Case Study · E-commerce · Product Discovery · Growth

The checkout wasn't broken.
The problem was earlier.

Every client said "fix our checkout." Checkout was fine. Users were leaving because they couldn't find what they wanted — not because paying was hard. The insight changed the entire product investment, from checkout optimization to a personalization engine. +16 percentage points of conversion followed.

Iksula Services Business Analyst → Product Lead Jul 2019 – Oct 2021 E-commerce · ITC · Eureka Forbes · ELC Toys
+16pp
Conversion Rate
Improvement
6
Products Shipped
End-to-End
95%
Client Satisfaction
Score
22%
Cart Abandonment
Recovery Rate
8w
A/B Test Duration
to Significance
01 Discovery
Rejected the default diagnosis
Session recordings and behavioral data revealed the drop-off was at product discovery, not payment
02 The Bet
Personalization over checkout
More expensive, harder to measure, invisible in a demo — but the right call based on data
03 Execution
Built and shipped across 3 accounts
Personalization engine, mobile PDP restructure, and abandoned cart flows — simultaneously
04 Measurement
8 weeks, statistically significant
+8pp homepage personalization, +6pp mobile PDP, 22% cart recovery — total +16pp
store Context
Where I Started My PM Journey

Iksula — The Company

Iksula helped premium Indian brands build their own e-commerce channels. In 2019–2021, brands like ITC, Eureka Forbes, and ELC Toys were watching Amazon dominate their categories and trying to compete on their own websites. Iksula built and operated those websites.

My Role

Started as "Business Analyst" — in practice, the PM on three accounts simultaneously. No PM above me. Owned the product lifecycle from discovery to launch to performance optimization. I was 23. I was figuring this out as I went.

This is where I learned product management. Not from a methodology or a framework, but from being accountable for numbers I didn't fully control, for clients I'd have to face in review calls, with data I couldn't ignore. Every product belief I hold today, I can trace back to a specific moment at Iksula where I learned it the hard way.

The Problem
The Common Complaint That Was Wrong

Every client had the same complaint: "Our website traffic is fine. Our conversions are terrible. Fix the checkout."

The industry default response was checkout optimization — fewer form fields, faster payment, reduced friction. We tried that first. The conversion needle barely moved.

When I mapped the full user journey using session recordings (Hotjar) and behavioral data (Google Analytics) across three client accounts for 4 weeks, a different picture emerged:

1

Users were abandoning at the product detail page, not payment

On mobile, 60%+ of checkout abandonment happened before the payment step. Users were leaving from the product detail page. The checkout was fine. The product presentation wasn't.

Finding: Wrong stage of the funnel — payment wasn't the problem
2

Returning users saw the same homepage as new users

A customer who'd bought baby toys last month was seeing ITC's cooking sauce promotions. A returning Eureka Forbes customer was shown a product category they'd never browsed. There was no personalization layer at all — every visit started from zero.

Finding: Discovery failure — the site didn't know its own customers
3

Category navigation was organized by internal logic, not customer mental models

ITC had 15+ product lines. The navigation was organized by their internal business unit structure. Customers couldn't find what they were looking for because the categories reflected how ITC was organized internally, not how customers thought about what they needed.

Finding: IA problem — the product architecture served the company, not the customer

The insight: The problem wasn't checkout friction. It was discovery failure. Users were leaving because they couldn't find what they wanted — not because paying was hard. Fixing checkout would have optimized the 40% of the funnel that was working. We needed to fix the 60% that wasn't.

The Bet
Personalization Over the Industry Default
The Non-Obvious Call

Build a personalization engine instead of optimizing checkout.

The personalization engine was the riskier call in every dimension:

More expensive to build than a checkout optimization. Checkout changes are incremental; a personalization layer requires a new data pipeline and recommendation logic.

Harder to measure — you need enough time and volume to see the impact. Checkout optimization shows results in 2 weeks; personalization needs 6–8 weeks minimum.

Invisible in a demo — you can't show a client "personalization working" in a meeting. A new homepage design is immediately visible. A recommendation engine that's learning is not.

I made the case to ITC's digital team using a competitive comparison: their Amazon.in performance (where their products converted well) vs. their own site performance. The difference was almost entirely personalization — Amazon knew what each user wanted; ITC's site did not.

They approved it. Here's what happened:

After 8 weeks of A/B testing, personalized homepage drove +8pp conversion vs. control. Mobile PDP restructure drove +6pp on mobile sessions. Combined with cart recovery emails, total uplift: +16 percentage points. The personalization bet was the right call.

Solution Design
Three Interventions, One Diagnosis
1

Personalization engine — homepage layer

Used browsing history and purchase data to serve personalized hero banners for returning users (category-specific), "recently viewed" and "you might also like" modules, and preference-based category surfacing. First time a returning user saw a homepage that reflected their history.

2

Mobile PDP restructure

Restructured product detail pages to surface price, availability, and the buy button above the fold on mobile. Previously, users had to scroll 3–4 screens to find critical purchase information. Rearranged the information hierarchy to match how mobile users actually make purchase decisions.

3

Personalized cart abandonment email flows

Triggered email sequences using the specific product the user viewed or added to cart — not generic "you left something behind" emails, but emails featuring the actual product with personalized copy. Measured against a control group receiving generic recovery emails.

Measurement
8 Weeks, Three Interventions, One Framework

Ran A/B tests across all three client accounts for 8 weeks — long enough to reach statistical significance given each account's traffic volume. The test duration was intentional: a 2-week test on these traffic levels would have produced false positives.

InterventionTest GroupControl GroupResultConfidence
Personalized homepage (returning users) Personalized hero + modules Generic homepage (all users) +8pp conversion 95%+
Mobile PDP restructure Price/CTA above fold Original layout (price below fold) +6pp on mobile 95%+
Personalized cart recovery email Product-specific email Generic recovery email 22% recovery rate 95%+

Combined effect: +16 percentage points overall conversion across accounts, anchored by ITC as the lead account. The +16pp is not additive across all three tests — it represents the net conversion improvement across the full user population after all three changes were deployed.

Key Decisions
The Calls I Made at 23
DecisionWhat I ChoseWhyAlternative Rejected
Where to investigate first Full user journey analysis before any solution Conversion problem wasn't where everyone assumed; needed to find where in the funnel before proposing what to fix Jump directly to checkout optimization — the industry default, and in this case, the wrong answer
Personalization vs. checkout Personalization engine first Data showed discovery was the gap; checkout was actually working fine; optimizing a working funnel stage has diminishing returns Checkout optimization — less risky, faster to build, but addresses the wrong failure mode
A/B test duration 8 weeks to statistical significance Traffic volume required 8 weeks to avoid false positives; shorter tests would have produced misleading results at conventional significance thresholds 2-week test — faster feedback but high risk of false positive leading to shipping the wrong change
Client communication Weekly written updates with data Builds trust, surfaces problems before they become surprises, gives clients visibility into work they can't see yet (the personalization engine learning) Monthly review — slower feedback loop, no mechanism for course correction
trending_up Results
What the Numbers Said
+16pp
↑ Improvement
Conversion Rate
Across All Accounts
6
Shipped
Products Delivered
On Time, In Scope
95%
Achieved
Client Satisfaction
Score
+8pp
↑ from personalization
Homepage Conversion
(Returning Users)
22%
Recovery rate
Cart Abandonment
Email Recovery
1st
Awarded
Excellence Award for
"Product Ownership Mindset" in BA role
Lessons Learned
What Iksula Built in Me
Lesson 01

The presenting problem is almost never the actual problem.

Every client said "checkout is broken." Checkout was fine. I learned to be professionally skeptical of the first diagnosis and to map the full journey before proposing any solution. This habit has surfaced the real problem in every role since.

Lesson 02

Risky bets need a strong narrative, not just data.

The personalization engine was the right call — but it wasn't obvious, and the data alone wasn't enough to close it. I had to build an argument: here's what Amazon does, here's what you do, here's the gap, here's what it's costing you. Data gives you the case. Narrative closes it.

Lesson 03

Product ownership is not about authority. It's about accountability.

I had no formal authority over the technology partners building these products. I couldn't stop a sprint or reassign anyone. What I had was clarity about the goal and willingness to be accountable for the outcome. That turned out to be enough — and it's still the most important PM skill I use.

Lesson 04

Ship fast, measure everything, change your mind when the data says to.

In 2 years I shipped 6 products. Not all of them worked perfectly the first time. The ones that succeeded are the ones where we measured closely and iterated quickly. The ability to change your mind based on evidence — without ego, without attachment to the original hypothesis — is a product superpower.