A/B Test Ideas for Ecommerce: 12 Offer Tests Ranked

I'll confess something. The first A/B test I ever ran on my own store was a button color. Orange against green. It ran for three weeks, and the answer at the end was a shrug.

Three weeks of traffic, gone, to learn nothing.

So if you searched "ab test ideas ecommerce" because you've got the testing app installed and no roadmap, here's the short answer. Rank your tests by how much they can change revenue per visitor, then run them top down:

1. Offer tests (bundles, gift thresholds, free shipping lines): biggest swing.
2. Price tests (price points and how the price is framed): next.
3. Headline and above-the-fold message tests: strong, cheap to run.
4. Image tests: worth it once the message is right.
5. Color and button tests: last, and usually never.

The reason is simple. Offer and price change what the buyer gets or pays. A button color changes what the buyer sees for half a second.

A test can only move revenue as far as the thing you changed. Change the deal and the number moves. Change a hex code and it mostly sits there.

Why do most ecommerce A/B tests fail?

Because most ideas are weaker than they feel in the meeting.

This isn't a small-store problem. Ronny Kohavi, who ran experimentation at Microsoft, has said that over two-thirds of ideas tested there failed to move the metric they were built to move. In the same interview he describes Airbnb testing 250 ideas and finding 20 winners. That's 8%.

Now think about your store. Microsoft and Airbnb have millions of visitors to burn on misses. You might have 40,000 a month. Every test you run on a weak idea is a month of traffic you can't get back.

Here's the thing. Low traffic doesn't mean you should test less. It means you should test bigger. A small change needs a huge sample to show up. A big change shows up on the traffic you already have.

That's why I rank the list below by expected impact, not by how easy it is to set up. The easy tests are easy because they barely touch the buying decision.

Which A/B tests move revenue per visitor the most?

Here's how I rank the five families of tests. To be clear about what this chart is: it's our judgment from building product pages for 6 to 8 figure brands, scored 1 to 5. It's a ranking, not measured lift data from a study.

Offer 5

Price 5

Headline 4

Image 3

Color 1

RevenueFlows AI judgment ranking, not measured lift data. 5 = most likely to move revenue per visitor.

And here are the 12 specific tests, ranked inside those tiers. These are the ecommerce ab test ideas I'd put on your roadmap, in this order.

Rank	Test	Control vs variant	Tier (our ranking)	What it touches
1	Bundle offer	Single unit vs a 2 or 3 pack at a better per-unit price	5	Average order value
2	Gift threshold	Percent off vs a free gift above a cart value	5	Average order value, margin
3	Price point	Current price vs 5 to 10% higher	5	Margin, conversion rate
4	Free shipping line	Current threshold vs one set just above a typical order	4	Average order value
5	Price framing	Plain price vs cost per use or per night	4	Conversion rate
6	Subscribe and save	Current discount vs a different discount level	4	Repeat revenue
7	Headline	Product name vs the outcome the buyer wants	4	Conversion rate
8	The "why this price" line	Buried in a tab vs above the fold	4	Conversion rate
9	Guarantee framing	Policy text vs a plain promise near the button	3	Conversion rate
10	Hero image	Studio shot vs product in use	3	Conversion rate
11	Review placement	Reviews at the bottom vs count and stars beside the price	3	Conversion rate
12	Button color	Orange vs green	1	Almost nothing

Why do offer and price tests beat everything else?

Because they're the only tests that can move both halves of the math at once.

Revenue per visitor is conversion rate multiplied by average order value. A headline test mostly pushes on conversion rate. An offer test pushes on order size, and sometimes on conversion rate too.

Watch what happens on a store like this (a hypothetical, so the math is easy to follow):

• Control: conversion rate 1.2%, average order value $95. Revenue per visitor $1.14. On 10,000 visitors that's $11,400.
• Bundle variant: conversion rate drops to 1.1%, average order value climbs to $124. Revenue per visitor $1.36. On the same 10,000 visitors that's $13,640.

Conversion rate went down. Revenue went up $2,240.

If you judge that test on conversion rate, you kill the winner. I've watched founders do exactly that, and it's the most expensive mistake in testing. I wrote a whole post on why revenue per visitor beats conversion rate as the scoreboard, and this is the reason.

The winning variant can convert worse. If your testing app crowns winners on conversion rate alone, it'll hand you the wrong answer on every offer test you run.

Price tests carry more risk than offer tests, because two buyers can see two prices. Do them, but do them carefully: one price per buyer from first visit through checkout, and enough orders per group. My Shopify price testing playbook walks through the guardrails.

And gift thresholds deserve a special mention. A free gift above a cart value often protects margin better than a percent off, because the gift costs you its landed cost, while 15% off costs you 15% of everything. I broke that down in does a free gift with purchase increase conversion rate.

Where do headline and image tests fit?

Right after the offer is settled. Not before.

A headline test is cheap and it can be strong, because the headline decides whether the buyer reads anything else. The swap that usually earns its keep is product name versus outcome. "Cooling Bamboo Sheets" versus "Sleep cool all night, even in July." Same product. Very different first second. My guide to writing product page headlines that convert has the patterns.

Headlines can matter at giant scale too. Harvard Business Review reported that a single change to how Bing displayed ad headlines lifted revenue by 12%, worth more than $100 million a year in the US. It had sat in the backlog for months because nobody thought it mattered.

Image tests rank one tier lower. They matter, but a better photo can't rescue a page that never says why the product is worth the price. Fix the message, then test the picture. On cold paid social, the biggest message test can sit before the product page: send half your clicks through an advertorial you write in one afternoon and half straight to the page.

Here's a real example of what happens when the offer and the message get fixed together. One bedding client's Cooling Bamboo Sheets went from a 1.0% conversion rate and a $125 average order value (revenue per visitor $1.25) to a 4.3% conversion rate and a $254 average order value. Revenue per visitor: $10.92. On 10,000 visitors that's $109,200 instead of $12,500. See the full case study numbers. Real client numbers, not typical results, and not a promise of what your store will do.

That wasn't a button change. The page and the offer were rebuilt.

So should you ever test button colors?

Honestly? Only when everything above is done and you've got traffic to spare.

Button color is the most famous A/B test in the world and the least useful one for a store doing $30,000 a month. The effect, if there's one at all, is tiny. Tiny effects need enormous samples. You'll run it for a month and get a flat line.

There's one exception I'll grant. If your button is genuinely hard to see (grey text on a grey background, lost below a wall of images), fix it. Don't test it. Just fix it. That's a bug, and bugs don't need experiments.

Can AI pick your A/B tests for you?

Partly, and this is where testing has changed the most in the last year.

The big Shopify testing apps now ship AI assistants built for the "I have the software and no roadmap" problem. Intelligems has an assistant called Atlas that, in its own words, surfaces high-impact experiment ideas from your data, and its visual builder lets you describe a change and have AI build it. Shoplift has Lift Assist, which it says analyzes shopper sessions to recommend what converts and generates components like sticky carts and countdown timers.

Useful? Yes. Especially for speed. Writing six headline variants used to take an afternoon. Now it takes a minute.

But here's what I'd watch. An assistant trained on what's easy to build will happily suggest a sticky cart or a timer. Those are tier 3 ideas at best. The tier 5 ideas (a new bundle, a new price, a new gift threshold) need you to make a business decision about margin. AI can draft the variant. You still pick the lever.

If you're choosing a tool, my ranking of the best Shopify A/B testing apps covers which ones can test offers and prices, not only themes.

AI made test variants nearly free to produce. It didn't make traffic free. The scarce thing is still the visitors you spend on each idea.

What to do next

Open your testing app and set up test #1 on your best-selling product only: control versus a 2-pack at a better per-unit price. Set revenue per visitor as the primary metric before you launch, not after. Run it at least two full weeks.

If you want the setup details (sample size, run length, reading the results), the ecommerce A/B testing guide and the step-by-step on how to A/B test a Shopify product page cover it.

---

Book Your Profit Audit

Most stores burn their testing traffic on tier 1 ideas while the offer and the price sit untouched. On a profit audit we'll show you which lever on your best product is leaking revenue per visitor, and how to rebuild a high-converting product sales page in less than 15 minutes.

Book Your Profit Audit →

Or go here to check it out → revenueflows.ai

P.S. A button color test asks which shade people click. An offer test asks what people will pay for. Only one of those questions prints money.