Two Thirds of Australian Retail's Page Weight Isn't Theirs

We ran Google Lighthouse against thirty of Australia’s largest online retailers. Twenty-one returned valid measurements.

Not one scored 50 or above. The median was 30. Every site failed the Largest Contentful Paint threshold.

That much we expected to be bad. What we did not expect was the reason. The median retailer in this sample ships 7.1MB to a mobile homepage, and 4.1MB of it — 62% — comes from domains they do not control.

The headline numbers

• Sites scoring ≥50: 0 of 21.
• Sites passing LCP (≤2.5s): 0 of 21.
• Sites passing CLS (≤0.1): 15 of 21.

That last line is the interesting one, and we will come back to it.

The full table

Lighthouse 12.6.1, mobile, homepage, September 2026. Sorted by performance score.

Site	Perf	LCP	Page weight	Third-party	3P share	Requests
petcircle.com.au	44	4.0s	9,306kB	6,177kB	66%	360
templeandwebster.com.au	44	7.0s	6,017kB	5,923kB	98%	319
chemistwarehouse.com.au	42	8.6s	6,304kB	4,132kB	66%	273
showpo.com	41	8.1s	4,973kB	4,453kB	90%	250
adairs.com.au	40	9.5s	3,047kB	1,235kB	41%	140
bunnings.com.au	37	26.5s	7,324kB	2,474kB	34%	372
theiconic.com.au	37	8.7s	6,883kB	4,235kB	62%	306
petbarn.com.au	34	15.3s	4,029kB	2,824kB	70%	255
babybunting.com.au	31	18.4s	7,128kB	4,700kB	66%	285
harveynorman.com.au	31	13.6s	5,747kB	2,981kB	52%	276
anacondastores.com	30	8.9s	10,379kB	2,190kB	21%	311
bcf.com.au	30	17.1s	5,555kB	2,386kB	43%	255
kmart.com.au	30	52.7s	14,036kB	10,953kB	78%	339
universalstore.com	28	33.4s	24,907kB	7,471kB	30%	819
princesspolly.com.au	27	25.4s	18,431kB	14,500kB	79%	707
myer.com.au	26	10.0s	6,285kB	3,510kB	56%	267
supercheapauto.com.au	26	10.3s	18,089kB	2,713kB	15%	300
thegoodguys.com.au	26	10.8s	3,405kB	2,605kB	77%	323
jbhifi.com.au	23	37.8s	8,574kB	6,388kB	75%	531
sportsgirl.com.au	22	64.6s	30,830kB	10,530kB	34%	385
rebelsport.com.au	2	34.9s	8,253kB	2,969kB	36%	414

These are not badly built sites

This is the finding that changed how we read the data.

If these were sloppy builds, layout stability would be the first casualty — images without dimensions, fonts swapping late, content jumping as the page assembles. Instead 15 of 21 sites pass Cumulative Layout Shift comfortably, with a median of 0.018 against a 0.1 threshold. Somebody did that work. It does not happen by accident.

So the picture is not incompetence. It is competent teams building competent storefronts, and then loading four megabytes of other people’s code on top.

Every single site runs the same two things

Across 21 retailers, we found 36 distinct third-party companies on the median site. Two appear on every single one:

Google Tag Manager and Facebook together account for over ten seconds of main-thread blocking across the sample. Not page weight — blocking, the time where the page is visible and does not respond to a tap.

Worth being precise about what this does and does not show. Tag Manager is a container: much of what it costs is the tags loaded through it, not the container itself. But that is rather the point. It is the mechanism by which a marketing team can add a script to a production storefront without a developer, a deploy, or a performance review — and 36 companies per site is what that mechanism produces over a few years.

The extremes are instructive

templeandwebster.com.au: 98% third-party. Of 6MB shipped, roughly 94kB is the retailer’s own. Their score of 44 is among the best here — the site itself is light. It is carrying almost nothing but other people’s code.

sportsgirl.com.au: 30.8MB and a 64.6-second LCP, driven by 17.8MB of images. This is not a tracking problem, it is an image pipeline problem, and it is the most straightforwardly fixable site in the sample.

supercheapauto.com.au: 15% third-party — the cleanest ratio here — and still only 26, because it ships 13.9MB of images.

rebelsport.com.au scored 2, with CLS of 0.937. That is close to the entire page being rebuilt after first paint.

The lesson from the spread is that there is no single Australian retail performance problem. There are two: too much third-party JavaScript, and unoptimised images. Most sites have one badly. A few have both.

A megabyte of JavaScript that never runs

The median site ships 1,145kB of unused JavaScript — downloaded, parsed, and never executed on that page load. On a mid-range Android on mobile data, that is real seconds spent on code that does nothing.

This is the signature of accumulation. Every script had a reason. Marketing needed a tag manager. Growth added an A/B testing tool. Personalisation went in for a campaign that ended two years ago and was never removed. No single addition looks like the problem, nobody owns the total, and the site still feels fine on the laptops of the people who build it.

The commercial read

The obvious conclusion is that Australian retail has a speed problem. The useful one is different.

Not one site in this sample scored 50. If every large competitor in your category is between 20 and 44, then loading properly is not hygiene — it is a differentiator, in conversion and in organic search, where Core Web Vitals remain a ranking signal none of these sites are sending well.

And the fix is more tractable than a replatform. The median site here would gain more from an audit of its 36 third-party tags than from any amount of storefront rework. That is a governance exercise as much as an engineering one: who can add a script, who reviews the cost, and who removes the ones whose campaign ended.

Method, and what this does not show

• Tool: Google Lighthouse 12.6.1, mobile emulation, default simulated throttling.
• Scope: homepage only, September 2026, run from Sydney.
• Sample: 30 attempted, 21 valid.

Exclusions, and why they matter. Seven sites were dropped: six returned no score (blocked automated browsers or exceeded our time limit), and catch.com.au was excluded because it served a bot-challenge interstitial — 25 requests, a <div class="announcement"> as its largest element, HTTP 202. It would have scored 79 and been the only site above 50 in this study. Including it would have been the most misleading number on the page.

Three limitations we would want a reader to hold:

Homepages are not the whole story. Product and category pages are where buying happens and they behave differently. This measures the front door.

The excluded sites skew slow, not fast. Several dropped out because Lighthouse timed out, which correlates with being slow. The measured set is therefore unlikely to be the slower end of the distribution, so these figures are conservative.

Lab data indicates, it does not prove. Lighthouse simulates a device and a connection. Field data from real users is the authority, and only the site owner has it. A retailer with a poor lab score and healthy field data is in better shape than this table suggests.

Every figure here is reproducible, and deliberately so. Run:

npx lighthouse "https://<site>/" --only-categories=performance \
  --chrome-flags="--headless=new" --output=json

The medians above come from the resource-summary and third-party-summary audits in the resulting report. Point it at any homepage in the table and you will get numbers close to ours — allowing for the variance any lab measurement carries between runs, and for the fact that these sites change weekly.

---

If you are on this list and the number surprises you, your own Search Console field data is the place to check next. If it agrees, the first move is an audit of what is loading and why — not a rebuild. Performance work on Adobe Commerce and Shopify is a large part of what we do, and we are happy to talk it through.