Conversion rate optimisation for ecommerce: the honest sequence that actually moves the number

CRO demystified (it's not button colours)

"Conversion rate optimisation" sounds technical, and the industry leans into that mystique to sell expensive services. The useful version is much simpler: find where visitors drop off, figure out why, fix the cause, measure the result. That's it. Button-colour A/B tests are the least important part.

Most stores that "do CRO" spend months A/B testing trivia (button text, hero image variants, pricing-font choices) and move conversion by 0.3%. The stores that actually move the number by 30-100% in a year do the unglamorous work: fix the checkout form, speed up the mobile experience, rewrite the product pages that are bleeding traffic.

This post is the honest version of CRO. The sequence that actually works for most stores.

1. Know your starting number (and what it means)

"Our conversion rate is 2%." OK, but 2% of what? The number only matters if you're measuring the same thing consistently.

• Site conversion rate: orders ÷ total sessions. Includes everyone, useful for overall trend
• New-visitor conversion rate: orders ÷ new sessions. Isolates acquisition effectiveness
• Returning-visitor conversion rate: orders ÷ returning sessions. Isolates retention effectiveness
• Product-page conversion rate: product-page-viewers who buy. The most actionable
• Cart-to-checkout conversion rate: people who start checkout ÷ people who add to cart. Isolates checkout friction
• Checkout completion rate: people who complete ÷ people who start checkout. Isolates final-step friction

Generic benchmarks:

• Under 1%: something is broken. Fix before optimising
• 1-2%: average
• 2-3%: good
• 3-5%: strong
• 5%+: excellent (or your traffic is too narrow to trust the number)

Most "my store has low conversion" calls turn out to be "my store has a checkout problem" or "my product page is confusing". Segment the number before trying to move it.

2. Measure the full funnel, not just the overall rate

Overall site conversion is a single number with many hidden failures. The funnel breaks it down:

• Visitor → product page view (%)
• Product page view → add to cart (%)
• Add to cart → checkout start (%)
• Checkout start → checkout complete (%)

Where visitors drop off tells you what to fix:

• Visitor → product page weak: navigation / category layout / search is the issue
• Product page → cart weak: product page itself (copy, photos, price, social proof) is the issue
• Cart → checkout weak: shipping cost surprise or checkout complexity is the issue
• Checkout start → complete weak: form friction, payment issues, or trust signals is the issue

Shoprocket's Conversion Funnel analytics tracks this automatically. export the data from Google Analytics 4 into a spreadsheet; the analysis is the same.

3. Pick fixes by leverage, not by ease

Most stores optimise the easiest thing to change (button colour, hero copy) rather than the thing with the most leverage. Leverage is: how many visitors are affected, multiplied by how big the fix could be.

High leverage, usually:

• Mobile page speed (affects 70% of traffic; can lift conversion 10-30%)
• Shipping cost transparency (affects every cart; can lift 10-20%)
• Checkout form simplicity (affects everyone who checks out; can lift 5-15%)
• Product photography consistency and quality
• Product page copy (first line especially)

Low leverage, usually:

• Button colour tests
• Homepage hero carousel slide order
• Footer link placement
• About page copy

Start with leverage. A single high-leverage fix beats 20 low-leverage tests.

4. Make one change at a time

Most stores implement CRO as "redesign the site and measure". Then conversion changes, and nobody knows which of the 40 changes caused it.

Rule: one test at a time. Measure for long enough to be statistically meaningful (see next point). Implement the winner. Move on to the next test.

This feels slow. It's not. It's honest. "We redesigned and conversion went up 20%" is a story with no learning. "We changed the product page hero to show the product in use, conversion went up 12%" is a learning you can reapply.

5. Sample size math: the boring part that stops you wasting months

A/B test running 2 weeks with 200 sessions per variant, variant B up 15% vs variant A: not a winner. That's within normal week-to-week variance.

You need real sample size to conclude:

• For a store at 2% baseline conversion
• To detect a 20% improvement (to 2.4%)
• With 95% confidence

You need roughly 3,500-4,000 sessions per variant. For a store doing 10,000 sessions/month, that's a 3-week test. For 1,000/month, that's 7-8 months. At which point the test is wasted: the traffic mix changes, the season shifts, the result is noise.

Stores under 1,000 monthly orders shouldn't A/B test most changes. Make the change based on qualitative evidence (heatmaps, session recordings, user feedback) and ship the version you think is better. Measure trend, not test outcome.

A/B testing becomes genuinely useful past 5,000-10,000 orders per month. Under that, qualitative research beats quantitative testing.

6. Heatmaps and session recordings beat A/B tests for small stores

Tools like Hotjar, Lucky Orange, or Microsoft Clarity (Clarity is free) record sessions and generate heatmaps. What they tell you:

• Where visitors click vs where you want them to click
• What they scroll past vs what they stop on
• Where they struggle (rage-clicks, repeated scrolling)
• Which form fields make them bounce

For small stores, this qualitative evidence tells you more per hour than any A/B test. Watch 20 session recordings of visitors on your product page. You'll see confusion, misclicks, and drop-off moments. Fix the top 3 and ship.

7. The 5 highest-ROI CRO fixes for most stores

After watching a lot of stores go through this, the fixes that pay back most reliably:

1. Mobile speed under 3 seconds. Compress images, remove unused scripts, lazy-load below-fold content
2. Show shipping cost on product page, not at checkout step 3
3. Simplify the checkout form. Guest checkout default, one page or few steps, auto-fill friendly
4. Rewrite the first line of your top 5 product pages as specific benefits (see product descriptions post)
5. Add trust signals near the Add to Cart button. Returns policy, review rating, guarantee

If you haven't done these, don't bother with anything fancier. These are the 80/20.

8. What to measure post-fix

Don't declare victory on day 1. Give every change 2-4 weeks to settle before judging.

Track:

• Overall conversion rate (week-over-week)
• The specific metric your fix targeted (product page CR, checkout completion, etc.)
• Revenue per session (sometimes conversion goes up but revenue per session goes down; net negative)
• Bounce rate on the page you changed
• Time on page (decrease can be good or bad; combine with other metrics)

Don't move on until the fix has clearly worked or clearly not. Accepting "meh, might have helped" wastes the learning.

9. When to hire a CRO agency

Honestly: usually not.

• Under $1M annual revenue: the ROI isn't there. Agency fees eat the conversion gain. Do it yourself
• $1M-$10M: maybe, if you don't have in-house resources. Vet carefully; many CRO agencies sell tests, not outcomes
• $10M+: yes, almost certainly. The leverage is big enough to justify a dedicated team or specialist agency

Red flags in CRO agencies:

• Pitching button-colour A/B tests as the core offering
• Promising specific percentage lifts before seeing your data
• Not asking about your funnel data before proposing tests
• Wanting to run 20 tests simultaneously
• Not willing to share their statistical method

Good CRO agencies start by reading your analytics and talking to you before recommending anything. They propose tests with expected lift and duration. They don't promise magic.

What Shoprocket helps with

• Real-time analytics with visitor tracking
• Advanced analytics (traffic, geo, device, entry/exit points)
• Conversion funnel visualisation
• UTM campaign tracking for attributing conversions to sources
• Automatic sitemap + product schema for SEO-driven traffic that compounds over time
• Abandoned cart recovery to capture the biggest single drop-off in most stores
• Apple Pay + Google Pay + Stripe Link for one-tap mobile checkout (massive mobile CR lift)
• Image optimisation (automatic WebP, resize, lazy-load) for page speed

<!-- EXAMPLE ONLY. Grab your real snippet from Sales channels → Embeds in your dashboard -->
<script src="https://cdn.shoprocket.io/loader.js" data-pk="pk_yourkey"></script>
<div data-shoprocket="catalog" data-embed-id="emb_xxx"></div>

What Shoprocket doesn't ship: native A/B testing on product pages (use Google Optimize replacement like VWO or Optimizely alongside), heatmap/session-recording tools (use Hotjar, Clarity, or Lucky Orange).

Tip: Before booking an agency or subscribing to a CRO tool, do this for free: watch 10 mobile session recordings on your top-revenue product page (Clarity is free). List the 3 moments of confusion you see. Fix them. Measure for a month. You'll often get more lift than a $3,000/month agency engagement would deliver.

TL;DR

• Segment your conversion rate. Overall is a bad signal; product-page and checkout-completion are actionable
• Measure the full funnel. The drop-off point tells you what to fix
• Pick fixes by leverage. Mobile speed + shipping transparency + checkout simplicity > button colour tests
• One change at a time. Avoid the redesign-and-measure trap
• Sample size math is real. Under 10,000 monthly orders, A/B testing is usually noise
• For small stores, use heatmaps and session recordings. Qualitative research beats quantitative testing
• The top 5 fixes cover most of the available lift. Do them before anything fancier
• CRO agencies make sense at $1M+ revenue, not before

Start a free trial and get funnel + attribution analytics from day one. 14 days, no card.