The Last Mile Is Where Demand Changes Hands

1. Where the handoff starts to fail

Manufacturers invest heavily to create demand through advertising, co-op programs, dealer locators, and brand marketing. Most of that demand eventually routes through a dealer website, and increasingly, it arrives on a mobile device.

The problem is that the dealer layer is not performing the way the broader system assumes it is.

Across approximately 7,300 dealer websites in the flooring, tile, cabinet, and countertop trades, the median mobile Largest Contentful Paint (LCP) measured 8.7 seconds. Google considers anything under 2.5 seconds “good,” while performance above 4 seconds falls into the “poor” range.

We expected to find a tail of bad sites. We found the median.

What this suggests is that the final handoff between manufacturer-generated demand and dealer-level conversion is structurally weaker than most participants in the channel can currently observe.

2. How this started

This project began as an attempt to better understand dealer visibility and digital presence at the local level.

While researching dealer networks, a consistent pattern kept appearing: businesses with strong reputations, established showrooms, and credible local operations were often paired with websites that performed poorly under mobile conditions. In many cases the issues were not obvious from the dealer’s perspective. The site loaded adequately in-store, on office wifi, or on a newer device. Under those conditions, nothing appeared broken.

The customer experience looked different.

The typical customer interaction happens offsite, usually on a mobile connection with variable network quality and older hardware. Under those conditions, load delays become much more visible. The result is a type of operational blind spot where neither side of the interaction fully sees the failure.

The dealer often has no direct visibility into abandonment caused by performance degradation. The customer rarely reports it. Traffic simply disappears into the next available result.

At first this looked like a prioritization issue affecting a subset of businesses. As more data was collected, it became clear the pattern extended well beyond isolated cases.

The same issues appeared across categories, geographies, and platforms with enough consistency that the project evolved from individual auditing into population-level measurement.

That shift created a different question entirely:

What does the dealer layer actually look like, at scale, under standardized mobile conditions?

3. What the data shows

These findings are based on approximately 7,300 confirmed dealer websites. Methodology is described in Section 6.

Slow mobile performance is the norm

The median dealer website recorded an 8.7 second mobile LCP.

Fewer than 3% of sites met Google’s recommended mobile threshold.

This was not concentrated in one segment of the channel. The pattern held across:

• flooring retailers
• tile dealers
• cabinet showrooms
• countertop businesses

It also persisted across regions and web platforms.

At population scale, slow mobile performance appears to be a structural characteristic of the dealer layer rather than an isolated quality issue affecting a minority of sites.

Dealers are often evaluating a different experience than customers receive

Desktop and mobile results diverged significantly.

Approximately 71% of dealer sites passed Google’s desktop LCP threshold, while fewer than 3% passed on mobile. The median gap between desktop and mobile performance was roughly seven seconds.

Operationally, this matters because most internal evaluation happens under desktop or office-network conditions. That environment tends to mask the mobile experience customers actually encounter.

The result is a visibility gap where many dealers reasonably believe their websites are functioning adequately because the version they personally interact with performs far better than the one customers receive on mobile connections.

Most optimization effort appears to happen after load

One of the more consistent findings involved the relationship between LCP and CLS.

Many dealer sites performed reasonably well on Cumulative Layout Shift (CLS), meaning page layouts generally remained stable after loading. At the same time, those same sites failed mobile LCP measurements by a wide margin.

Roughly three out of four dealer websites measured within Google’s “good” range for CLS while simultaneously measuring “poor” for mobile LCP.

The implication is not that sites are entirely neglected. In many cases there is evidence of attention being paid to visual presentation and post-load usability.

The larger issue is that the page frequently does not render quickly enough for those improvements to matter under real mobile conditions.

4. Why this was difficult to see previously

Until recently, there was no reliable way to observe dealer-layer performance consistently across the channel.

Individual site audits existed. Vendors could report on sites they managed. Dealers could self-evaluate their own experience. But none of those approaches produced a population-level view under standardized testing conditions.

Two things changed:

• dealer discovery became scalable
• performance auditing became programmatic

Once those capabilities were combined, the broader pattern became easier to measure.

That matters because many decisions around digital programs, dealer support, and demand routing have historically operated under assumptions that were difficult to validate at scale.

This dataset does not solve every visibility problem in the channel, but it does make the operational condition of the dealer layer substantially easier to observe than it was previously.

5. The manufacturer layer

Manufacturers create and fund demand through national campaigns, dealer programs, co-op advertising, and search visibility initiatives. Much of that demand ultimately routes into dealer websites that manufacturers do not directly control.

The implicit assumption behind that system is that the dealer layer can absorb and convert traffic at a reasonable rate.

The dataset raises questions about that assumption under mobile conditions.

If the median mobile experience approaches nine seconds, then a significant portion of paid and organic demand is encountering friction at the exact point where conversion is expected to begin.

This does not mean individual dealers are failing operationally, nor does it suggest that specific manufacturer programs are ineffective in isolation. The issue appears more systemic than individual.

A common response is that these industries remain relationship-driven, and that local trust, in-person expertise, and showroom experience still determine outcomes. That is largely true.

The findings here are narrower.

The website increasingly functions as the first interaction in the buying process, particularly on mobile. If that first interaction performs poorly, the relationship may never begin in the first place.

It is also important to acknowledge what this dataset does not capture. Dealers with little or no search visibility are underrepresented because the discovery process relied heavily on search-surfaced businesses. Whether those dealers perform better or worse digitally is outside the scope of this study.

What this dataset does provide is a clearer operational view into the portion of the dealer network most manufacturer-funded demand currently flows through.

6. Methodology

The dataset was built before conclusions were drawn from it.

Defining the population

Dealer discovery used two primary methods.

The first surfaced approximately 21,000 candidate websites through search-based discovery using flooring, tile, cabinet, and countertop queries paired with geographic modifiers.

Each site was classified as DEALER, ADJACENT, or NON_DEALER using a structured language-model classification pipeline applied to extracted site content. Validation against a 100-site manually reviewed sample produced approximately 87% agreement.

Of the search-surfaced candidates:

• 33% classified as DEALER
• 37% classified as ADJACENT
• 18% classified as NON_DEALER
• 12% remained UNCERTAIN due to insufficient content for confident classification

A secondary discovery method added 926 websites sourced from a manufacturer-aligned dealer network. These were treated as confirmed dealers without additional classification.

Combined, the final population included approximately 7,300 confirmed dealer websites.

Because discovery depended heavily on search visibility, the dataset likely overrepresents dealers with at least some digital presence and underrepresents businesses operating primarily through referrals or offline relationships.

Measuring performance

Dealer websites were audited programmatically using Lighthouse measurements through the PageSpeed Insights API under standardized testing conditions aligned with Google Core Web Vitals methodology.

Mobile and desktop environments were tested separately.

These measurements represent lab conditions rather than field telemetry collected from live users.

Testing measurement stability

A 200-site sample was re-tested three times per site to evaluate variance.

LCP measurements showed a mean variation of approximately 19%, while CLS demonstrated greater instability. Population medians shifted by roughly 8% under repeated testing conditions.

Headline statistics and supporting language were adjusted accordingly where variance meaningfully affected interpretation.

The exact values move somewhat depending on methodology and timing. The broader distribution pattern remained consistent across repeated measurement.

Scope limitations

This paper does not claim a direct revenue impact from improving site speed.

It does not assign blame to individual dealers or vendors.

It does not prescribe a single operational solution.

The goal is narrower: to describe how the dealer layer currently behaves under standardized mobile testing conditions at population scale.

7. Closing

Manufacturers generate demand with the expectation that dealer networks will convert it into showroom visits, calls, consultations, and sales opportunities.

What this dataset suggests is that the digital handoff between those layers is weaker than most participants in the system currently realize.

The issue is not limited to a handful of poorly maintained websites. At population scale, the pattern appears structural.

The last mile of the customer journey is where demand changes hands.

Right now, under mobile conditions, it is also where a meaningful portion of that demand is likely being lost.