When the Client Wants a Facelift and the Data Says Otherwise
How I challenged a design brief, reframed a pharmacy app's core problem, and led a targeted intervention on the two flows responsible for 93% of conversion failure.
Company
Emami Frank Ross Ltd
Role
Solo Product Designer
Duration
8 weeks
Platform
iOS and Android
ACT 01 ·
The Brief I Challenged
Frank Ross wanted to modernise the UI. They were expanding from Kolkata to Bengaluru, Mumbai, and Chennai. New markets. Same app. Just fresher.
I asked for their analytics before agreeing to scope.
Shipping a prettier version of a broken product into new markets would not accelerate growth. It would accelerate churn at scale.
I went back to the client with a different brief.
ACT 02 ·
Building the Case
I got access to Google Analytics, app analytics, and Play Store reviews. Not to audit the app. To form a point of view before walking into a room with one.
of conversions came through prescription upload and telephone orders. Not search. Not browsing.
came through direct search. At 9% conversion rate, the funnel was barely functioning.
The flow most digital products treat as an edge case was the primary revenue engine.
Pharmacy ordering is not quick commerce. A user fulfilling a specialist prescription is navigating handwritten instructions they often cannot read, ordering medicines whose names they may not recognise, in a category where a mistake has real consequences. Of course they reach for a pharmacist.
The telephone order flow was not a workaround. It was load-bearing trust infrastructure.
Day 1 uninstall rate
Customer retention rate
Search-to-conversion rate
Cart abandonment rate
ACT 03 ·
Negotiating Under Constraint
Stakeholders agreed the problems were real. The pushback came on cost.
A full overhaul was not on the table. So I made a call.
Go deep on the two areas responsible for 93% of conversion activity. Modernise the surface layer across the rest.
The two interventions: prescription ordering including the call back flow, and the end-to-end search to checkout funnel.
Everything else got a visual and language refresh. Plain consumer language replacing pharmaceutical jargon. Navigation restructured around user mental models, not internal taxonomy. No deep structural change unless the data justified it.
This was not a compromise. It was a prioritisation.
INTERVENTION 01 ·
Prescription Ordering: Designing for Two Mental Models
66% of conversions depended on this flow. But the data revealed two distinct user mental models within the same task.
Digital user
Willing to transact if the friction is low enough.
Human-bridge user
Needs a pharmacist to confirm the order before committing.
Upload flow
4 to 5 screens. No progress signals. No upfront indication of which items needed a prescription. High cognitive load in a context where uncertainty produces abandonment, not hesitation.
Redesigned as a single screen using progressive disclosure. Items requiring a prescription surfaced at the top using recognition over recall. One interaction to upload and move forward.
Collapsed from a 5-screen maze to a single, clear interaction.
Call back flow
Users reaching for the telephone option were already signalling hesitation. The original form asked them to manually enter name, phone number, pick a date, and scroll through 30-minute time slots.
A critical affordance failure. The interaction communicated "fill this form to get help" at the exact moment the experience needed to say "we've got you."
Name and number removed entirely. The app already had this information. Topic selector added upfront so the right pharmacist could call back prepared. Time slots reframed as contextual windows: "In next 30 mins", "Today 2pm to 5pm", "Tomorrow 9am to 12pm." Cognitive model shifted from scheduling a calendar slot to simply saying when you are free.
From a friction-heavy form to a single-tap handoff to a pharmacist.
INTERVENTION 02 ·
Search to Checkout: One Connected System
27% of conversions came through search. At 9% search-to-conversion the funnel was not just underperforming. It was leaking at every stage.
Search, PDP, cart, and checkout were not failing independently. They were failing as a system.
We mapped the entire path from search query to order confirmation and treated it as one intervention.
Search
In a pharmacy app the first search failure is a trust problem, not a usability problem.
A user searching for a specific medicine who gets irrelevant results does not try a different query. They question whether the app stocks what they need and leave. Search was restructured around result relevance and scannability. Best Seller and Best Deal tags introduced to accelerate purchase confidence without adding cognitive load. Out-of-stock items surfaced upfront with a Notify Me affordance, converting a dead-end loop into a retention touchpoint.
First search failure in a pharmacy app destroys trust before a single purchase can happen.
Search List Page
Search results are not the destination. They are the evaluation layer between a query and a purchase decision.
The original list page buried the details users needed to choose confidently. Stock status, pricing, product variants, and savings were not scannable at a glance. Users had to tap into individual products just to compare basics.
Redesigned around faster evaluation. Pricing hierarchy tightened, Best Deal and Best Seller tags introduced, and out-of-stock status surfaced upfront so users could compare and move forward without dead ends.
Search results redesigned as a clearer comparison layer between query and product decision.
Product Detail Page
The highest-intent moment in the funnel. The only job of this screen is to remove every remaining reason not to buy.
No expiry date. No delivery timeline. No return policy. No offers. An information desert at the moment users were ready to commit.
Redesigned around the mental model of a cautious first-time buyer. Trust signals, expiry date, delivery window, return policy, and ratings surfaced above the fold. Frequently Bought Together added to increase average order value at peak purchase intent. Product trend indicator introduced to leverage social proof as a decision accelerator.
Every element tied directly to a conversion barrier identified in research. Nothing added for visual completeness.
From an information desert to a trust-first, conversion-optimised page.
Cart
A user reaching the cart has already made a purchase decision. Abandonment here is almost always friction-induced, not intent-driven.
Coupon apply required 5 clicks. A self-inflicted friction tax at the final conversion stage. Collapsed to a single inline action targeting the 30% abandonment rate at its highest-friction point.
Prescribed and non-prescribed items visually separated for clarity and regulatory compliance. Removal confirmation added as an error prevention measure. Contextual cross-sell introduced to drive incremental average order value without interrupting the checkout flow.
Removing the coupon friction tax at the moment users were closest to converting.
Checkout
By the time a user reaches checkout the funnel has done its job. The only design responsibility left is to not break the momentum.
Interaction cost reduced at every step. Visual hierarchy restructured to guide users through without decision paralysis. Payment and delivery information surfaced clearly with no buried steps.
A user who started with a search query could now reach a confirmed order without hitting a single dead end, a missing trust signal, or an unnecessary interaction cost along the way.
Reduced interaction cost at every step so momentum built in the funnel could reach conversion.
REFLECTION ·
What This Project Actually Taught Me
01
Data kills assumptions faster than anything else
I went in expecting a product discovery problem. The conversion data told a different story in the first session. 66% of revenue was flowing through a flow I had mentally categorised as an edge case. That single data point collapsed my hypothesis and restarted my thinking.
02
Domain nuance is not a UX problem
Pharmacy ordering is not quick commerce. A user cannot always read their own prescription. A pharmacist reading handwritten instructions and confirming an order is not a workaround to be designed away. It is trust infrastructure the product depends on. The designer's job is to make that handoff frictionless, not eliminate it.
03
The interface comes after the trust question
This changed how I approach any new domain. The first question is no longer "what is broken in the interface." It is "what do users in this context need to trust before they will transact." The screens follow that answer. Never the other way around.