Good is the minority: dealer site speed, Core Web Vitals, and the script tax

A vehicle detail page took 6.2 seconds to show a photograph of the car. That is a field number, not a lab score: 75th percentile, mobile, a mid-range Android two generations behind the current handset, on cellular in a mid-size metro. The shopper had already tapped back, opened a competitor's listing and read a price. The dealer's page eventually loaded perfectly.

This is the part of retail marketing that no creative review touches. A page that loads is not a feature. It is the precondition for every other feature, and on dealer sites it is usually the thing that has been traded away one vendor request at a time.

What is actually being measured

Core Web Vitals are three numbers taken from real Chrome usage rather than a test device in an office, reported at the 75th percentile and segmented by device class. A URL passes when three quarters of its real loads clear the thresholds.

Interaction to Next Paint replaced First Input Delay in March 2024, and that single change explains why a great many sites that passed for years stopped passing. First Input Delay measured only the delay before the first interaction could be handled. Interaction to Next Paint measures every interaction on the page and takes the worst one. A site whose first tap was fine but whose third tap stalled behind a chat widget was never fast. It was unmeasured.

The sample, and what it says

For this piece we measured 3,300-plus dealer domains, pulled from brand store-locator files rather than from a vendor's customer list, using mobile field data over a trailing 28-day window. A domain counted as good only if all three metrics cleared their thresholds at the 75th percentile. Under that rule, good is the minority: most dealer domains fail at least one metric, and the failures are not evenly spread. Cumulative Layout Shift is the one most sites pass. Largest Contentful Paint and Interaction to Next Paint are where the sample breaks.

That distribution tells you where the cause is. Layout shift is a design discipline problem, and most dealer themes have solved it. Largest Contentful Paint on a vehicle detail page is one element, almost always the hero photograph or the inventory module, and both are typically lazy-loaded or injected by JavaScript that has to arrive first. Interaction to Next Paint is main-thread contention, which in practice means the total JavaScript on the page and the order in which it wakes up.

Neither of those is a design decision. Both are a consequence of the vendor list.

Method note: these are point-in-time figures, so re-run the sample before quoting a number in a review. The thresholds are Google's published ones, unchanged. The point of the number is the shape of the distribution, not the third significant figure.

The experiment that removes the usual excuse

The standard objection is that fast sites are fast because they belong to better operators, and better operators sell more cars. That is a real confound, and it is why cross-sectional speed data can never prove causality on its own.

One national brand answered it with a randomized experiment across its dealer pages. Visitors were assigned at random to receive the same page through a faster delivery path or through the standard one: same inventory, same offers, same photography, same forms. Assignment was random, so the two groups were alike in every respect except delivery. The faster arm produced more shopper action, including more vehicle-detail views per session and more lead submissions.

That design is the whole value of the study. It converts a correlation everyone already believed into a causal claim you can defend to a fixed-ops director who does not care about milliseconds. If your brand's digital team ran something similar, ask for the readout and cite the design, not a lift percentage lifted from a summary slide.

The script tax

Open a dealer vehicle detail page in a browser and count the third-party origins. A typical set looks like this.

Every one of these was bought to answer a question. Not one of them was bought with a millisecond budget attached, and none of the people who bought them owns the total. That is the script tax: a page assembled by eight stakeholders, each optimizing their own line item, with nobody accountable for the sum.

The pattern is visible in the metric that breaks first. Tags that fire on idle are expensive but survivable. Tags that boot on scroll, re-render a header, or hold the main thread for 300 milliseconds when a shopper taps a trim selector are what push Interaction to Next Paint past 500 milliseconds, and the shopper experiences that as a site that ignores them.

What a speed program actually looks like

1. Measure field data per template, not per site. A vehicle detail page, a search results page, a service scheduler and a parts lookup are four different products with four different budgets.
2. Write the budget down: transferred weight, compressed JavaScript, third-party origin count, per template. Review against it the way you review against brand guidelines.
3. Keep a tag register with a named owner per tag and the question it answers. Any tag without an owner is removed. This is the same register as the stack audit, and it is the single highest-leverage document in the whole discipline.
4. Fix the three things that weigh most, in order: font loading, hero imagery, and the boot sequence of third-party scripts. Most dealer pages are fixed by deferring what does not need to run before the first paint.
5. Attack the HTML latency you can control: server response time and edge caching for the document itself. A cached document that arrives in 200 milliseconds changes what the rest of the page can do.
6. Refuse the speed tool that solves speed by adding a script. There are legitimate performance vendors. A tag that claims to optimize the page while becoming part of the page is not one of them.
7. Prove the change in the field with a holdout: hold a third of traffic or a control set of rooftops, change one template, and compare 28 days of field data and leads. Without a holdout you have an opinion with a chart.

The uncomfortable part

Good Core Web Vitals will not sell a car on their own. There is no version of this where a 2.1-second Largest Contentful Paint becomes a headline in a monthly review. What speed does is stop taxing everything else you already pay for: paid clicks that land on a page nobody waits for, crawl budget spent on templates that time out, merchandising work on vehicle descriptions a shopper never reaches, and a lead form that loads after the decision was made somewhere else.

It is also the only conversion lever in the building that can be fixed without changing price, offer or inventory. That is rarer than it sounds, and it is why the deck has a section on it.