Mobile Checkout Optimization Case Study in Pakistan

The Client

An Islamabad-based demi-fine jewellery brand — gold-plated and sterling silver pieces spanning everyday studs, statement earrings, and occasion-adjacent bridal sets priced between PKR 4,500 and PKR 48,000 — selling nationwide through its own Shopify storefront. The brand was founded in 2021 by two sisters who still ran creative and operations respectively, with a team of around thirty people covering the studio, packing, and a two-person marketing function.

Acquisition was Instagram-first. A following of roughly 180,000 drove organic demand, layered under Meta ad spend of about PKR 1.6M a month. The storefront drew around 86,000 sessions a month, 78% of them mobile — the norm for jewellery brands in Pakistan, where the category is browsed on phones during commutes and evening scroll sessions and often shared to family group chats before purchase.

The commercial shape of the store mattered to everything that followed. Monthly volume sat near 990 orders, but the value split was uneven: mobile orders averaged PKR 9,800 while desktop orders averaged PKR 14,600, because the biggest bridal-adjacent baskets were researched and completed on larger screens. Roughly 58% of orders were cash-on-delivery. Demand was also seasonal in the way Pakistani jewellery always is: the October-to-March wedding corridor and the two Eids carried the year, which meant any checkout fix shipped in the third quarter would be stress-tested by the highest-traffic weeks the brand had — a timing reality that shaped the whole engagement calendar.

The brand’s founders engaged WeProms after a growth-agency pitch promised to fix revenue through bigger ad budgets; our read of their analytics was that they were already buying enough traffic and converting too little of it on the device four-fifths of it arrived on. The brief we agreed on was narrower and more honest than “grow revenue”: find where mobile buyers stall, fix those steps specifically, and prove each fix before the wedding-season traffic arrived.

The Problem

The store’s mobile economics were quietly bleeding margin and volume at once. The specific blockers, confirmed in the first fortnight of diagnosis:

• A conversion gap that compounded monthly. Mobile converted at 0.90% against desktop’s 2.06%. With 67,000 mobile sessions a month, that gap was the difference between roughly 600 mobile orders and the 1,380 the same traffic produced at desktop rates — not a target, but a measure of how much room the device was leaving on the table.
• Checkout completion at 41% on mobile. Nearly three in five shoppers who had selected a piece, tapped checkout, and begun entering details abandoned before ordering — against 63% completion on desktop for the same catalog.
• A hostile address form. Fourteen fields, free-text city entry, no autocomplete, and a courier-zone mismatch that threw “delivery not available” errors for perfectly serviceable addresses. Session recordings showed shoppers retyping the same address two and three times before quitting.
• Payment methods that leaked at the edges. JazzCash and Easypaisa redirects returned to a generic page that had lost the cart; bank-transfer instructions pushed buyers off-platform to confirm receipts manually; COD fees surfaced only at the door, generating refusal-driven returns.
• No trust architecture at the moment of commitment. For a category where the shopper is sending money to an Instagram-native brand for an item they cannot touch, the payment step offered no insured-delivery marker, no exchange policy, and no payment logos — nothing to answer “is this safe?” at the exact moment the question peaks.

This is a pattern we see constantly in Pakistani ecommerce: acquisition machinery tuned to a scroll-first mobile audience, attached to a checkout designed for a desktop, card-first market that no longer describes how the country shops.

Phase 1 — Funnel Diagnosis and Baseline (Weeks 1-2)

The engagement opened with the discipline that determines everything downstream: decompose the funnel until the leak is a step, not a sentiment. We instrumented the Shopify checkout funnel in GA4, tagged each checkout step, and pulled four weeks of session recordings focused on abandonment within the checkout window.

Funnel stage	Mobile	Desktop
Session → product view	74%	79%
Product view → add to cart	7.2%	8.2%
Add to cart → checkout started	42%	50%
Checkout started → completed order	41%	63%
Session → order (conversion)	0.90%	2.06%

The table settled an internal argument the brand had been having for a year. Browsing and cart-add rates were nearly at parity — the product pages, imagery, and pricing were doing their job on mobile. The collapse happened after the checkout button: a 22-point completion gap versus desktop. The problem was not demand. It was the checkout.

Payment-method forensics. Segmenting completion by chosen method exposed how uneven the leak was: cash-on-delivery completed at 66%, card at 51%, JazzCash and Easypaisa at a combined 47%, and bank transfer at a measured 22% — the true bank-transfer number was unknowable because the flow pushed buyers off-platform to confirm receipts manually, and many never returned to mark the order.

Exit-voice evidence. An exit survey on the checkout collected 214 responses over twelve days. The leading reasons: payment felt risky for this brand and amount (31%), too many fields and keyboard friction (26%), unclear delivery date or charges (18%), and the demand to create an account (12%). The survey and the funnel decomposition pointed at the same two steps.

Phase 1 closed with the baseline frozen — conversion, completion, payment mix, revenue split — in a dashboard shared with the founders, so every later claim would be judged against a fixed starting point.

Phase 2 — Checkout Restructure and Payment Rebuild (Weeks 3-5)

Phase 2 was the structural build, drawing on the ecommerce conversion optimization playbook we run for Pakistani D2C stores. The organizing principle: rebuild the checkout for how Pakistanis actually pay, then get out of the shopper’s way.

Checkout element	Before	After
Account	Forced login or registration before address	Guest checkout default; optional OTP-verified account after purchase
Address form	14 fields, free-text city, zone-mismatch errors	7 fields with area autocomplete over ~1,900 Pakistani localities
Payment step	Flat radio list, static ordering	Value-ordered methods, upfront COD fee line item, trust strip
Wallet redirects	Generic return page, cart lost	Purpose-built return page with order state and cart restore
Bank transfer	Off-platform confirmation	In-checkout instructions with receipt upload
Order summary	Collapsed behind a tap	Sticky summary with a persistent pay button
Trust markers	None	Insured delivery, 7-day exchange, payment logos at payment step

The address model. The single highest-leverage change. We replaced free-text entry with an autocomplete over a curated dataset of roughly 1,900 areas and localities mapped to their city and courier zone. Shoppers type three characters, tap their area, and city, province, and zone populate automatically. The mismatch errors that had generated support tickets and abandons alike simply stopped occurring.

Recovering the wallet redirect. JazzCash and Easypaisa flows previously landed on a generic post-payment page that had forgotten the cart, leaving buyers who had just authenticated a payment staring at a homepage. The rebuilt return page reconstructs order state, confirms success or explains failure with a retry path, and restores the cart in one tap. Wallet completion moved from 47% toward parity with cards within three weeks of the fix.

Bank transfer brought inside. High-value bridal-adjacent buyers disproportionately chose direct transfer and disproportionately vanished confirming it. In-checkout instructions with the brand’s account details, the order reference pre-filled, and a receipt upload turned an off-platform hope into a measurable step.

Guest-first, account-later. The forced login-or-register wall — the direct complaint of 12% of exit survey respondents — was removed in favour of guest checkout as the default path. Account creation moved to the thank-you page, where a phone-number OTP verifies and attaches the order to a customer record in two taps. The brand lost nothing it actually needed: post-purchase account completion ran above 40% because the shopper creates the account when the order already exists and anxiety is low, not mid-checkout when it reads as one more gate between them and the payment step.

COD fees stated where decisions happen. Cash-handling fees moved from the courier’s doorstep into the checkout as a visible line item beside a stated delivery date range per courier zone. It is a small transparency change with an outsized behavioural effect: the shopper evaluates the fee while still in buying mode, with the product in front of them, instead of discovering it at the door in refusal mode with the parcel in the rider’s hand.

Trust where the question is asked. The payment step gained a compact strip: insured delivery, 7-day exchange, and the logos of every payment method offered. It is a small addition that answers the exit survey’s top objection at the exact pixel where it was raised.

Phase 3 — Experiment and Optimize (Weeks 4-8)

With the structure sound, Phase 3 ran the sequenced experiment program that is the core of mobile conversion optimization work — one variable per test, a two-week minimum runtime, and a 95% confidence threshold before a call was made.

Experiment (mobile)	Result	Decision
Area autocomplete at the address step	+9 points address-step completion	Shipped, week 5
Sticky summary with persistent pay button	+4 points overall mobile completion	Shipped, week 6
Value-based payment method ordering	+5 points completion on carts above PKR 15,000	Shipped, week 6
Prepaid toggle with 3% incentive	+6 points prepaid share, completion-neutral	Shipped, week 7
Urdu reassurance line under the card field	+2 points card completion, inconclusive interval	Kept at zero cost

Two findings deserve note. First, the prepaid toggle — which surfaces the 3% prepaid saving as an explicit choice rather than burying it in payment method selection — moved prepaid share without depressing overall completion, meaning the incentive recruited deciders rather than alarming the COD-faithful. Second, the Urdu reassurance line under the card number field (“your card details are encrypted and never stored”) narrowly missed significance but was retained because it cost nothing and directionally helped the exact segment the exit survey flagged as anxious.

The sequencing discipline mattered as much as any single result. Because autocomplete shipped before the sticky button, and the sticky button before method reordering, each result is attributable. The 17-point completion gain in the final numbers is the additive product of named interventions, each of which survives inspection on its own.

Phase 4 — Roll Out, Measure, and Compound (Weeks 8-12)

By week 8 every structural and validated experimental change was live for all mobile traffic. Phase 4 turned from shipping to verification.

The weekly read. A shared dashboard tracked the mobile funnel stage-by-stage, completion by payment method, prepaid-versus-COD mix, and average order value on each device class. The founders reviewed it every Monday against the frozen Phase 1 baseline — the practice that keeps a 90-day claim honest.

The compounding economics. The prepaid shift carried an operational dividend beyond conversion. At 55% prepaid, more than half of orders stopped carrying cash-on-delivery refusal risk on high-value parcels, and delivery fee disclosure at checkout rather than at the door removed the “surprise fee” refusal pattern support had been logging since 2024. The brand’s ops lead began reporting refusal-driven returns as a separate line item, which is where the COD-fee transparency effect became visible.

Consistency checks. Desktop conversion held essentially flat — 2.06% to 2.11% — over the same window, which is what a mobile-specific intervention should produce, and average order value moved only modestly (mobile PKR 9,800 to PKR 10,200) as the prepaid incentive’s discount effect was offset by larger baskets completing instead of abandoning. Nothing in the final table required a hero assumption.

Operational noise fell with it. Two second-order signals arrived unplanned. Support tickets mentioning address or delivery-zone problems dropped from roughly forty a week to under ten, because autocomplete had eliminated the free-text entries that generated them. And the ops team’s log of refused-at-door deliveries — the costliest failure mode a COD jewellery business has, given insured parcel values — began trending down in the weeks after COD-fee transparency shipped, for the same behavioural reason the exit survey had predicted in reverse.

The wedding-season read. The engagement’s timing meant weeks ten through twelve overlapped the early wedding corridor, when session volume rises and casual browsers dilute conversion. The mobile conversion rate held its post-rollout level through that mix shift rather than regressing with it — about as clean a stress test as a seasonal business offers.

What comes next. With the checkout no longer leaking, the natural next frontier for this store is upstream: product-page trust depth for high-consideration pieces — sizing, materials, and care content that answers the family-consultation loop jewellery purchases pass through before the cart. That is a separate workstream, deliberately sequenced after the leakiest step was fixed.

Final Results at 90 Days

Metric	Before	After	Change
Mobile conversion rate	0.90%	1.25%	+39%
Mobile checkout completion	41%	58%	+17 points
Desktop conversion rate	2.06%	2.11%	+2% (untouched)
Prepaid order share	42%	55%	+13 points
Mobile share of store revenue	51%	59%	+8 points
Monthly orders (flat traffic)	~990	~1,240	+25%
Monthly store revenue	PKR 11.6M	PKR 14.4M	+24%

The revenue arithmetic is deliberately unglamorous: the same 86,000 monthly sessions and the same PKR 1.6M ad spend, converting at a higher rate on the device most of the traffic already used. Mobile orders rose from roughly 600 to roughly 840 a month; desktop held; the mix shifted. These are illustrative outcome ranges built from the patterns WeProms sees across Pakistani ecommerce conversion engagements — not audited third-party figures. They exist so a growth team can sanity-check what a checkout rebuild should plausibly return for a store of this size in this market.

What Made This Work

1. The funnel was decomposed before it was judged. Splitting conversion into five measurable stages ended the “our site doesn’t convert on mobile” conversation and replaced it with “the address and payment steps eat 22 points of completion versus desktop.” A problem you can locate is a problem you can fix; a mood is not.

2. The checkout respected local payment physics. Wallet redirect recovery, in-checkout bank transfer, area-based addressing, and upfront COD fees are not refinements — in this market they are the load-bearing structure. Every one of them addressed a failure mode the payment-method forensics had first proven existed.

3. Trust was placed at the point of maximum doubt. A PKR 20,000+ purchase from an Instagram-native brand is a trust decision before it is a checkout decision. Answering it with insured delivery, exchange policy, and payment logos at the payment step converted an unspoken objection the exit survey had ranked first.

4. Every claim was earned sequentially. One variable per test, minimum runtimes, confidence thresholds, and a frozen baseline. The final 39% is not a before-and-after flattered by seasonality — it is a stack of named interventions, each with its own reading.

5. Commercial effects were tracked beyond conversion. Prepaid share, refusal-driven returns, and per-device average order value were monitored from day one, so the work’s value showed up in margin and operations, not only in a conversion-rate cell.

What Teams Can Apply

For Pakistani D2C and ecommerce teams whose mobile conversion lags desktop by more than half:

1. Decompose before you redesign. Instrument every checkout step and split completion by payment method before touching anything. The leak is almost never “the site” — it is one or two steps, and the fix budget should go there first.
2. Build the address form on areas, not postcodes. An autocomplete over Pakistani localities mapped to courier zones removes the single most abandonment-heavy interaction a local checkout contains. Seven fields with autocomplete will outperform fourteen without, every time.
3. Treat the wallet redirect as part of your checkout. The experience does not end when JazzCash or Easypaisa opens their app — it ends when the buyer returns to a page that remembers them. A return page that restores the cart and states the order’s condition is a conversion asset, not a technicality.
4. Move COD fees and bank-transfer confirmation inside the flow. Fees discovered at the door become refusals; transfers confirmed off-platform become lost orders. Both are checkout features in disguise.
5. Match trust architecture to order value. The higher the consideration, the more the payment step must answer safety directly. For jewellery, electronics, and furniture, a trust strip at payment is routinely among the cheapest conversion gains available.

WeProms Digital has run this framework across jewellery, fashion, beauty, and electronics storefronts in Pakistan. The catalog changes the trust architecture; the sequence — decompose, rebuild for local payment behaviour, experiment with discipline, verify against a frozen baseline — stays the same.