What is the fastest way to improve your conversion rate?

Remove a cost the shopper meets for the first time inside checkout. On most stores this is the single largest lever available and it needs no redesign: the delivery charge, the surcharge or the minimum order value moves earlier, so the shopper meets it while they are still deciding rather than after they have committed.

How much can conversion rate realistically improve?

A store with untouched basics can often find a large relative gain in the first two or three changes, and the increments shrink quickly after that. The useful expectation is a compounding series of small validated wins rather than one transformation, which is why a testing habit matters more than any individual fix.

Should I improve conversion rate or average order value first?

Conversion first if people are failing to complete, order value first if they complete easily and buy little. The funnel tells you which: a store losing most people before payment has a completion problem, and raising order value on a leaking funnel raises revenue for a smaller and smaller group.

Does site speed still move conversion rate?

Yes, and mainly below a threshold rather than continuously. Moving from slow to acceptable is worth real money; moving from fast to slightly faster usually is not. Measure your slowest template on a mid-range phone before deciding whether this lever is yours to pull.

Do I need to test every change?

Test anything you intend to keep and anything you might roll out further. Genuine defect fixes can ship without a test, since a broken control is not a hypothesis. Everything else is a guess with a plausible story attached, and the cost of finding that out later is a year of changes nobody can attribute.

How to Improve Your Conversion Rate: The Levers That Matter

Q: How do I know which lever applies to my store?

By where the funnel loses people. Losses before the cart point at the product page and the information on it. Losses between cart and checkout point at cost and account requirements. Losses inside checkout point at effort and error handling. The step names the lever.

Six levers that improve your conversion rate, ordered by what they are worth against what they cost, and the three teams pull in the wrong order.

The CROBenchmark Team

September 21, 2026

Spot your biggest conversion leaks in 15 minutes.

Check best practices, accessibility, data hygiene, and customer sentiment - then compare results with competitors and unlock tailored A/B testing ideas.

Quick Answer

You improve your conversion rate by acting on the specific reason visitors are failing to buy, which your funnel already tells you. Six levers cover almost all of it: late cost disclosure, required effort, unanswered objections, trust gaps, speed below a usable threshold, and error recovery. They are not equal. On most stores a cost the shopper meets for the first time inside checkout is worth more than every design change combined, and it takes an afternoon to move. The order matters more than the list, because teams reliably start with the lever that is most visible rather than the one that is largest, and then conclude that conversion work does not pay.

Key Takeaways

The funnel step where you lose people names the lever; do not choose one from a list.
A cost first seen inside checkout is usually the largest single lever and the cheapest to move.
Required effort is the second: most checkouts ask for fields nobody needs.
Speed pays below a threshold, not continuously, so measure before you invest in it.
Test anything you intend to keep, because an obvious lever is still a hypothesis.

Improving a conversion rate is not a general improvement to a store. It is the removal of a specific reason that people who wanted to buy did not. That distinction sounds pedantic and it decides everything downstream, because a general improvement cannot be chosen by evidence, cannot be sized before you do it, and cannot be measured after. Last updated: September 2026.

Omniconvert has measured conversion behaviour across the CROBenchmark dataset of 7,000+ websites in 15+ industries, against 248+ audit criteria, over 13 years in eCommerce, and Omniconvert Explore averages a 23.2% conversion uplift across 70,000+ experiments. The gap between those two numbers is the subject of this article: knowing where stores leak is common, and acting on it in the right order is not.

What follows is the six levers, ranked by what they are worth against what they cost, and the three that get pulled out of order. If you do not yet know whether your rate is a problem at all, start with a good conversion rate. If completion is healthy and revenue is not, the companion piece on how to raise your average order value is the better use of a fortnight.

Let the funnel choose the lever

Where you lose people names what to fix. Losses before the cart point at the product page and the information on it. Losses between cart and checkout point at cost and account requirements. Losses inside checkout point at effort and errors. Choosing a lever from a list instead is how a quarter goes into the third-largest problem.

The measurement needed here is smaller than most teams assume. Four numbers: visitors reaching a product page, reaching the cart, entering checkout, and completing. The ratios between consecutive steps are what select the lever, and the absolute values barely matter for this purpose.

What you are looking for is the step with the worst ratio relative to how many people entered it, not the step with the largest raw loss. The largest raw loss is almost always the first step, because that is where the most people are, and fixing the first step is usually the hardest work available.

One warning about the reading. A step can look healthy because the people who reach it are already unusually committed, which happens when an earlier step is filtering hard. A checkout with a high completion rate and a tiny number of entries is not a good checkout. It is a good filter somewhere upstream.

Six levers that improve your conversion rate, ranked

Late cost, required effort, unanswered objections, trust, speed and error recovery. The first two are worth more than the other four on most stores and cost the least, which is the opposite of the order in which they are usually attempted.

The last row deserves a note, because small but concentrated is easy to misread as unimportant. Error recovery affects few sessions and affects them at the exact moment of payment, which means every one of those sessions was a completed order that did not happen. It is the cheapest revenue on the table and it is invisible in aggregate reporting.

Lever one: remove the cost they meet too late

A delivery charge, surcharge or minimum order value first seen inside checkout produces an abandonment rather than a reconsideration, because the shopper has already spent effort and now feels the price changed. Moving it earlier costs an afternoon and is the largest lever on most stores.

The mechanism is about sequence rather than amount. The same charge disclosed on the product page is a fact the shopper factors in while deciding. Disclosed at checkout it is a change to a decision they had already made, and people respond to the change rather than to the number.

Baymard Institute's long-running checkout research puts average cart abandonment near seventy percent, with unexpected extra costs consistently among the most cited reasons people give for leaving [Baymard Institute]. That finding has been stable for years and most stores still disclose delivery cost for the first time after the shopper has entered an address.

The fix is a content and logic change: show a delivery cost or a clear rule on the product page, state any threshold in the cart rather than enforcing it at payment, and make sure the total does not change after the payment screen appears.

Lever two: cut the effort you require

Count every field a shopper must complete and mark the ones not needed to take payment or fulfil the order. The count is almost always higher than the team expects, and each unnecessary field is a small tax paid by every single buyer.

Do this as a literal count rather than a review, because reviewing produces justifications and counting produces a number. Walk your own checkout, list every required field, and write beside each one what would break if it were removed.

The usual survivors of that exercise are a company field nobody reads, a second address line that could be optional, a phone number needed only by one carrier, and a mandatory account creation that exists because it was the default. Each is defensible alone. Together they are the difference between a checkout that takes ninety seconds and one that takes four minutes.

Account creation deserves separate attention because it is not a field, it is a decision, and it is made at the worst possible moment. A guest path that is as prominent and as short as the account path recovers a measurable share of the people who stop there.

Lever three: answer the objection where it happens

Every product has one question that stops the purchase, and it is rarely price. Fit, compatibility, delivery timing, what happens if it is wrong. The answer has to sit where the hesitation happens, because a shopper who has to leave the page to find it frequently does not come back.

Finding the question is easier than teams expect. It is the one your support inbox answers most often, the one that appears in your reviews, and the one that shows up in the search box on your own site. Three sources, all of which you already have.

Answering it is where the work goes wrong. A policy page is not an answer, because it is somewhere else. Nielsen Norman Group's usability research has repeatedly found that information required for a decision must be present at the point of the decision, since users reliably fail to seek out content that sits one click away from where they need it [Nielsen Norman Group].

So the answer goes on the product page, near the control that provokes the hesitation. Sizing guidance beside the size selector. Return terms beside the buy button. Delivery date beside the price.

The three levers that get pulled out of order

Speed, redesign and discounting. Each is highly visible, each feels like real work, and each is usually attempted before the two cheap levers above have been touched. That sequence is the most reliable way to conclude that conversion work does not pay.

Speed. Genuinely matters, and the return is threshold-shaped rather than linear. A slow store has real money in this lever; a store already loading acceptably on a mid-range phone has very little. Measure your slowest template before committing engineering time.

Redesign. The most expensive intervention available and the hardest to learn from, because it changes everything at once and produces a result nobody can attribute. If a redesign is happening for other reasons, insist that the funnel is measured before and after; if it is happening to improve conversion, pull the first two levers instead.

Discounting. It reliably lifts conversion rate and often lowers profit, which makes it the clearest case of optimising the metric rather than the business. Bain and Company's work with Fred Reichheld holds that a five percent improvement in retention can raise profits by twenty-five to ninety-five percent, and a discount habit trains exactly the customers you most want to keep to wait for the next one [Bain and Company].

Test the lever, even when it is obvious

A lever that looks certain is still a hypothesis, and the cost of not testing is not the occasional bad change. It is that after a year of confident improvements nobody can say which ones did anything, so the programme has no evidence with which to argue for its next quarter.

There is one sensible exception. A genuine defect, a control that does nothing, a message that says only that something went wrong, can be fixed without a test, because a broken thing is not a hypothesis. Everything else earns a control group.

Run one lever at a time. Changing cost disclosure and form length together produces a result that cannot be decomposed, and the next store you work on will need to know which of the two mattered.

Declare the metric and the decision rule before the data arrives. This single habit removes most of the arguments that make testing programmes unpleasant, because it converts a debate about interpretation into a check against something already written down.

What to do this week

Five actions, none needing a redesign and all finishable inside a week. Start with the free audit so the order is chosen by your store rather than by this article.

Run the free CROBenchmark audit. It scores your store against a dataset of 7,000+ audited websites and names which of the six levers applies to you.
Measure the four funnel numbers. Product page, cart, checkout entry, completion. The worst ratio is your lever.
Buy something from your own store on a phone. Note every cost you meet that you had not met before. That list is lever one.
Count the required fields. Mark the ones not needed to take payment or fulfil the order. That count is lever two.
Test the top item against a control. Omniconvert Explore is the CRO platform for that step, with A/B and multivariate testing, on-site surveys, heatmaps and session recordings. Where findings arrive faster than they can be tested, Nexus by Omniconvert is an AI for eCommerce growth engine that unifies commerce data, ranks experiments by True Profit, and generates campaigns and creative you approve before they go live. For the finding step itself, croaudit.marketing covers how to audit it.

The bottom line

Most stores do not have a conversion problem so much as a sequencing problem. The two levers worth the most cost the least and are invisible in a design review: a cost that arrives too late, and a set of fields nobody needs. They get skipped because they do not look like work, and the quarter goes instead into a redesign, a speed project or a discount, each of which is visible, expensive and frequently worth less than the thing that was skipped. So let the funnel choose. Buy from your own store on a phone and write down every fact that surprised you. Count the fields. Fix the largest item and prove it moved something before you fix the second. That is an unglamorous programme and it compounds, which is more than can be said for the alternative.

Source: Omniconvert, conversion levers by effort and value across the CROBenchmark dataset
Lever	What it removes	Effort	Typical value
Late cost disclosure	A price the shopper meets after committing	Low, a content change	Highest
Required effort	Fields and steps nobody needs	Low to moderate	High
Unanswered objection	A question that has no answer on the page	Moderate	High
Trust gap	Doubt about whether it is safe to buy	Moderate	Moderate
Speed below threshold	Waiting long enough to reconsider	High, engineering	Moderate, then flat
Error recovery	A dead end at the last step	Moderate	Small but concentrated