Austin Natural Mattress: AOV +88%, Orders +250%, Net Sales +582% — After a Recent Rebuild Made the Site Worse

When Shane of Austin Natural Mattress first paid us for an audit in June 2025, he’d been running the business for 25 years. Two stores — Austin (the flagship) and Houston — carrying Avocado, Naturepedic, Vispring, and his own in-house line. 99% of revenue runs through the showrooms. The website is the billboard that drives foot traffic and phone calls into the stores.

Shane wasn’t asking for an online revenue machine. He wanted a website on par with the brands he carried and the in-store experience he’d built. As he put it:

“The website needs to kick ass.”

Shane, Founder, Austin Natural Mattress

What happened between that audit and the case study you’re reading is the reason we don’t do rebuilds — and the clearest proof we’ve ever published that rebuilds, done by the wrong hands, can leave a site measurably worse than it was before.

The TL;DR:

• AOV up 88%, Orders up 250%, Net Sales up 582%. First 3 weeks post-launch (May 6–27, 2026) vs 3 weeks pre-launch on the previous rebuild.
• Mobile PageSpeed 28 → 97 on the homepage. 93 on category, 95 on product. SemRush site health 74 → 91 and climbing.
• The rebuild before us made things worse. LCP got slower (16.5s → 18.6s). Layout shift went from 0 to 0.464. Site health dropped from 90 to 74. Zero of the 250 UX audit recommendations were implemented. Order history was deleted in the process.
• Same site. No new URL, no new content, no broken customer data. We replaced the failing theme and resolved the audit findings against the Baymard 239-point standard. Same store, working properly.
• $50K mattresses. 99% showroom revenue. The website is the billboard that drives foot traffic. The trust gap on the previous site was the gap. We closed it.

Project went live May 6, 2026.

What the audit found in June 2025

We delivered two audits. The performance audit drew on 11 years and over 300 WooCommerce engagements — we know what slows these sites down and what resolves it. The UX audit ran against the Baymard Institute’s 239-point methodology, distilled from around 180,000 hours of research on the top ecommerce sites globally.

The findings, summarised:

• Mobile PageSpeed: 28 out of 100
• Largest Contentful Paint: 16.5 seconds (Google wants 2.5)
• 250 UX failures across product discovery, search, navigation, and trust signals

For a retailer selling $50K mattresses, the trust gap was the gap. A premium-brand buyer comparing the Austin Natural Mattress website against the Avocado website, the Naturepedic website, and the Vispring website wasn’t seeing a site that read as the dealer worth visiting in person. They were seeing a site that read as a reason to keep looking.

Shane’s existing developer was already mid-redesign when our audit landed. He gave her free rein to implement everything we’d recommended. Six weeks later the new site was live.

How the rebuild made things worse

We ran the post-launch numbers on December 2, 2025. The data was unambiguous:

• Mobile PageSpeed: 28 → 44. Improvement on the surface metric, but…
• Largest Contentful Paint: 16.5s → 18.6s. The metric that determines whether a visitor sees the page before they leave became 2.1 seconds slower.
• Cumulative Layout Shift: 0 → 0.464. Layout shift didn’t exist on the old site. The rebuild introduced it — pages where elements jump around as they load, frustrating for users, penalised by Google.
• UX recommendations implemented from our audit: 0 of 250.
• SemRush site health: 90 → 74. Shane and Amanda (his partner) had been monitoring this themselves.
• Google Analytics: still not configured to track revenue. Shane was flying blind on whether the new site was working at all.

On top of the SEO and performance damage, the rebuild had destroyed the order history. The previous site’s data didn’t carry over. Shane couldn’t compare the new site to the old site on any longitudinal sales metric. The information you need to evaluate a rebuild’s impact was the first thing the rebuild deleted.

This is the case against rebuilds in one project. We don’t do them, and this is exactly why.

What we said to Shane

On December 2 we wrote to Shane with the numbers. We didn’t pitch. We told him what we were seeing, included the verbatim data, and ended with one principle:

“I know you value relationships, Shane. I could see that in how you handled the dynamic on our calls. But when that relationship is costing you — in time, in quality, in opportunity — it stops being loyalty and starts being expensive.”

Campbell Angus, in an email to Shane, December 2, 2025

The developer’s own position, captured in our November call with her: she hadn’t gone into the UX audit deeply. Shane’s account exceeded her capacity. Her sub-contractor was running behind. None of this is held against her in the case study — it’s a structural problem, not a personal one.

Capable generalist developers hit the same ceiling on WooCommerce that capable owner-operators hit. The platform out-of-the-box has poor UX, no off-the-shelf theme addresses the purchase journey, and page-builder themes add performance overhead nothing resolves from the outside. An audit identifies the work. It doesn’t do the work. Handing a 239-point Baymard audit to a generalist developer mid-redesign is asking them to do something the platform itself isn’t built to support.

Three weeks after the email, Shane brought us in to take over.

What we did — without rebuilding the rebuild

We don’t do rebuilds. In 11 years, we never have. Austin Natural Mattress’s site after we shipped was the same site they came to us with — same URL, same content, same orders, same customer data. What changed was the underlying theme, the failing UX areas across the purchase journey, and the performance configuration. Same store. Working properly.

The work was diagnosing the specific failures against the Baymard guidelines and resolving each one.

UX changes across the purchase journey — mobile-first

Most of Austin Natural Mattress’s traffic was mobile. So was most of the broken experience. The audit findings were applied with mobile as the design lead, not as an afterthought to the desktop layout. Tablet and desktop followed.

• Header restructured. The previous version was consuming around 50% of the mobile screen before a product appeared. Reduced to the recommended 20%, with categories surfaced where customers would find them.
• On-site search replaced. New search with proper indexing and typo tolerance. Predictive results pulling categories first, then products, in the layout Baymard guidelines specify.
• Product page redesigned. For a $50K mattress purchase, trust signals do more work than copy. Customer reviews surfaced on product pages, descriptions delivered in both bullet and paragraph form (different buyers scan differently), supporting product information laid out for decision-making instead of placed below the fold.
• Cart and checkout redesigned per Baymard guidelines — guest checkout default, fewer fields, clearer steps.
• Visual polish. The design Shane’s previous developer shipped looked good. We kept what worked, replaced what didn’t, made the parts that mattered for conversion work the way the audit said they should.

Performance work supporting the UX changes

• Replaced the page-builder theme that had been carrying the rebuild’s performance debt
• Standard performance optimisation across homepage, category, product, cart, and checkout
• Mobile PageSpeed across homepage, category, and product pages: 97 / 93 / 95
• LCP back below Google’s target. CLS resolved.

Project went live May 6, 2026.

Results: AOV +88%, Orders +250%, Net Sales +582%

Comparing the first three weeks on the new site (May 6–27, 2026) against the three weeks immediately before launch (April 14–May 5, 2026) — the last three weeks of the previous rebuild:

Metric	Change
Average order value	+88%
Orders	+250%
Net sales	+582%
Items per order	+21%
Mobile PageSpeed (homepage)	28 → 97
SemRush site health	74 → 91

AOV climbed 88%. That’s where the UX work shows up most directly. For a high-end mattress retailer where individual mattresses run into five figures, the redesigned product page does the decision-support work the old one wasn’t doing — reviews surfaced, technical specifications presented for both scanners and readers, trust signals in the places where buyers make the decision.

Orders climbed 250%. Net sales climbed 582%. The combination is what compounds — more orders, each at a higher value, across a site that for the first time in months wasn’t actively losing buyers at the cart.

SemRush site health hit 91 and climbing — already above the 85–90 baseline before the previous rebuild dropped it to 74. Mobile PageSpeed 97 across the homepage, 93 on category, 95 on product. The platform fundamentals are now where Shane had been told they would be after the previous rebuild — six months and one failed project late.

If your WooCommerce store looks like this

If you’ve paid for an audit and handed it to your existing developer to implement, look at what has changed on the metrics — not what’s been promised. The audit identifies the work. It doesn’t do the work.

If you’ve gone through a recent rebuild and something feels off — the site looks newer but performs worse, or pages jump around as they load, or your site health score dropped after the launch — those are measurable problems. They don’t resolve themselves, and they don’t resolve with the same hands that built them.

You don’t need a new site. You need the one you have working properly. That’s what we do.

Frequently asked questions

Can a WooCommerce rebuild actually make things worse?

Yes — and it’s more common than the industry admits. Austin Natural Mattress’s previous rebuild improved the surface PageSpeed score (28 → 44) while making Largest Contentful Paint slower (16.5s → 18.6s), introducing Cumulative Layout Shift from zero to 0.464, dropping SemRush site health from 90 to 74, and deleting the historical order data needed to evaluate any of it. A rebuild that improves one metric while degrading three others isn’t an improvement. It’s a regression dressed as progress.

What’s the difference between PageSpeed and Largest Contentful Paint?

PageSpeed is a composite score from 0 to 100 that summarises a basket of metrics. Largest Contentful Paint (LCP) is one specific metric inside that score — it measures how long it takes for the largest visible element on the page to render. PageSpeed can improve while LCP gets worse if other component metrics offset the loss. For real-world conversion, LCP matters more than the score: a visitor decides whether to stay based on how fast the page appears, not on a composite number they’ll never see.

What is Cumulative Layout Shift and why does it matter?

Cumulative Layout Shift (CLS) measures how much the page moves around as it loads. When images load late and push the text down, or buttons appear after the user has already started reading, that’s CLS. Google penalises high-CLS sites in search rankings. Users tap the wrong thing because the page shifted between their decision and their tap. Austin Natural Mattress’s previous rebuild introduced a CLS of 0.464 — Google’s “good” threshold is 0.1. Anything above 0.25 is rated poor. The old site, before any rebuild, had a CLS of zero.

Should I let my existing developer implement a UX audit?

Only if they have specific WooCommerce purchase-journey experience and the bandwidth to actually do the work. Generalist developers hit the same WooCommerce platform ceiling that owner-operators hit — the out-of-the-box UX is poor, no off-the-shelf theme addresses the purchase journey, and page-builder themes add performance overhead that nothing resolves from the outside. An audit identifies the work. It doesn’t do the work. Handing a 239-point Baymard audit to a generalist developer mid-redesign is asking them to do something the platform itself isn’t built to support.

What happens to order history during a WooCommerce rebuild?

It depends on the rebuild approach. If the database is preserved, order history carries over. If the rebuild starts from a fresh database, customer orders, purchase history, and longitudinal sales data can be lost. Austin Natural Mattress’s previous rebuild destroyed the historical order data — meaning Shane couldn’t compare new-site performance against old-site baseline on any sales metric. The information needed to evaluate the rebuild’s impact was the first thing the rebuild deleted. Our approach preserves the database, the orders, and the customer data because we don’t rebuild — we work on top of what’s already there.

How do you measure if a website rebuild was successful?

Six measurements, in order of importance: (1) Core Web Vitals — LCP, CLS, INP — against Google’s thresholds. (2) Mobile PageSpeed across homepage, category, and product pages. (3) Baymard 239-point UX score on mobile and desktop. (4) Conversion rate, AOV, and items per order in the same period year-over-year or quarter-over-quarter. (5) SemRush site health vs the pre-launch baseline. (6) Revenue tracking in Google Analytics actually configured and reporting. If any of these went backwards after the launch, the rebuild has a problem the project lead may not be telling you about.

Why did Austin Natural Mattress’s AOV go up 88%?

The product page redesign did most of the work. For a $50K mattress purchase, trust signals matter more than copy — reviews surfaced on the product page where buyers make the decision, technical specifications laid out in both bullet and paragraph form (different buyers scan differently), supporting information placed for decision-making instead of below the fold. The cart and checkout redesign per Baymard guidelines closed the gap at the final step. For premium-brand retail, the trust gap is the AOV gap.

Is mobile-first design really different from mobile-responsive design?

Yes. Mobile-responsive means a desktop design that adapts to a smaller screen. Mobile-first means the mobile experience is the design lead, with tablet and desktop following. The difference shows up in the header (Austin Natural Mattress’s old header consumed 50% of the mobile screen before a product appeared — reduced to 20%), the navigation (categories surfaced rather than buried in a hamburger menu), and the product page (decision-support information sized for thumb-scrolling rather than print-catalogue scanning). When 70-80% of WooCommerce traffic is mobile, the mobile-first/responsive distinction is the difference between a site that converts and one that doesn’t.

Can a developer’s capacity be the real issue, not their skill?

Often, yes. Capable generalist developers running multiple client accounts hit a capacity ceiling that isn’t about skill — it’s about the time and depth a 239-point UX implementation requires. Austin Natural Mattress’s previous developer was capable, but Shane’s account exceeded her capacity, her sub-contractor was running behind, and she hadn’t gone into the UX audit deeply. That’s a structural problem, not a personal failing. A capable generalist developer working at capacity will deliver a worse outcome than a specialist team with the bandwidth to do the work properly. The platform requires both depth and focus that one-person shops can rarely sustain across accounts at this complexity.

What is SemRush site health and how does it relate to performance?

SemRush site health is a composite SEO health score from 0 to 100 measuring crawlability, indexability, internal linking, page speed, and on-page issues. A drop in site health after a rebuild typically indicates broken internal links, lost canonical tags, structured data regressions, or new performance issues affecting crawl budget. Austin Natural Mattress’s site health was around 85-90 before the previous rebuild dropped it to 74. After our work, it hit 91 and is still climbing. For a site driven by organic search and local foot traffic to brick-and-mortar showrooms, a site health drop after a rebuild is a measurable revenue risk.

The pattern: same site, working properly — without the rebuild that made it worse

Austin Natural Mattress is the fifth case study we’ve shipped where the headline outcome wasn’t more traffic — it was the same store finally able to convert the traffic it already had. AOV +88%. Orders +250%. Net sales +582%. Mobile PageSpeed 28 to 97. Done without rebuilding, without breaking the URL, without starting from scratch.

What makes this case study different is the proof on the other side of it — the six months between June 2025 (when we delivered the audit) and December 2025 (when Shane saw the numbers) is one of the cleanest controlled experiments we’ve ever published. Same audit, two different sets of hands implementing it. One set produced a site where LCP got slower, layout shift was introduced, site health dropped, and 0 of 250 recommendations got implemented. The other set produced what you see in the table above.

If you’ve just been through a rebuild and something feels off, audit what the rebuild actually delivered. The site looking newer doesn’t mean it’s performing better. The metrics either improved or they didn’t. The data doesn’t lie, even when the project status updates did.