B2C Website & Conversion Optimization Playbook

Conversion optimization for a consumer site means finding the step where shoppers drop out, usually on a phone and often in checkout, fixing what is plainly broken and testing only what your traffic can measure. It takes a funnel you can trust, an honest ranking of fixes and a review habit, not a redesign.

I wrote this edition for brands that sell to consumers online: ecommerce stores, subscription services and apps that sign people up on the web. If your site’s job is to win a demo request or a sales conversation, read the B2B edition instead.

The steps run in three phases over about 12 weeks: diagnose in weeks 1 to 3 (Steps 1 to 4), improve in weeks 4 to 8 (Steps 5 to 8), and make it routine in weeks 9 to 12 (Step 9).

Agree what success means before anyone touches a page. Pick one goal for the sprint and write it as a rate: purchase conversion (sessions that end in an order), subscription sign-up rate or first purchases from new visitors. Repeat purchase matters too, but it is mostly won after the first order, so keep it as a secondary number for now.

Then map the stages a shopper passes through, one path for each way of buying. A store usually has a product path (product view, add to cart, cart, checkout started, order) and, if it sells subscriptions, a plan path (landing page, plan page, checkout started, subscription active). Name each stage after the analytics event that records it, so the map and the reports use the same words (Figure 1).

Last, choose the cuts you will read every stage by: device (desktop, phone, tablet), traffic source (paid, organic, direct, email, social), new against returning customers, and any high-value group such as email subscribers or loyalty members.

Every later step reads these numbers, so check three things first.

• Events. Each stage in your map fires its event once, on desktop and on phones, including orders paid with express wallets and orders that finish on a confirmation page your platform hosts.
• Orders and revenue. Purchases in analytics match orders in your store platform for the same period, and each purchase carries revenue, tax, shipping and any discount code. Send the order number as the transaction ID: Google Analytics counts purchase events with the same transaction ID once, so a reloaded confirmation page can’t record a second order^[2].
• Campaign tags. Paid, email and social links carry consistent UTM tags, so you can see which channels bring revenue, not just clicks.

In my view some gap between analytics and the store always remains; what matters is that you know its size and its cause. Use the store’s orders for revenue, and analytics for the steps before the order.

Work out three stage rates: add-to-cart rate (product views to add to cart), cart-to-checkout rate (cart to checkout started) and checkout completion rate (checkout started to order). Cut each by the segments from Step 1, then rank the leaks by the revenue a realistic improvement would recover: the extra orders times your average order value. The biggest drop is not always the most valuable one.

Don’t judge your store against the famous average. Baymard Institute’s cart abandonment rate of 70.22% is the simple average of 50 studies published from 2006 to 2025, whose results run from 55% to 84%^[3]. It says most carts are abandoned everywhere; it can’t say which of your stages is broken. Your own segments can.

The numbers show where shoppers leave; the pages show why. Audit on a phone first if most of your traffic is mobile, following a new visitor’s path from the landing page or homepage through category and product pages to the cart and checkout, and put real orders through, one with a card and one with an express wallet. Three checks earn their time:

• Clarity. Is it obvious what you sell and why you’re different? The Five-Second Message Test shows a page to people for five seconds, then scores whether they understood what it is, who it’s for and what to do next.
• Friction. Note every surprise and every extra step: costs that appear late, fields to fill in, an account to create, size or fit details that are hard to find, filters that don’t match how people choose, and an add-to-cart button below the first screen.
• Speed. Google’s Core Web Vitals measure loading (Largest Contentful Paint), responsiveness (Interaction to Next Paint, which replaced First Input Delay in 2024) and visual stability (Cumulative Layout Shift) from real Chrome users^[4]. PageSpeed Insights reports them for any busy page over the last 28 days, at the 75th percentile of page loads, so the score reflects your slower visits, not the average (Figure 3).

Speed is worth fixing, but read the promises made for it with care. The study most often quoted, by Deloitte and 55 for Google, followed 37 retail, travel, luxury and lead-generation sites for 30 days and found that when four speed measures each improved by 0.1 seconds, retail conversion rates rose 8.4% and retail spend 9.2%^[5]. Google paid for it, and it observed changes in speed rather than running a controlled test, so read it as a direction, not a forecast.

The free Website Conversion Audit is a fast first pass: about 50 checks across speed, mobile, conversion and analytics, weighted by what each costs you in conversions.

Put every finding from Steps 3 and 4 on one list and score each 1 to 10 for impact (how much revenue it could recover, from Step 3), confidence (how strong the evidence from Step 4 is) and ease (how little design and development it needs). Rank by the average of the three, then sort each line into one of three buckets:

• Quick wins: simple, low-risk changes you ship this week.
• Template changes: updates to the category, product, cart or checkout templates every shopper passes through; ship them and watch the stage rate.
• Experiments: changes where you can’t predict the answer, tested in Step 8 on pages with enough traffic.

Don’t test what is plainly broken; fix it. Then put the roadmap for the next six to twelve weeks on one page: each change, the stage it acts on, who owns it and roughly when it ships.

On a phone, shoppers judge a product by its first screen. Fix product pages in this order:

• Say what it is and why it’s worth it. A clear name, benefits rather than only features, and the facts that settle hesitation: size and fit, ingredients or composition, how to use it.
• Keep the price and the button together. Price, offers, bundles and stock sit on the first screen with the add-to-cart button, which is large, high-contrast and the only main action nearby.
• Answer cost and risk before the cart. Shipping cost, delivery time and returns in one line under the button.
• Put proof beside the decision. The average rating and review count by the price; full reviews and FAQs that handle common objections lower down.
• Keep the page light. A strong first image, and a gallery and review widget that don’t slow the page or make it jump as it loads.

On category pages, use filters that match how people choose (size, colour, price), consistent product cards with image, price and rating, and put bestsellers or higher-margin products near the top.

When Baymard Institute asks US online shoppers why they abandoned a checkout, extra costs top the list by a wide margin, and most of the other common reasons are things a store can fix (Figure 6). In the cart and checkout:

• No surprises in the cart. Clear line items, totals, shipping, tax and discounts, an obvious path to checkout beside a quieter “continue shopping”, and upsells that don’t crowd it.
• Guest checkout. Let people buy without an account and offer one after the order. Where an account is needed, as for a subscription, ask only for a password on the confirmation page.
• Fewer fields. Baymard’s benchmark found the average checkout had 11.3 form fields in 2024, while most sites need only 8. One full-name field, the second address line behind a link and billing set to the shipping address by default do most of the work^[7].
• Payment and trust. Offer the methods your shoppers use (cards, express wallets and, if your audience uses it, buy now, pay later), and show returns, security and how to reach support beside the payment step.

Test what you genuinely can’t decide, on the stage with the leak. Good starting tests for a consumer site: on product pages, the first image (in use against a product close-up), the order of the benefits and how reviews are shown (stars only against full snippets); on offers, bundles, subscriptions and how a discount is framed (percent off, dollars off, buy one get one); in the cart and checkout, the free-shipping message and the order of payment options. Use urgency only when it is true.

Write each test on one page before you build it: the hypothesis, the control and the variant, one primary metric (add-to-cart rate, checkout start rate, checkout completion rate or average order value), the sample you need, and a decision rule set in advance: ship, iterate or roll back.

The sample depends on your baseline rate and the smallest lift worth detecting, and it grows fast as that lift shrinks. Measure as close to the change as you can: a checkout change read on checkout completion needs a small fraction of the visitors it needs when read on sessions that end in an order (Figure 7). Run whole weeks, so weekday and weekend shoppers both count, and don’t stop early because the dashboard looks good.

Three things keep the gains.

A revenue dashboard. Sessions by channel and device, conversion rate (sessions to orders), average order value, revenue by channel and campaign, and the three stage rates from Step 3; views for new against returning customers, top regions and key categories or hero products; and a list of tests in progress with their results. Google Analytics with Looker Studio, or your own BI tool, is enough.

A review rhythm. A short weekly snapshot, a monthly conversion review and a quarterly strategy session, each with a fixed agenda (Figure 8), so conversion work stays a habit without becoming anyone’s full-time job.

Rules for changes. Name who can edit copy and images, change templates, change tracking, and start or stop tests. Decide which changes need a brief and a test and which can go straight live, and log every change with a date, an owner and a reason. Write down standards: UTM tagging, event names, how discount codes and offers are named, and the checks every new page passes, such as one main call to action, visible trust points and a test on a phone.

What changes by stage?

What mistakes should I avoid?

• Benchmarking against the 70% average. Compare your own segments instead (Step 3).
• Trusting analytics you haven’t reconciled. Check purchases against store orders first (Step 2).
• Auditing on a laptop. Review the site on the device most of your shoppers use (Step 4).
• A/B testing what is plainly broken. Fix it and move on (Step 5).
• Stopping a test when it looks good. Decide the sample and the end date before launch (Step 8).
• Editing a page while a test runs on it. Check the change log first (Step 9).