the designers something to build on
I was the only UX researcher on the agency team, responsible for the research from start to finish and for the first low and mid-fidelity wireframes. The agency’s designers would take those to high fidelity. I needed to answer three questions:
- Why are shoppers leaving the product page instead of adding to cart?
- What does someone need to know before spending on gold or diamonds online?
- What should the team fix first, with limited design and build time?
A product page with a high exit rate.
What stops someone trusting a high-value purchase they can’t touch?
ActionListen first, watch people shop, then prioritise
How the ten months ran
01Started with what shoppers were already saying
Before recruiting anyone, I read the app-store reviews, studied the Indian online gold market, and compared the PDP with three competitors. The reviews pointed at the same gaps the interviews would later confirm.
Competitor feature analysis: 15 product-page features
On paper Aurumgemé was level with the market. It was missing customer reviews, EMI (instalment) options and related products, and its load time scored 3.5 of 5. The feature list wasn’t the problem; how the page worked was.
Where the four product pages differed
| BlueStone | CaratLane | Aurumgemé | PC Chandra | |
|---|---|---|---|---|
| Customer reviews | Yes | No | No | No |
| Related products | Yes | Yes | No | Yes |
| EMI (instalment) options | Yes | Yes | No | No |
| Large, high-quality images | Yes | Yes | Yes | No |
| Detailed descriptions | Yes | No | Yes | Yes |
| Availability shown | Yes | Yes | Yes | No |
| Product comparison tool | No | No | Yes | No |
| Social sharing | No | Yes | Yes | Yes |
| Search results | Poor | Good | Good | Poor |
| Overall design | Okay | Okay | Okay | Poor |
| Load time | 4/5 | 5/5 | 3.5/5 | 3/5 |
All four had specifications, pricing, add to cart and wishlist, filters, buy-in-store and a responsive layout.
02Recruited 30 people and watched them shop
I interviewed 15 people on the current site and kept 15 for the redesign, so we could compare like with like at the end. Using everyone up front would have given richer early data, but no clean before-and-after.
Each session followed real shopping tasks: find a ring, check the size, understand the price, add to cart. I annotated the PDP with what people did, next to the heuristic issues I found.
21 participants had bought from a jewellery website before and 9 were new to this one, so we heard from both habits and first impressions.

Who took part
Aged 18 to 65 (most 22 to 52), Indian shoppers living in India and abroad, from a range of professions. Gap I noted at the time: no unemployed men, and not everyone was the buying decision-maker at home.
Task analysis: what people did in each part of the journey
03Found six barriers, all about confidence
11 of the 15 participants didn’t read the description. Several asked why the details that mattered most sat under collapsed headings.
“This looks really cluttered.”
Most participants were unsure how to measure a ring size, and the size charts weren’t consistent between products.
No breakdown by gold rate, weight or making charges, and worries about hidden charges at checkout. Instalment (EMI) options were missing.
“I felt discouraged due to the lack of transparency in pricing and the presence of hidden charges.”
No reviews, no certification on the page, few close-ups. A trusted brand name wasn’t enough on its own.
“I did not feel assured due to the absence of any kind of certification of the jewellery.”
Slow loading, filters that returned irrelevant results, and pages that didn’t look like each other. Compare, currency and quantity controls didn’t work.
“One page looks different from the next.”
Few ways to ask a question before buying, and worries about payment security.
“I’m always worried about my payment information being secure.”
“I wish there were more ways to get help.”
From the frustrations board, grouped
Trust and security
“I worry about the security of my payment information.”
“I am concerned about the authenticity of the jewellery.”
“I find it difficult to trust reviews without verified purchase badges.”
Seeing the piece
“I am disappointed by the lack of multiple angles and close-up shots.”
“I am frustrated by the absence of videos showing the jewellery being worn.”
“Lack of detailed information about materials and craftsmanship.”
Size and fit
“Lack of clear size guides for necklaces, bracelets, etc.”
“Inaccurate size information for rings and bracelets.”
“Inaccurate product measurements and sizing information.”
Finding things
“Limited filtering options make it hard to refine search results.”
“Inconsistent website layouts and designs across pages.”
“I recommend adding more filters to narrow down searches.”
Speed and mobile
“Slow website response times during peak shopping hours.”
“I feel the mobile version of the website could be more user-friendly.”
“Lack of intuitive and user-friendly mobile apps.”
After the order
“I am unsure about the return and refund policies.”
“I feel the customer service response time is too slow.”
“I am dissatisfied with the lack of regular updates on my order status.”
Shoppers weren’t put off by the jewellery. The page left out the reassurance a store counter gives: the real price, the right size, a close look and someone to ask.
04Turned findings into something designers could use
I coded the data in a thematic analysis, built five personas and mapped the current journey from first ad to repeat purchase. Then I mapped each problem to a research fix, a design fix and a feature, and ranked them by severity so the designers could work in phases.
Pieces she can trust are genuine and ethically sourced.
Struggles withNo certification or proof of authenticity on the page.
To browse and buy quickly on his phone.
Struggles withAn app and website that behave differently.
Versatile pieces that fit her budget.
Struggles withToo many options and unclear return policies.
High-value pieces that hold their worth, often as gifts.
Struggles withTrusting an expensive order to arrive safely and be genuine.
Unique, affordable pieces to give to friends.
Struggles withHidden charges that appear at checkout.
Five personas, each a composite of patterns from the interviews.
Thematic analysis: 6 themes, 35 coded problems
Slow loading · Navigation · Mobile experience · Search
Sizing · Descriptions · Images and close-ups · Care and maintenance
Payment security · Seller credibility · Authenticity
Size and quality issues · Returns and exchanges · Support response
Different behaviour per platform · Mobile-first expectations
Reviews shape decisions · Customer photos and posts · Requests for more detail
Priority by severity, which set the design phases
Current journey, from first ad to repeat purchase
05Wireframed, tested, then scaled up
I drew the low and mid-fidelity wireframes in Figma, more than ten variations of the PDP hierarchy, and the agency’s designers took the chosen direction to high fidelity. Before building at scale, I ran moderated usability tests on the prototype with 10 of the original participants. Most earlier concerns were resolved. What remained (filters still returning irrelevant items, stock information out of date, live chat not always available, a few errors) went back to the team before the full build.
Prototype test round: what 10 participants said and hit
Observations
Checkout felt straightforward and quick.
Participants noticed recommendations based on their browsing history.
Positive quotes
“The website’s interface is sleek and modern. It was a joy to browse through the collections.”
“I loved how quickly I could find exactly what I was looking for, thanks to the enhanced search features.”
“The product images are stunning and give a real sense of what the jewelry looks like in person.”
“The reviews and ratings… helped me make an informed decision. I felt confident purchasing.”
Negative quotes
Filters sometimes didn’t work as expected and showed irrelevant products.
“The filters and categories can have small images next to the headings.”
“The live chat support is not always available.”
“The availability information seems outdated. I tried to buy a necklace that showed as available, but it was out of stock when I went to check out.”
Errors
“I encountered an issue where my cart was emptied unexpectedly.”
Broken links on product pages or during checkout.
Search missed synonyms and related terms.
Prices differed between the product page and checkout.
ResultThe redesign shipped, and the numbers moved


After launch I ran the second half of the study: the 15 held-back participants (Group B) on the redesign, against the original 15 (Group A), over three weeks with interviews and the same tasks. A t-test supported the hypothesis that the redesign improved satisfaction and task completion. After launch, the client’s team also shared their own website numbers for the old and new site.
Reported by the client’s team after launch, old site vs new site
The 68% in the original brief is a different measure: the exit rate, how often a visit ended on the product page. Bounce rate only counts visits that left after viewing a single page.
From the moderated sessions, Group A (15) vs Group B (15)
- Group B mentioned faster loading, easier menus and filters, and better product images. One participant said the close-ups made them confident in their choice.
- I reported back module by module, with next-phase recommendations: personalised recommendations, virtual try-on, a support chatbot and WCAG accessibility.
- On attribution: the agency’s designers built the high-fidelity UI and the client’s team shipped it. My part was the research, the prioritisation and the first wireframes.
ReflectionWhat I’d do differently
- Treat the comparison as directional. Two groups of 15, tested months apart, is not an A/B test, and the website numbers cover different periods, so seasons and campaigns may play a part. Next time I’d push for a true A/B test on live traffic, both versions running at the same time.
- Keep the hardest problem on the table. Price transparency was one of the strongest findings, but a gold-rate price breakdown didn’t make it into the first release. I’d argue harder for it in phase two.
- Widen the sample. We had no unemployed men, and not every participant made the buying decision at home. Both shape how people shop for jewellery.
- Pick the idea that adds the most value. We had 35 coded problems. Ranking them by severity is what made the build manageable.
Client brand details are covered in screenshots, and the low and mid-fidelity wireframes stay private under the client agreement. Personas are composites, not real individuals. The research, process and outcomes are real.