How to Increase Website Conversions: Stage-by-Stage Guide

You know what they say about a website conversion rate: you are never really done optimizing it.

That applies to new stores and to ones already generating good revenue.

In this post, you will find proven ways to increase website conversion rate and keep it above the average 2%. Whether you are running a new store or one already gaining traction, you will find a strategy to apply today.

Here is what this guide covers:

• What a good website conversion rate looks like

• The five stages where conversions are lost

• Strategies to increase website conversions

What a good website conversion rate looks like

Across ecommerce, conversion sits between 2% and 3%, running higher for food and apparel and under 2% for furniture, travel, and electronics. Compare against your own previous quarter.

At 2%, the other 98% left without buying. Small moves matter at high order values: 45,000 sessions at 0.3% books 135 orders, at 0.4% it books 180.

Campaign conversion rate is a separate number, the share of viewers who acted on a campaign. For the ones below, acting means giving an email, so this is a lead capture rate:

The five stages where conversions are lost

Strategies to increase website conversions

1. Find the stage you are losing people at

2. Say what you sell in the first screen

3. Make the page fast enough to keep mobile visitors

4. Build the mobile version first

5. Strip the page down to the next step

6. Segment visitors before you write the offer

7. Make the thing they came for findable in two clicks

8. Make the product page answer the buying question

9. Compete on something other than price

10. Put reviews and customer photos where the hesitation is

11. Remove the risk with a guarantee you can state as a number

12. Put recommendations where visitors already look

13. Show shipping costs early and set a free-shipping threshold

14. Cut the checkout to the fewest steps and fields

15. Add the payment methods shoppers already use

16. Use the shipping cutoff as the deadline

17. Recover carts onsite, then by email and SMS

18. Give a slow decision somewhere to go

19. Build the channels that let you reach visitors again

20. A/B test every change, with a control group

1. Find the stage you are losing people at

Stage:All

The five stages above tell you which numbers to check. Shopify Analytics and GA4 both report them:

In Shopify:

• Analytics, then Reports, then Conversion over time. Set the range to a full month.

• The report gives sessions, sessions that added to cart, sessions that reached checkout, and sessions that completed, as both counts and percentages.

• Add the Device type and Referrer name breakdowns and run it again for each.

In GA4:

• Explore, then Funnel exploration, with the steps session_start, view_item, add_to_cart, begin_checkout, purchase.

• Turn on Show elapsed time and set the breakdown to Device category, then to Session default channel group.

• For a lead rather than a purchase, replace the last two steps with your form view and form submit events.

Read it as four step-to-step rates rather than five totals: cart adds against product views, checkouts against cart adds, orders against checkouts. The step with the largest percentage drop against the one before it is where your money is going.

Then compare segments inside that step. A store at 2% overall is usually desktop at 3% carrying mobile at 1%, and the gap between them is a better estimate of the opportunity than any published benchmark.

What each stage points to:

• Weak arrival or browsing: speed, message, navigation and search, tactics two to seven.

• Weak product page: answers, proof, differentiation, tactics eight to eleven.

• Weak cart or checkout: recommendations, shipping, steps, payments, recovery, tactics twelve to eighteen.

AI analytics tools skip the report building. You ask a question and get the answer back. We compared the options in our roundup of web analytics tools, and asked this way the exchange looks like this:

2. Say what you sell in the first screen

Stage:Arrival

A visitor arriving from an ad or a search result knows only the words they clicked. The first screen has to name the product, who it is for, and what it does better than the alternative they were also considering.

Industry language is what usually breaks it. Wording that sounds precise inside the company means nothing to someone who arrived 12 seconds ago.

How to test the first screen:

• Show the first screen to someone outside the company for five seconds, then ask them what the site sells.

• Rewrite anything you would not say out loud to a friend.

• Name the one dimension where you are stronger than the alternative, and put it in the subhead.

• Write a separate value proposition for the company and for each product, since they are different promises.

Flaus sells a single product, an electric flosser, and leads its homepage with the value proposition and press mentions before any product detail:

3. Make the page fast enough to keep mobile visitors

Stage:Arrival

Every second before a page becomes usable is a second in which a share of visitors leave, and on mobile connections that share grows quickly. A change further down this list only reaches the visitors who waited for the page.

The usual causes are known: uncompressed images, render-blocking scripts, and third-party tags that were added for a campaign and never removed.

Where to start:

• Serve images in WebP or AVIF at the size they actually display.

• Audit third-party tags and remove the ones no longer tied to a live campaign.

• Defer anything that is not needed for the first screen to render.

• Measure on a mid-range Android phone over mobile data.

4. Build the mobile version first

Stage:Arrival

A design approved on a desktop monitor and checked on a phone afterwards arrives cramped for the share of the audience on a phone. Reversing the order catches that before it ships.

Ideal of Sweden sells phone cases and accessories, and roughly 90% of its visitors are on a phone, so every campaign is built there first. Its capture campaigns have collected 698,000 emails across 12 markets:

The brand also tested the offer itself. A flat 15% off outperformed a gift-card prize draw by three to one, because a shopper on a phone trades an email for a certain reward sooner than for a chance at a bigger one:

How to run a mobile-first check:

• Build and approve the phone version before anyone opens the desktop one.

• Keep the close button reachable with a thumb and away from the form field.

• Test a certain reward against a chance-based one and keep the winner.

• Check the whole path on a phone, since a page that reads well can still hide a checkout step behind a scroll.

5. Strip the page down to the next step

Stage:Arrival · Browsing

Pages accumulate. A banner for a campaign that ended, a link added for a partner, a menu item nobody removed, and the action the page exists for is now one option among many.

Each of those is a competing choice, and competing choices make the decision harder. A page built to serve every visitor at once has no single next step.

The primary action should be visibly more important than everything else on the screen: one button, high contrast, wording that names the action itself.

What to remove:

• Sidebars and secondary navigation on landing and product pages.

• Stock photography that carries no information about the product.

• Second and third buttons of equal visual weight.

• Menu items that exist for internal reasons.

Black Ember keeps its homepage to one hero and one action, and moves product announcements into an onsite feed so nothing competes with it. It is one of 35 stores we take apart in our ecommerce website design examples:

6. Segment visitors before you write the offer

Stage:All

A first-time visitor, a returning visitor, and someone with items in the cart are asking different questions of the same page. One message for all three answers none of them.

Four segments sit in almost every store, and the session already identifies them:

Traffic source, device, and location split the same audience again. Each condition pays on its own:

Segments can also be collected rather than inferred. Foncia manages property across 600 agencies, so renters, landlords, buyers, and sellers all arrive on the same homepage.

A short quiz asks which one the visitor is, and each answer opens the matching page:

Campaigns written for a visitor's own city raised main call-to-action clicks by 6.5% on top of that.

The same split works at country level. carVertical writes each European market its own campaign in the local language, and averages a 19% click rate across a 574,000-contact list.

7. Make the thing they came for findable in two clicks

Stage:Browsing

Catalogs with hundreds of items lose orders at the point of discovery. A shopper who cannot locate the item they arrived for often leaves rather than browsing for a substitute.

Search and filters carry that load on a large store. Filters that match how shoppers actually choose, by fit, occasion, or room rather than by internal category, cut the number of pages between arrival and the product.

What to check:

• Run your ten most common site searches and count the clicks to the right product.

• Add filters for the attributes shoppers name in support tickets.

• Handle zero-result searches by offering close alternatives.

Allbirds splits visitors at the top of the homepage into Shop men and Shop women, so the first click already narrows the catalog:

8. Make the product page answer the buying question

Stage:Product page

Thin product pages assume the shopper already knows what they want. A list of specifications and a price gives someone comparing two options nothing to decide with.

Nutrimuscle sells sports nutrition, where first-time buyers cannot tell which protein or dose fits them. The brand built a comparison tool that walks a shopper through the differences and ends on one product.

The tool ran a 7.1% click rate. It is one campaign inside a wider onsite strategy that attributes €278K across all of its onsite campaigns, so the click rate is the part that belongs to the tool:

How to find what the page is missing:

• Read the last 50 pre-sale support tickets and write down the question that repeats.

• Answer it on the page itself, and in a short tool where the answer depends on the shopper.

• Add zoomable images, scale references, and video for anything where fit or size matters.

• Compare your product against the alternative before the shopper leaves to do it themselves.

9. Compete on something other than price

Stage:Product page

In a crowded category, competing stores often settle at roughly the same price. A shopper comparing three of them sees three identical numbers and picks on whatever else is on the page.

The answers are often true of the store already and missing from the page: lead time, stock held, warranty length, who picks up the phone, whether an expert will size the order.

State them as numbers: ships in two days, three-year warranty, answered within an hour. A figure can be compared, and an adjective cannot.

How to find the non-price argument:

• Open three competitor product pages beside your own and write down what only you can say.

• Turn each one into a number: days to ship, years of warranty, hours to a reply.

• Put the strongest one beside the price, where the comparison actually happens.

• Ask the shopper what nearly stopped them, in a one-question survey after the order.

10. Put reviews and customer photos where the hesitation is

Stage:Product page

Copy written by the brand answers a different question from copy written by a buyer. Ratings, written reviews, and customer photos work at the moment of doubt, which is the product page and the cart rather than a reviews page a shopper has to go looking for.

Where proof sits decides whether it is read at the moment of doubt. Star ratings under the product title, photos from buyers inside the gallery, and the most useful review beside the add-to-cart button all sit at that moment; a long list further down does not.

Where to place proof:

• Star rating and review count directly under the product title.

• Customer photos inside the main gallery.

• Filters on reviews for the attributes shoppers hesitate over, such as fit or durability.

• Payment marks and security signals beside the checkout button.

Proof placed at the point of hesitation, rather than on a page the shopper has to go looking for:

11. Remove the risk with a guarantee you can state as a number

Stage:Product page · Cart

Buying online means paying before receiving. A guarantee moves part of that risk back to the store, and a specific one carries further than a vague promise of satisfaction.

Emma Sleep replaced the form in its cart campaign with a single reminder of the 200-night trial and free returns. Measured against a control group, it cut abandonment 5.54% and raised average order value 5.5%:

How to state a guarantee:

• Give the number: days of trial, days to return, length of warranty.

• Put it next to the price and next to the add-to-cart button.

• Repeat it in the cart, where the objection actually surfaces as a hesitation.

12. Put recommendations where visitors already look

Stage:Browsing · Cart

A shopper who has found one item costs nothing more to show a second. Recommendations work when they use what the session already shows about what the visitor wants.

Pierre Hardy sells luxury footwear and accessories, where a sitewide discount costs more in brand than it returns in orders. So the brand matched recommendations to what it already knew about each shopper.

First-time visitors saw best-sellers, shoppers with something in the cart saw items bought alongside it, and returning visitors saw what they had looked at last time.

The feed carrying those recommendations held an 11.6% click rate, most of it on mobile. Across every onsite campaign the brand runs, those campaigns influenced 22% of online revenue over the year.

Naming the item works better than recommending a category. Vepsalainen sells high-end furniture, and its customers responded six times better to messages about the specific pieces they had already viewed. Best-sellers shown to a first-time visitor:

How to match recommendations to intent:

• Show best-sellers to first-time visitors, since there is no history to personalize on yet.

• Trigger frequently-bought-together the moment an item enters the cart.

• Bring recently viewed products back for returning visitors.

13. Show shipping costs early and set a free-shipping threshold

Stage:Product page · Cart

A shipping cost that appears on the final screen turns a decided purchase back into an open question. Stating it on the product page removes the surprise, even when the number itself is unchanged.

How to use the threshold:

• Set it slightly over the current average order value, within reach of a single add.

• Show the remaining amount in the cart as a currency figure.

• Recommend an item that closes the gap instead of leaving the shopper to find one.

A spend threshold turns that same cost into a reason to add an item.

L'Atelier d'Amaya sells jewelry and runs one in tiers: €5 off from €50, €10 from €80, €15 from €100, with free delivery on top and an end date. A shopper sitting just under a tier has a reason to add one more piece.

Across all its onsite campaigns the brand attributed €971,000 over six months, reaching 95.8% of site visitors:

14. Cut the checkout to the fewest steps and fields

Stage:Checkout

A shopper who reaches checkout has decided to buy. Every screen, field, and interruption between the cart and the confirmation page gives them a chance to stop.

Forced account creation asks for commitment to the store at the moment the shopper only wanted the product. Guest checkout removes it.

Field count works the same way. Each one is another decision before the order completes, and a field that does not affect fulfillment or contact costs more than it returns.

What to change:

• Offer guest checkout, and put the account option after the order is placed.

• Ask only for what is needed to fulfill and contact, and use address autocomplete for the rest.

• Show a progress indicator so the shopper knows how much is left.

• Keep the discount code field small, so shoppers without a code do not leave to look for one.

15. Add the payment methods shoppers already use

Stage:Checkout

A shopper who has entered an address and chosen delivery is one step from paying. A missing payment method ends the order there.

Digital wallets also remove the card-entry step on mobile, where typing a 16-digit number is the slowest part of the whole process.

Installment options change what counts as affordable. An order that feels like too much in one payment can pass as three smaller ones, so the effect is largest where the single payment is largest.

What to check:

• Add the wallets your market uses, and show them above the card form rather than below it.

• Offer installments on the price bands where your cart abandonment is highest.

• Match local payment methods in every country you ship to.

16. Use the shipping cutoff as the deadline

Stage:Cart · Checkout

A deadline removes the option of deciding later. The open question is which one to put on the page.

A discount clock is the default and the weakest, because a shopper who has seen your last sale expects the offer to return.

A shipping cutoff carries information as well as urgency. Order in the next four hours for delivery Friday is a deadline the shopper wanted anyway, since it answers the delivery-date question at the same time.

How to build the deadline:

• Show the cutoff as a delivery date alongside the running clock.

• Bind timers to a fixed end timestamp so they survive a reload and a second device.

• Wire stock messages to live inventory, so the number updates itself and never needs maintaining.

• Pair the end date with a spend threshold, so the campaign carries both a reason to act now and a reason to add one more item.

17. Recover carts onsite, then by email and SMS

Stage:Cart

Cart recovery email reaches people whose address you already have. Everyone else leaves with no way for you to reach them again, which is the case for first-time and anonymous sessions.

An onsite campaign reaches them while the session is still open. 4murs, a home decor retailer running on organic traffic with no retargeting budget, added AI cart recovery in January 2026 and recorded a 24.5% click rate against zero in a control group.

The trigger is what separates the two. AI-triggered campaigns convert at 15.98% across our data against 3.94% for exit-intent, because the model reads the session instead of waiting for the cursor to leave.

Email and SMS then handle the shoppers who did leave an address, which is why the two run together. An onsite recovery campaign, fired while the session is still open, looks like this:

How to set up the sequence:

• Trigger onsite on sessions that have items in the cart and show signs of ending.

• Show the actual cart contents inside the campaign.

• Follow with email within a few hours, and SMS only for shoppers who opted in.

• Test your strongest guarantee against a discount before you give margin away.

18. Give a slow decision somewhere to go

Stage:Product page · Cart

Some purchases take more than one visit, and a shopper who left to compare needs a route back that does not cost you the click a second time.

An announcement channel is the cheapest one. Black Ember sells technical bags and posts product launches to an onsite feed rather than interrupting the browse, which drove 4,019 product page visits in 14 days:

How to make the return trip easy:

• Save the cart across sessions, so a shopper returning a week later finds it still filled.

• Offer back-in-stock and price-drop alerts, which recover demand you already paid for.

• Answer the delivery-date question on the product page, since it is often what the visitor left to find out.

• Publish launches and restocks to a channel the visitor opens themselves.

19. Build the channels that let you reach visitors again

Stage:All

Around 97% of first-time visitors leave without buying, and a store that captured nothing has to pay for the click again. Email, SMS, web push, and an onsite feed are the channels that make the second visit free.

What the visitor is offered decides the rate. Capture campaigns carrying a concrete incentive collect an email from 7.45% of viewers, against 4.60% without one.

A one-question first step raises completion further, with multi-step capture campaigns collecting from 5.17% of viewers against 4.62%. The format is covered in our notes on micro-commitment campaigns.

émoi émoi runs a gamified capture campaign offering products, free delivery, and discounts. It added 8,000 subscribers in a single month at a 17% signup rate:

Emma Sleep asks what the visitor is shopping for before asking for an email. Of the people who pick an option, 68.6% go on to give an address, which raised signups by 50%:

The incentive can carry a threshold of its own. Tikamoon sells furniture and offers €20 off orders over €200, in each market's own language and currency, inside a capture strategy that averages 11,650 new subscribers a month:

How to build the list email and SMS run on:

• Give each page type its own offer, so the incentive matches what the visitor came for.

• Ask the question first and the email second, then collect the phone number on a third step.

• Tag every contact with the segment that captured it, so the first message continues the same conversation.

• Exclude existing subscribers and give them an offer instead of another form.

20. A/B test every change, with a control group

Stage:All

An A/B test splits traffic between two versions of the same campaign or page and reports which one performed better. Our A/B tests raised click rate from 4.64% to 5.86%, an improvement of up to 26% from changing a headline, a delay, or a call to action.

That comparison still leaves one question open. Version B beating version A does not tell you whether either of them improved on showing nothing at all.

A control group answers it. Hold back a slice of the same audience, show them nothing, and report both variants against it.

4murs ran its cart recovery this way and recorded a 24.5% click rate on the variant against zero in the control, on the same audience in the same period.

Emma Sleep measured its cart reminder the same way and reported a 5.54% reduction in abandonment against the held-back group.

How to run an A/B test properly:

• Split traffic between the two variants, and hold back a third slice that sees nothing.

• Judge on revenue per visitor or order rate as well as click rate.

• Change one variable per test and run to statistical significance.

• Prioritize tests by the revenue at stake, so the queue reflects value.

Frequently asked questions

What is a good website conversion rate?

Ecommerce sites typically sit between 2% and 3%, with low-cost categories running higher and furniture, travel, and electronics under 2%. Compare against your own rate from the previous quarter, split by source, since a global average blends very different industries.

What makes a website convert?

Losses cluster at five stages: arrival, browsing, the product page, the cart, and the checkout. Checkout problems cost the most, because the shopper had already decided to buy.

How do you segment website visitors for conversion?

Start with three segments already in the session: new visitors, returning visitors who browsed and left, and sessions holding items in the cart.

New-visitor capture campaigns collect an email from 8.30% of viewers, against 4.60% for a campaign with no condition on it.

How long does it take to increase website conversions?

Targeting and timing changes to existing campaigns show a result within one to two weeks on steady traffic. Checkout and product page changes need a full month of A/B testing against a control group before the revenue figure is reliable.

How do you optimize a website for conversions?

Walk the five stages and fix the latest one first, since a checkout problem wastes every visitor the earlier stages won. Change one thing at a time and measure it against a held-back group before moving on.