Case Study: How a 40,000-Product Auto Parts Store Cut Page Weight by 71% Before Peak Season

Last spring, a reader who runs an independent auto parts ecommerce operation — let's call him Marcus, since he asked us not to use his real name — sent us a frustrated email. His storefront listed roughly 40,000 SKUs, from brake rotors to infotainment harnesses, and every product page carried four to six high-resolution photos. Google's Core Web Vitals had been flagging his category pages for months, and his mobile conversion rate was sliding. He'd already upgraded hosting twice. The bottleneck wasn't the server. It was the images.

We followed the project from intake to results, and it turned into one of the cleaner case studies we've documented on automotive ecommerce performance. It also became our first hands-on look at Pixlog, an AI image optimization platform that compresses, reformats, and CDN-delivers images automatically.

The Starting Point: Great Photos, Terrible Payload

Marcus's catalog photography was genuinely good — clean white-background shots, plus install-angle images for the repair-guide content he publishes alongside many listings. The problem was that those files were shipping as full-resolution JPEGs, most between 1.2 MB and 3.5 MB. His median page weight on product templates sat around 6.8 MB, and Lighthouse performance scores hovered in the mid-50s on mobile. A category page listing 24 parts could cross 12 MB.

He had tried a manual optimization workflow: exporting smaller JPEGs in batches, then re-uploading. It worked for about two weeks. Then new inventory arrived, the process broke down, and the old heavyweight files quietly crept back in. That's the classic failure mode we see with manual image pipelines — they don't survive contact with a busy operations schedule.

Decision Point: Automate or Keep Tweaking?

In late April, Marcus evaluated three paths:

  • Manual re-exporting — free, but already proven unsustainable at his catalog size.
  • Self-hosted compression scripts — more control, but required ongoing maintenance and a developer on retainer.
  • An automated optimization layer — a service that would intercept, compress, convert, and deliver images without touching his theme code.

He chose the third option, partly because his team had no dedicated front-end developer. The setup took an afternoon: he pointed the service at his image domain, and delivery switched over progressively rather than all at once. No template edits, no plugin conflicts with his parts-lookup widget. For a store running on a heavily customized theme, that mattered — we've seen image plugins break fitment filters before.

What Actually Changed Under the Hood

Three technical moves drove most of the gain. First, format conversion: legacy JPEGs and PNGs were served as WebP (and AVIF where the browser supported it), typically cutting file size 60–80% at visually identical quality. Second, AI-driven compression that adapts per image rather than applying one global quality setting — a flat product shot on white compresses far more aggressively than a textured engine-bay photo before artifacts appear. Third, CDN image delivery, which pushed assets to edge nodes near the shopper instead of serving everything from a single origin.

The combination is where the platform earns its keep. Compression alone helps; compression plus edge delivery plus automatic format negotiation is what moves real-user metrics. According to the write-up on how the optimization pipeline works, the median page-weight reduction across sites using it is 71%, with Lighthouse scores climbing from the 50s into the 90s.

Obstacles Along the Way

The rollout wasn't frictionless. Two issues surfaced in week one.

The first was a handful of zoom-on-hover galleries where customers expected pixel-level detail on part numbers stamped into metal. Marcus's team flagged about 200 SKUs and excluded their hero images from aggressive compression. That was a judgment call, not a platform failure — but it's the kind of nuance any workshop or parts retailer should plan for.

The second was caching: his existing CDN layer held onto old image URLs for several days, so some shoppers saw mixed results during the transition. Purging the cache resolved it, and after that the delivery path was consistent.

Measurable Results After 60 Days

By mid-July, roughly seven weeks after switching on, the numbers looked like this:

  • Median page weight: 6.8 MB down to about 1.97 MB — a 71% reduction.
  • Mobile Lighthouse performance: 54 → 92 on product templates.
  • Largest Contentful Paint on category pages: 4.1 s → 1.8 s.
  • Mobile conversion rate: up 19% year over year, against a flat prior quarter.
  • Bounce rate on mobile product pages: down 11 percentage points.

Marcus was careful not to attribute all of that to images alone — he also cleaned up a third-party chat widget in the same window. But the timing lined up: the biggest movement in Core Web Vitals happened in the two weeks after image delivery switched over. With Pixlog handling delivery automatically, his team stopped thinking about image weight entirely, which is arguably the real win for a lean operation.

What We Took Away

For anyone running an automotive ecommerce catalog — parts, accessories, tools — the lesson isn't "compress your images." Most teams already know that. The lesson is that image weight is an operational problem, not a one-time design task. Catalogs grow, seasonal inventory lands, and manual pipelines decay. Automating the compression, conversion, and delivery layer removes an entire class of recurring work.

We'd add one caveat: audit your detail-critical images before you flip the switch. Fitment diagrams, stamped part numbers, and wiring-pin close-ups deserve a quality floor. Everything else can be optimized hard. Do that, and the performance gains tend to show up faster than most teams expect — often within a single billing cycle.