01

What has to keep working, and who decides

Before anything changes on a dealership website, the things that must never break are written down and agreed with the dealership. Inventory and pricing, the lead forms, consent and privacy behavior, accessibility for keyboard and assistive-technology users, and whatever else that specific store depends on to trade.

That list is the dealership's, not ours. It is agreed in writing before installation, it can be added to at any time, and nothing on it is touched. Consent, accessibility and pop-up widgets get particular attention, because those are the failures that turn a performance change into a legal problem.

02

What happens when something goes wrong

The original dealership website is always the fallback. If Speed Engine cannot operate safely on a page, it stops and the site the dealership already has is what the shopper gets: unchanged, with nothing missing. A shopper never sees a broken intermediate state, and the dealership is never left worse off than before installation.

This isn't a promise we're hoping holds up: it's built into how the product works, and we check it on your specific website before you ever depend on it. We can't predict every possible failure. If something does fail, you land back on the site you already had.

A map of what a shopper receives in each of three situations: when the speed layer can operate safely on a page, when it cannot, and when the dealership has asked for it to stop. Two of the three paths end on the original dealership website, unchanged. The drawing describes designed behavior; it states no measurement and no result.
Where each of the three paths ends. Two of the three end on the original dealership website, unchanged, which is what fail open means in practice.
03

What gets checked before a dealership relies on it

A faster page is not an acceptable result if something a shopper needs stopped working. Every item above is exercised on the dealership's own site, on real devices, and the outcome is recorded. Before the change is live for shoppers, not after.

  • Navigation, search, filters, pagination and scrolling
  • Vehicle media, pricing, incentives and availability
  • Chat, trade, retailing, service, finance and lead forms
  • Consent, keyboard focus, reduced motion and mobile behavior
  • The analytics events the dealership has said it needs
04

How a dealership stops it

One instruction, and the site returns to exactly what it was. There is no unwinding, no migration back and nothing left behind, because the original website was never replaced. A dealership that wants Speed Engine gone gets the site it had.

Who to call, how fast we answer and what we commit to belong in the agreement covering a specific implementation, not on a public page: they differ by dealership. What gets measured afterwards is in the measurement methodology.

What can be committed to, and what can't
CommitmentWhat it means
The original site is the fallbackIf Speed Engine cannot operate safely, the shopper gets the site the dealership already has, unchanged
What must not move is agreed in writingThe protected list is set for that specific website before installation, not a generic promise that applies to every dealership
Checked on your own site, not a demoEvery safeguard is exercised on the dealership's own website, on real devices, before it is relied on
Removable in one instructionTurning it off doesn't wait on a software update or a scheduled release
05

What is the worst thing that can go wrong?

A lead form that looks normal and silently stops submitting. Not a visible error: a form that accepts input, appears to send, and never arrives. The failure can run for days before anybody notices. Nothing on the dealership's side looks broken, and it is expensive in a way no speed improvement offsets.

Every failure worth worrying about has the same shape. An optimization layer changes when something runs. Another script depended on that timing. The tool that breaks does not announce itself. Chat that loads but never connects an agent. A trade valuation that returns nothing. A consent banner that never fires: a compliance problem, not a performance one.

The safeguards are written as acceptance criteria, not as reassurances. A vendor who cannot describe their failure mode has not thought about it, and the dealership discovers it in production.

06

How is the safeguard verified rather than asserted?

By exercising the functions rather than reading the metrics. Before an implementation is accepted, the shopper paths that matter are run by hand on the optimized experience. Filter inventory, open a gallery, reveal pricing, start a trade valuation, open chat, complete a lead form. Then confirm the consent behavior still fires. A faster trace with a broken lead form is not a result, it is a regression with good numbers attached.

We check every safeguard on your specific site before you rely on it, not once, and not as a formality. What has to stay untouched differs at every dealership. It depends on your platform, your tools and how your store actually sells cars. A safeguard that works on one rooftop only tells you about that rooftop. The safe sequence for third-party script work walks through that same check at the level of a single script.

Sources and further reading

External sources support the general technical guidance on this page. They do not represent a DealerSpeed Engine performance result.