D2C ecommerce · October 2026

Conversion rate optimisation in New Zealand: what's different about a small market

Most CRO advice assumes you have traffic to burn. "Test everything." "Run rapid experiments." "Let the data decide." That works for a US store with half a million visitors a month. On a New Zealand store, the same advice produces tests that run for a quarter and end in a shrug, or tests that get called early and ship a win that was never there.

The problem in a small market isn't tactics. It's arithmetic. A test that reaches a verdict in two weeks on a big US store can take four months on a good NZ one, and some pages will never get there at all. Good conversion rate optimisation in New Zealand is mostly about deciding what not to test, and what to do instead.

Here's how that plays out.

1. The maths nobody puts on the sales call

A/B tests need a certain number of buyers in each arm before the result means anything. How many depends on your conversion rate and how big a lift you're trying to detect. The smaller the lift, the more buyers you need, and the curve is brutal.

Take a store converting at 2.5%, testing at the usual standard (95% confidence, 80% power):

To detect a 20% lift, you need roughly 34,000 visitors across both arms.
To detect a 10% lift, roughly 128,000.
To detect a 5% lift, roughly 500,000.

Now put that against a New Zealand store doing about $2M a year online at an average order of $120. That's around 55,000 sessions a month. A sitewide test hunting a 10% lift takes a bit over two months. A 5% lift would take the best part of a year.

Most real test ideas don't produce 20% lifts. Most produce small ones, if any. So the honest answer to "can we test this?" is often "not at our size, not for that effect." You can run your own numbers on the A/B test calculator before you brief anything.

2. What you can test at New Zealand volume

The pages worth testing are the ones every buyer walks through:

The product page template. One change, applied across every product, gets most of your traffic. On the store above, a PDP template test hunting a 20% lift reads in about a month.
Cart and checkout. Fewer visitors, but they convert at many times the site average, so each one carries more signal. A change here needs far fewer sessions to read.
Sitewide elements. The header, the shipping bar, the offer. Everyone sees them, so they borrow the whole site's traffic.

The common thread is reach. If the change touches most of your buyers, you can probably test it. If it touches a slice of them, you probably can't.

3. What usually can't be tested

This is the list most agencies won't give you, because it's where their test roadmap lives:

Single collection or category pages. On the store above, one collection taking 5% of traffic would need about four years to detect a 10% lift.
Single-campaign landing pages. The campaign ends before the test does.
Submenus and deep navigation. You'd be surprised how few people open them. Check before you spend a month redesigning one.
Anything split by source or device. "Let's see how it did on mobile from Meta" cuts a small sample into smaller ones. You can look at the breakdown. You just can't trust it.

It also catches high-value pages. I've had pages with the highest revenue per visitor on the site that still didn't get enough visitors to test. Valuable and testable aren't the same thing.

4. What to do instead

When a page can't give you a clean answer, you still have to make a call. You have five good options, roughly in the order I'd reach for them:

Check the power before you brief. Run the numbers before anyone writes code. If the test can't detect the lift you'd realistically get inside six to eight weeks, don't run it. That one habit kills most wasted tests before they start.

Make the change bigger. Small tweaks need huge samples. If you've got four trust fixes for the product page, bundle them into one variation instead of testing them one by one. You'll learn less about which fix did the work, but you'll get an answer.

Widen the window, with rules. Borderline surfaces can run longer, but set the stopping rules up front. When I stop a test early it's because it hit a rule I wrote before launch, like being well down on revenue, not because the dashboard looked exciting on day six. (Calling tests on a feeling is how you end up with false positives that cost real money.)

Measure something closer to the change. If conversion rate can't move enough to read, watch behaviour instead. Clicks on the thing you changed, scroll depth, heatmaps. You're checking that people behave the way you expected, not proving a revenue lift.

Ship it on judgment, and say so. Sometimes the right move is to just roll it out. That isn't guessing if the change comes out of research. Surveys, session recordings, reviews and support tickets tell you what's broken. When you can't test the fix, research is the evidence. Write down that it wasn't tested, so nobody quotes it as a proven win in six months.

This is also why research does more work in a small market than a big one. On a new engagement the first three or four weeks are research: customer surveys, on-site polls, heatmaps, analytics, user tests. A US store can afford to test its way to the answer. A New Zealand store mostly has to research its way there, then test the few things big enough to read.

5. Where the money usually is on a small store

If you can't win by testing lots of small ideas, win by picking changes that pay off on every order. On most New Zealand stores I'd start with these:

Post-purchase upsells. The offer appears after the customer has paid, on the thank-you page or in a one-click add-on before the order confirms. Because the first purchase is already done, it can't cost you a sale, which makes it one of the few changes you can ship without a test and still sleep at night. It lifts revenue per order on the traffic you already have. I've written up the full approach, from picking the product to writing the offer, in the post-purchase upsell guide.

The cart drawer. Most Shopify themes ship a cart drawer that shows the items and a checkout button, and nothing else. Add a free-shipping progress bar set in NZD, one relevant add-on, and your delivery promise. Every buyer passes through it, so it's also one of the few surfaces with enough traffic to test.

The free-shipping threshold. Set it just above your typical order, not at a round number someone picked years ago. Nudging people up to it raises average order value. Set it too high and people give up at the cart, so watch conversion when you move it. If rural delivery surcharges eat into your margin, check the threshold still covers them.

Bundles on the product page. "Frequently bought together" or a set with a small saving. It moves order value, and it sits on the product template, which has the traffic to read.

Fixing what's broken. Slow pages, a mobile menu that hides half the range, a checkout that asks for too much. You don't need to test a fix for something research shows is broken. Fix it and move on.

The pattern: in a small market, raising the value of each order often beats chasing a higher conversion rate, because you can ship it with more confidence and it compounds on every sale.

6. The Australia question

Most New Zealand D2C brands end up selling into Australia sooner or later. If both markets run through the same store, you can often test on the combined traffic, which can double the pages you're able to test. Watch the differences that don't carry across, like GST display, payment methods and shipping thresholds. The Australian version of this post covers those.

The short version

A/B testing is an arithmetic problem first. Run the numbers before you brief.
In New Zealand, test the surfaces every buyer touches: product page template, cart, checkout, sitewide.
Collection pages, campaign pages, menus and segment breakdowns rarely have the volume.
When you can't test, bundle the change, measure behaviour, or ship on research and label it untested.
Raise the value of each order: post-purchase upsells, a working cart drawer, the right free-shipping threshold, bundles.
Research is worth more to a small store than to a big one.

If your testing program is full of tests that never reach a verdict, it was probably sized for someone else's traffic. The conversion rate optimisation in NZ page shows how we size a program to what a New Zealand store gets. Or book a 15-minute call and we'll work out which of your pages can be tested and which can't.