Conversion Rate Optimisation · ServiceMOFU
A/B testing services that produce real answers
The short answer
Most A/B tests declare winners that aren't. They're stopped too early, run on too little traffic, measured on the wrong metric, or checked daily until random noise produces a "significant" result. The outcome is a backlog of wins that never show up in revenue.
PMG runs experiments with pre-calculated sample sizes, pre-registered success criteria, full weekly cycles and one primary metric — so a win actually means something.
How we run a test
- Write the hypothesis — "Because we observed X, we believe changing Y will cause Z, measured by metric M."
- Calculate sample size for the minimum effect worth detecting, before launch.
- Run full weeks to capture weekday and weekend behaviour.
- Don't peek and stop — checking daily and stopping at the first significant result inflates false positives dramatically.
- Read secondary metrics — a checkout test that lifts conversion but lowers AOV may not be a win.
- Document everything — winners, losers and inconclusives — in a searchable test log.
What's worth testing
- Offers and pricing presentation — often the biggest effects
- Headlines and value propositions
- Proof placement — reviews, guarantees, trust badges
- Form length and fields
- Checkout steps and payment options
- Page layout for mobile
Button colours are rarely worth the traffic.
When not to A/B test
- When traffic is too low to reach significance in a reasonable time
- When the change is an obvious fix — a broken form doesn't need a test
- When multiple major changes are bundled so you can't learn what worked
- During abnormal periods like sale events, unless that's what you're testing
How to design an A/B test
- Hypothesis — "Changing X for audience Y will improve metric Z because…"
- Primary metric — conversion rate, revenue per visitor, qualified leads per visitor.
- Sample size — estimate how many visitors and conversions are needed to detect the expected change.
- Duration — at least one to two full weekly cycles.
- Run without peeking decisions — monitor for errors, not for early winners.
- Analyse — statistical confidence, segment checks, secondary metrics.
- Decide and document — roll out, iterate or discard.
What to test first
| Priority | Examples |
|---|---|
| High impact | Offer, price presentation, headline, page structure, form length |
| Medium | Proof placement, images, CTA copy, trust badges |
| Low | Button colours, minor copy tweaks |
A/B testing mistakes
- No hypothesis — testing random ideas
- Too little traffic for the change being tested
- Stopping as soon as results look significant
- Running overlapping tests on the same audience
- Measuring clicks instead of conversions or revenue
- Ignoring segment differences (mobile vs desktop)
A/B testing by traffic level
| Monthly conversions | Testing approach |
|---|---|
| Under 200 | Research-led changes, not formal tests |
| 200–1,000 | Big, bold tests on key pages |
| Over 1,000 | Regular testing programme |
A worked example
An eCommerce brand tested button colours for months with no clear results. We switched to testing bigger ideas — delivery-date messaging, product-page layout and offer framing — one at a time, with proper sample sizes. Two of the first five tests produced clear wins.
Before A/B testing
- Hypotheses from research
- Sample-size estimate
- Tracking verified
- Test calendar
How a A/B Testing engagement with PMG runs
- Audit (week 0–1). Accounts, analytics, tracking, website, CRM flow and competitors reviewed, with a written list of what's broken, what's wasted and the three moves we'd make first.
- Plan (week 1–2). Break-even ROAS or a target cost per lead, a measurement plan and a 90-day roadmap for A/B Testing, with owners and dates.
- Launch (week 2–4). Clean structures, verified conversion signals, the first creative sprint and landing-page fixes.
- Optimise (months 2–3). A weekly test cadence, with budget following the cheapest incremental results rather than platform-reported ones.
- Scale (month 3 onwards). New channels, geographies and retention loops, added only when the economics hold.
Every week you get a one-page scorecard, a change log, the test board and a 30-minute call with your strategist. See how we work and pricing for details.
Why teams choose PMG for A/B Testing
- Measurement first. Unit economics and clean tracking come before any increase in spend.
- You own everything. Ad accounts, data and creative stay in your name; ad spend goes straight to the platforms.
- Senior hands on the account. The strategist you speak to is the person running it.
- Weekly transparency. Scorecard, change log and test board — every week.
- Short minimum terms. We keep clients with results, not contracts.
- India and international. One Mumbai team running campaigns across Indian cities and overseas markets, Mon–Sat, 7am–9pm IST.
Where we run A/B Testing
We manage A/B Testing for businesses across India — including Mumbai, Delhi NCR, Bengaluru, Pune, Hyderabad, Chennai, Kolkata and Ahmedabad — and for companies in the UAE, Saudi Arabia, the United States, the United Kingdom, Canada, Australia, Singapore and Germany. Campaigns are planned market by market, with language, currency, platform mix and privacy rules matched to each. See all Indian locations and international markets.
Free growth audit
Get a written audit for A/B Testing.
No sales call required. Tell us where to look; a strategist replies within working hours and sends the written audit within 72 hours.
- Tracking check
- Pixels, Conversions API, GA4 events and offline conversions — what is firing, what is double-counted, what is missing.
- Wasted spend
- Search terms, placements, audiences and overlap that cost money without producing qualified leads or sales.
- Conversion path
- Landing page speed, message match, form friction and follow-up time — where interested people drop off.
- 90-day priorities
- The three changes we would make first, with the metric each one should move.
Frequently asked questions
What if we don't have enough traffic?
Test bigger changes, test on higher-traffic pages, or make research-led improvements without formal testing. Underpowered tests produce confident nonsense.
Client-side or server-side testing?
Client-side is faster to set up but can cause flicker and slow pages. Server-side is cleaner and faster for users. For high-traffic sites and anything touching performance, server-side is worth the effort.
How long should a test run?
Until it reaches the pre-calculated sample size, and for at least one to two full weeks.
What is multivariate testing?
Testing several elements and their combinations at once. It needs much more traffic than A/B tests and suits high-traffic sites.
Can we A/B test ads as well as pages?
Yes. Meta and Google have experiment features; the same principles apply — one variable, enough volume and a fixed duration.
Does A/B testing hurt SEO?
Not when done correctly — using proper redirects or canonical tags and running tests only as long as needed.
What should we test first?
Changes closest to revenue with the biggest potential impact — offers, pricing presentation, key page layouts and checkout steps.
How do we know a test result is reliable?
By reaching the planned sample size and running full weekly cycles before deciding.
Can A/B tests hurt SEO?
Not when done properly with temporary tests and no cloaking.
Is there a long contract?
No. Minimum terms are short, because we would rather keep clients with results than with contracts.
Will we own the ad accounts and data?
Yes. Everything runs in your own accounts with partner access for PMG, and you can see it all at any time.
What do you need from us to start?
Access to ad accounts and analytics, your margins and order values (or deal values), brand assets, and a decision-maker available for a weekly 30-minute call.
Works best alongside
More in this family
All servicesTerms used on this page
Talk to a strategist
Thirty minutes. Your numbers. A straight answer.
Book a strategy call with the people who'd actually run your account. We'll tell you what we'd do — or that you don't need us yet.