Industries

Conversion optimization for online retailers, where testing actually works

This is the one industry on this site where conversion optimization means what it originally meant. There is enough traffic to run real experiments, the conversion event is observable, and the whole path from arrival to purchase happens in one system. That makes it the rare case where testing is genuinely available rather than a technique borrowed from businesses with more volume. It also means the basics matter more, because a store with a shipping surprise at checkout is losing money that no amount of testing recovers.

Shipping cost revealed late ends more carts than anything else

Unexpected cost at checkout is consistently the largest single reason people abandon, and it is largely self-inflicted.

A customer builds a cart with one number in mind and meets a different one at checkout. The gap feels like a bait rather than a fee, and they leave.

The damage is not the cost itself but the timing. Somebody who knew the shipping charge from the start factors it in; somebody who discovers it at the final step feels misled.

Showing it early is the fix. On the product page, in the cart, or as a threshold clearly stated, so nothing at checkout is a surprise.

Free shipping thresholds work and only when they are visible while somebody is shopping. A threshold discovered at checkout is an obstacle; one shown in the cart is a reason to add another item.

Where shipping is genuinely expensive, saying so plainly costs some conversions and prevents the abandonment that comes from concealment. It also reduces the support contacts and the complaints.

Site search is the highest-intent behavior on the store

Somebody who types into the search box has told you exactly what they want, and most stores handle it badly.

Search users convert at substantially higher rates than browsers in most stores, and they are also the group most likely to leave immediately when the results are poor.

Zero-result searches are the clearest signal available in retail. They are a list of things people wanted and could not find, and it is usually a mixture of stock gaps, wrong terminology and search that cannot handle a typo.

Synonyms and misspellings are the common failure. Customers use different words from your catalog, and a search that cannot bridge that returns nothing for products you stock.

Search analytics are usually available and rarely examined. What people search, what returns nothing, and what they search and then do not buy are three reports with direct actions attached.

It is also product research. Repeated searches for something you do not carry are a demand signal arriving free of charge.

The returns policy is a conversion asset

It removes the risk of buying something you cannot touch, and most stores hide it.

Buying online means committing to something unseen. The returns policy is what makes that acceptable, and a generous, clearly stated one converts.

Hiding it does not prevent returns; it prevents purchases. Customers who cannot find the policy assume the worst, and the assumption is usually worse than the reality.

Stating it where the decision happens, on the product page rather than only in a footer link, is where it does its work.

The cost is real and it is measurable. A more generous policy increases both conversion and returns, and whether that trades favorably depends on your margins and categories, which you can determine.

Return reasons are also a diagnostic. A category with high returns usually has a listing problem about sizing, scale or expectation, which is fixable and reduces the cost rather than accepting it.

Checkout friction is measurable, so measure it

Every field, step and requirement costs a share of the people who reached it, and the store can see exactly where.

Forced account creation is the most reliably expensive requirement in retail checkout. Guest checkout with an option to create an account afterward captures the same customers without losing the impatient ones.

Each additional field costs something. Asking for anything not needed to fulfill the order is a choice with a price attached.

Payment options matter more than they used to, since a customer whose preferred method is missing frequently leaves rather than switching.

Error handling is the overlooked half. A form that clears entries on an error, or rejects a valid address format, loses people who had already decided.

All of this is observable at a step level, which is the advantage this industry has over every other one on this site. The funnel is visible, so the argument can be settled with data rather than opinion.

The second purchase decides the economics

Acquisition rarely pays for itself on a first order, which makes the repeat rate the number that matters.

In most retail categories the first order barely covers what it cost to acquire the customer. The business is built on the second and third.

That makes post-purchase the highest-leverage stretch and the least designed. The order confirmation, the shipping updates, the arrival and what happens afterward.

Delivery experience does more for repeat purchase than most marketing. Arriving when promised, packaged properly, with a straightforward path if something is wrong.

The first repeat purchase is the threshold worth watching, because customers who order twice behave very differently from those who order once, and the interval between them is measurable.

Email and messaging carry consent obligations, so the list built at checkout has requirements attached. Worth setting up properly rather than assembling informally.

This is the one place testing genuinely applies

Every other industry we write about lacks the volume for reliable experimentation, and this one does not.

A store with meaningful traffic can detect real differences in a reasonable period, which makes structured testing available rather than aspirational.

The discipline still matters. Running until a predetermined sample rather than until a result looks good, and accepting that most tests will show nothing.

Test things large enough to matter. Shipping presentation, checkout structure, returns messaging and product page layout produce readable differences; small visual changes generally do not.

Seasonality and promotions confound results, so tests spanning a sale or a peak period are measuring the calendar rather than the change.

The basics still come first. There is no point testing button placement in a checkout that surprises people with shipping cost, and fixing that is worth more than any experiment.

Questions we actually get

What is our biggest conversion problem likely to be?: Shipping cost revealed late, in most stores. The damage is the timing rather than the amount, since somebody who knew from the start factors it in and somebody who meets it at the final step feels misled. Show it on the product page or in the cart so nothing at checkout is a surprise.
Is site search worth investing in?: Search users convert substantially better than browsers and leave immediately when results are poor. Zero-result searches are a free list of things people wanted and could not find, usually a mix of stock gaps, different terminology and search that cannot handle a typo. It is also product research arriving at no cost.
Should we make our returns policy more generous?: It increases conversion and returns together, and whether that trades favorably depends on your margins and categories, which you can measure. What is clear is that hiding the policy prevents purchases rather than returns, since customers assume the worst. State it on the product page, not only in a footer.
Should we require account creation at checkout?: It is the most reliably expensive requirement in retail checkout. Guest checkout with an option to create an account afterward captures the same customers without losing the impatient ones. Each additional field costs something, so anything not needed to fulfill the order is a choice with a price attached.
Is A/B testing worth it for us?: This is the one industry where it genuinely is, provided you have the traffic and the discipline: run to a predetermined sample, accept that most tests show nothing, and avoid periods with promotions or seasonality. Fix the basics first, since there is no point testing anything in a checkout that surprises people with shipping cost.

What is different here

Two areas set the constraints. Advertising claims, subscription and auto-renewal disclosure, and how reviews are gathered and displayed are regulated federally and increasingly at state level too, with negative-option billing drawing particular enforcement attention. Separately, selling into many states raises sales tax collection questions that turn on where a business has nexus. Neither is a marketing decision, so both are worth settling with an advisor before spend scales.

Other disciplines for online retailers

Worth reading alongside this

Written by KC Thompson, Morgul Marketing. Updated September 6, 2026.

Talk to us about conversion optimization for online retailers

Tell us the market you work in and what the account looks like now. We will say where we would start, and what we would not touch yet.