Service · conversion rate optimization agency

Conversion rate optimisation

Yes, when a small lift on existing traffic is worth more than the programme costs: a 10% relative lift on ₹1 crore of monthly site revenue adds ₹10 lakh a month.

Performance Marketing practice: abstract red and ink artwork from the Red Balloon reel

Redesigns are the expensive part; a conversion rate optimization agency should spend more of its time on research, using recordings, heatmaps, surveys and analytics to see where visitors stall and why. This page covers our CRO services for e-commerce stores, lead-generation sites and SaaS pricing pages, where most traffic is mobile and small frictions cost real money. We score hypotheses, confirm that sample sizes are reachable, then design and QA variants across devices, including budget Android phones. Tests run in full weeks and are checked for sample ratio mismatch before anyone reads results. Winners are shipped, losing ideas are documented, and landing page optimization feeds straight back into paid media. Test code and results stay with you.

Specialist services

Go deeper

Also covered on this page

Landing page optimisationA/B testing
Where this works

Real-world situations we're built for

Illustrative · not a client
D2C fashion (Gurugram)

Mostly mobile traffic; many shoppers drop at the COD vs prepaid choice

What we'd do
Test prepaid incentive placement, UPI-first payment order and delivery promises, backed by heatmaps
What moves
Checkout completion up, prepaid share up
Insurance (Mumbai)

Quote form has 14 fields; mobile abandonment is high

What we'd do
Step form with progress bar, save-and-resume, fewer mandatory fields
What moves
Form completion rate up; tracked as form completion rate before and after, over a fixed period
B2B SaaS (Bengaluru)

Pricing page gets traffic but few trial starts

What we'd do
Test plan comparison clarity, INR pricing with GST note, proof near the CTA
What moves
Trial start rate up
How a project runs

5 phases, each with a deliverable

What happens, how long it takes, what you get and what we need from you.

Research

2–3 weeks

Analytics review, recordings, heatmaps, surveys, speed and accessibility audit

You get: Insight report

We need: GA4, store and tool access

Prioritise

1 week

Hypotheses, scoring, sample size checks

You get: Test roadmap

We need: Business priorities and constraints

Build and QA

1–2 weeks per test

Variant design, development, QA across devices including budget Android phones

You get: Test ready

We need: Developer approvals, brand sign-off

Run

2–6 weeks per test

Full weeks only, at least two business cycles; sample ratio mismatch checked on day 3; no early calls on partial data

You get: Test result

We need: No other site changes on tested pages

Ship and learn

1 week

Winners implemented, learnings documented

You get: Updated site and test library

We need: Developer time

A typical engagement

From first call to result

Illustrative · not a client

Delhi NCR furniture e-commerce brand on Shopify Plus, about 3 lakh sessions a month, 70% mobile

Weeks 1–2ResearchGA4 funnel, Microsoft Clarity recordings and heatmaps, on-site polls, five customer interviewsProblem list with evidence
Week 2PrioritiseICE scoring; each test sized in the VWO or Convert sample size calculator against traffic and minimum detectable effectTest roadmap
Weeks 2–4Quick winsSpeed fixes checked against CrUX field data, not Lighthouse lab scores alone; bug fixes; clearer delivery and returns copyBetter baseline
Weeks 4–7Test 1Product page: delivery date and EMI options above the fold vs controlShip or drop
Weeks 7–10Test 2Checkout: UPI-first payment order and clear COD fee messageShip or drop
Weeks 10–13Test 3Cart: cross-sell placementShip or drop
Week 13LearnResults archived, hypotheses updatedQuarterly test library
Where it starts0.9% conversion rate, AOV ₹18,000, mobile LCP 4.1 s, no testing tool
Where it endsThe retailer has a testing rhythm and an evidence library. Weekly it watches conversion rate by device, checkout completion and revenue per visitor.
What we measure

The numbers we report on

Every metric has a formula, a target, a source and a reporting rhythm.

MetricHow it's calculatedTargetSourceCadence
Conversion rateOrders or leads ÷ sessions, split by device and sourceSet from baseline; aim for a measurable lift each quarterGA4 + backendWeekly
Revenue per visitorRevenue ÷ sessions (or users) in the testSet from baseline; the usual primary metric for ecommerce testsGA4 or testing toolPer test
Funnel step completionUsers completing a step ÷ users starting itRule of thumb: work first on the step losing the most ₹ (users lost × downstream conversion × AOV), not the highest drop-off %GA4 funnel explorationWeekly
Test significance and powerp-value or probability to beat control, against a sample size fixed before launch95% confidence and 80% power planned up front (common convention)Testing tool (VWO, Optimizely, AB Tasty)Per test
Core Web VitalsLCP, INP and CLS at the 75th percentile of real usersLCP ≤ 2.5 s, INP ≤ 200 ms, CLS ≤ 0.1 (Google thresholds)Search Console, CrUX, PageSpeed InsightsMonthly
Form completion rateForm submissions ÷ form startsAbove your current rate for that form, with field-level drop-off checkedGA4 events or form analyticsWeekly
Test velocityTests concluded per monthRule of thumb: only as many concurrent tests as traffic allows each to reach its fixed sample within 2–6 weeksTest logMonthly
Automated by default

What runs itself after launch

The manual work we take out of this service, and the tools we use.

When: Test endsTest result reportingTesting tool integration with GA4 plus a Looker Studio templateRemoves: No manual result pulls
When: Every deploySpeed regression checkLighthouse CI in GitHub Actions or the PageSpeed Insights APIRemoves: Slow releases do not undo conversion gains
When: WeeklyRecording triageMicrosoft Clarity rage-click and dead-click filters with a weekly digestRemoves: Broken elements found quickly
When: Spike above thresholdForm error alertsGA4 custom insights on form error eventsRemoves: Broken forms caught within hours
When: Exit or idleExit surveysHotjar surveys triggered on exit intentRemoves: Continuous 'why' data
Rules we work to

Standards and regulations

W3CWCAG 2.2

Accessible forms, labels and contrast help every user; level AA is the common target.

GoogleCore Web Vitals

Published speed and stability thresholds that affect experience and search.

Central Consumer Protection AuthorityGuidelines for Prevention and Regulation of Dark Patterns 2023

Tests cannot use false urgency, basket sneaking, confirm-shaming or drip pricing.

MeitYDPDP Act 2023

Recordings and heatmaps must mask personal data and match your privacy notice. Main duties apply from 13 May 2027.

What's changing

Platform and policy updates that affect this work

We track: web.dev blog (Core Web Vitals), Shopify developer changelog, Microsoft Clarity blog, W3C WAI news.

GoogleINP replaced FID as a Core Web Vital.

Measure interaction delay on budget Android phones; heavy scripts and testing tools can hurt INP.

Shopifycheckout.liquid stopped working for Information, Shipping and Payment pages on Plus; Checkout Extensibility became required.

Checkout tests must use checkout UI extensions, not template edits.

Shopifycheckout.liquid support ended for Thank you and Order status pages, with additional scripts moving to apps and pixels.

Post-purchase tracking and upsell tests need rebuilding on extensibility.

MeitYDPDP Rules 2025 notified; main duties apply from 13 May 2027.

Review masking and notices for session replay tools.

Before you hire anyone

Questions to ask any agency, including us

What happens when a test is inconclusive?

Why ask

Most tests do not produce a clear winner

A good answer

Learning captured, next iteration planned, no forced winners

How will you stop tests slowing our site?

Why ask

Client-side tools can cause flicker and INP issues

A good answer

Server-side or edge options, anti-flicker limits, speed checks

Can we see research work, not just test results?

Why ask

Good tests come from research

A good answer

Sample insight reports, survey and recording analysis

How do you guard against false winners?

Why ask

Peeking and novelty effects mislead

A good answer

Fixed sample sizes, full weeks, guardrail metrics

Who builds the variants: your team or our developers?

Why ask

Development bottlenecks kill test velocity

A good answer

Build ownership, QA process, developer hours needed

FAQ

Questions buyers ask us

Straight answers from the practice lead. Highlighted text marks Red Balloon figures we confirm before publishing.

How long does each A/B test need to run?

At least one to two full weeks, and until it reaches the sample size calculated before launch. Stopping the moment a result looks significant produces false winners.

  • At a 3% baseline conversion rate, detecting a 10% relative lift at 95% confidence and 80% power needs roughly 50,000 visitors per variant.
  • Full weeks matter because Indian traffic shifts by weekday, around salary dates and during festive periods.
  • Avoid launching tests during sale events or major campaign changes, which distort the baseline; check the marketing calendar before scheduling.
Which tools do you use for testing and behaviour analysis?

Microsoft Clarity or Hotjar for heatmaps and session recordings, GA4 and BigQuery for funnels, and VWO, Optimizely, AB Tasty or PostHog for experiments. Google Optimize shut down in September 2023.

  • Client-side testing tools can cause flicker and slow pages; server-side or edge testing avoids this on high-traffic templates.
  • Clarity is free with no traffic cap, which suits Indian sites with large volumes of mobile visitors.
  • Test on real budget Android phones over 4G, since fast laptops miss the problems most Indian users face.
  • Run an A/A test on any new tool first: if two identical variants show a significant difference, the setup is broken.
Are session recordings and heatmaps compliant with the DPDP Act?

They can be, if your privacy notice discloses them, inputs are masked so no names, phone numbers or payment details are captured, and recordings are deleted after a defined period.

  • Clarity and Hotjar mask form inputs by default; check custom fields, chat widgets and order confirmation pages, which default masking can miss.
  • Never record payment pages; card data falls under RBI rules and PCI DSS.
  • Keep retention as short as the analysis allows, and limit recording access to named team members.
What do you test first, and how do you decide?

Start where the funnel loses the most money: the biggest drop-off multiplied by traffic and order value, scored for impact, confidence and effort before anything is designed.

  • On Indian ecommerce checkouts, look first at COD and UPI options, shipping cost shown upfront and form length; on lead sites, form length, pricing cues and mobile speed.
  • Each hypothesis records its evidence, the metric it should move and the minimum lift worth detecting, so a test without a winner still teaches something.
Who owns the test results and the code changes?

You own the test log, results, designs and code changes. Winning variants should be built into your site's codebase rather than left running permanently inside the testing tool.

  • A shared test log records hypothesis, screenshots, sample size, result and decision for every test, including losers, which stops the same idea being retested a year later.
  • Winners left in a testing tool add page weight and create a dependency on that tool's licence.
  • Testing tool accounts should sit under your company's billing, keeping test history with you.
Back to the practice

Performance Marketing →

Start with an audit

Ready for
lift?

A paid two-week Growth Audit covers your funnel, ads, site, tracking and automation gaps, and ends with a 90-day plan you can run with or without us. The fee is credited against your next engagement.

Prefer to talk? Call +91 97114 48083, or email info@redballoon.in.

A person from the team replies to every enquiry.

What do you need?
Your details are sent over a secure, encrypted connection and only used to reply to you.