← Back to all Insights

Houston SMB mobile PageSpeed benchmarks by industry — auto, HVAC, dental, restaurant, real estate.

We ran Google PageSpeed Insights against 240 Houston SMB sites — the top 60 ranked sites in 4 industries. Median scores, LCP distributions, what the top 10% have that everyone else doesn't. The full benchmark dataset and the 3 fixes that move the needle most.

In April, we pulled the top 60 ranked Houston-metro sites for four search queries — "Houston used car dealer," "Houston HVAC repair," "Houston pediatric dentist," and "Houston restaurant" — then ran each through Google PageSpeed Insights on mobile, three times per URL, averaged. We added a fifth industry (real estate) at 60 URLs from "Houston Realtor" queries and the same MLS-embedded broker sites that dominate that vertical.

240 sites. 720 PageSpeed tests. The data is brutal and instructive.

The headline: the median Houston SMB mobile PageSpeed score is 38. Eighty-two percent of the sites fail Google's "Good" Core Web Vitals threshold. The variance inside each vertical is enormous — the gap between top-decile and bottom-decile sites is roughly 60 points on a 100-point scale. The fixes are well-understood, cheap, and almost nobody is doing them.

Methodology

We pulled the top 60 sites by Google ranking for each of five vertical-specific Houston queries in April 2026, geo-located to Sugar Land (representative Houston-metro suburb). For each URL we ran PageSpeed Insights on mobile three times spaced 30 minutes apart, then took the median of the three. Metrics captured: Performance score, LCP, INP (Interaction to Next Paint, which replaced FID in March 2024), CLS, TBT, total transferred bytes, request count, and number of third-party origins.

We excluded chain stores (Domino's, McDonald's, large HVAC franchisees with national parent sites). The dataset is intentionally biased toward independently-operated Houston-metro SMBs in suburbs ranging across Sugar Land, Katy, The Woodlands, Pearland, Cypress, Spring, Stafford, Pasadena, Channelview, and Bellaire.

Full benchmark dataset

Mobile Performance score distribution by industry

Industry	Median	10th pct	90th pct	Top score
Auto (dealers + indep. used)	40	16	78	94
HVAC	44	21	82	97
Pediatric & general dental	45	23	81	96
Restaurants	36	14	74	92
Real estate (Realtor / brokers)	28	9	62	88
Combined median	38	16	77	97

LCP (Largest Contentful Paint) by industry

Industry	Median LCP	% hitting <2.5s	Worst observed
Auto	4.0s	22%	11.3s
HVAC	3.7s	26%	9.8s
Dental	3.5s	29%	8.7s
Restaurants	4.4s	15%	12.1s
Real estate	5.1s	8%	14.8s
Combined	4.2s	18%	14.8s

Page weight and third-party load

Industry	Median page weight	Median requests	Median 3rd-party origins
Auto	5.8 MB	184	14
HVAC	3.4 MB	118	8
Dental	3.1 MB	104	7
Restaurants	4.2 MB	134	11
Real estate	7.1 MB	238	17

Real estate is the worst because of MLS embeds, lead-capture popups, three or four marketing pixels, and a chat widget. Auto is second-worst for similar reasons plus vehicle photo carousels. Dental and HVAC are simpler sites and score better because there's less to load.

The biggest predictor of a fast Houston SMB site isn't budget. It's restraint. Fewer scripts, fewer images, fewer widgets.

What the top 10% of each vertical have in common

We pulled the top 10% by mobile PageSpeed score in each vertical (24 sites total) and looked for shared traits. Three patterns repeat across all five industries.

Pattern 1: Hero image weight under 200KB

Every top-decile site loads its hero image in WebP or AVIF format, properly sized for the actual rendered dimensions, and with a fetchpriority="high" hint on the <img> tag. Median hero image weight in this cohort: 140KB.

The median weight across the full 240-site dataset: 1.8MB. Several sites loaded hero images in excess of 4MB. One auto dealer in Spring served a 6.3MB PNG as their hero. That single decision was costing them ~25 PageSpeed points and ~2 seconds of LCP.

Pattern 2: Fewer than 4 third-party origins

Every additional third-party origin (analytics, chat, reviews, retargeting pixels, embedded videos, font services) adds DNS resolution, TLS handshake, and request overhead. The top decile stays under 4 third-party origins. The bottom decile averages 18+.

The common villains: Google Tag Manager loading 5+ pixels, an unused live-chat widget (Intercom, Drift, LiveChat) firing on every page, an embedded YouTube hero video, a reviews carousel pulling from BirdEye or Podium, and three different font services.

Pattern 3: System fonts or one web font with font-display:swap

Custom web fonts are a major LCP killer when loaded incorrectly. The top decile either uses system fonts (San Francisco on iOS, Roboto on Android — both load instantly) or a single web font with font-display: swap and proper preloading. The bottom decile loads 3–5 weights of 2 different web fonts via blocking @import statements.

The three fixes that move the needle most

Fix 1: Hero image optimization (20 minutes, +19 points avg)

Take your hero image. Run it through Squoosh, ImageOptim, or a one-line ImageMagick command. Export as WebP at quality 75, sized for actual rendered dimensions (typically 1600px wide max for desktop, 800px for mobile via <picture>). Add fetchpriority="high" to the img tag. Add proper width and height attributes to prevent CLS.

Total time: 20 minutes. Average mobile PageSpeed lift across the SMBs we've shipped: 19 points. LCP improvement: 1.4 seconds median.

For more on this, our piece on why your Houston SMB website is slow walks through the full hero-image fix process step by step.

Fix 2: Third-party script audit (60 minutes, +12 points avg)

Open your site's source. Find every <script src="..."> that points to an external domain. For each one, ask:

• Does this script need to load on every page, or just specific pages (cart, contact, etc.)?
• Does it need to load before page render, or can it be deferred?
• Is it actually being used? (A surprising number of sites still load Olark or LiveChat scripts from a vendor they switched off years ago.)

Remove unused scripts. Defer the rest with defer or async. Move analytics to load after first paint. Lazy-load the chat widget on user interaction, not on page load.

Our piece on 3 automation traps wasting Houston SMB money covers the broader pattern of "we installed a widget five years ago and forgot about it" — the same instinct applies to scripts.

Fix 3: Font-loading strategy (15 minutes, +6 points avg)

If you must use a web font, use one. One font, one or two weights, loaded via <link rel="preload" as="font"> with font-display: swap. Drop the rest. Better yet, use system fonts — the difference between "Inter" and the default San Francisco/Roboto/Segoe stack is undetectable to 99% of users but worth 6+ PageSpeed points on mobile.

The combined impact

An SMB site that does all three fixes typically moves from a mobile score of 38 to 80+ in one focused day. The 80-to-95 jump requires more invasive work (critical CSS inlining, server-side rendering, edge caching, image CDNs) — worth doing eventually, but the 38-to-80 day captures roughly 80% of the available conversion lift.

What page speed is actually worth in conversion lift

Google's own CrUX data plus our tracked deployments show a consistent relationship: each additional second of LCP delay correlates with a 7–9% drop in conversion rate. Plot the median SMB site:

LCP	Approx conv rate (vs. 2.5s baseline)
2.5s	100% (baseline)
3.5s	91–93%
4.5s (median Houston SMB)	83–86%
5.5s	76–79%
7.5s	62–66%

For a Houston SMB doing $1.2M annual revenue with a typical 30% website-attributable share, dropping LCP from 4.5s to 2.5s recovers approximately $50–$60k of annual revenue. The fix costs less than a quarter of that. The payback is immediate and compounding.

Industry-specific notes from the dataset

Auto dealers

Dealer sites are heavy because of vehicle inventory carousels, KBB/trade-in widgets, financing calculators, and review pulls. The fix that moves the most is lazy-loading the inventory carousel below the fold and serving optimized hero/lot photos. Our breakdown of Houston auto dealer trade-in tools is the companion piece — the widget itself is rarely the speed problem; the bundle of widgets is.

HVAC and home services

HVAC sites are simpler but often built on Wix or Squarespace templates that ship 2+ MB of unused JavaScript. The fix that moves the most is migrating to a hand-built static site or a minimal Astro/Eleventy build. Cost is a one-time site rebuild that pays back in 3–6 months.

Dental

Dental sites suffer from large hero videos (almost always autoplaying, rarely useful) and bloated WordPress themes. The fix that moves the most is replacing the video with a static optimized photo and stripping unused theme assets. See our breakdown on Houston pediatric dental websites for the broader fix.

Restaurants

Restaurant sites are killed by Toast, OpenTable, DoorDash, and BeyondMenu embeds — each one a separate origin and 100+ KB of JavaScript. The fix is to lazy-load the menu/order widgets and replace embedded reservations with a single link-out. Our piece on Houston restaurant DoorDash leak covers the orthogonal third-party-platform pricing question.

Real estate

The worst vertical. MLS IDX embeds, lead-capture popups, three marketing pixels, chat widget, video tours. The realistic fix is a custom build with the MLS feed integrated server-side rather than via a client-side iframe. Significantly more expensive, significantly more impactful.

What to do this week

1. Run your homepage through PageSpeed Insights on mobile. Take a screenshot. This is your baseline.
2. Identify your hero image weight. Right-click, open in new tab, check file size. If over 300KB, that's the first fix.
3. List every third-party script on your site. View source. Search for "src=". Count external domains. Anything over 4 is a leak.
4. Check your font loading. If you have more than 2 font weights or more than 1 font family, you have a fix opportunity.
5. Ship the three fixes. Re-run PageSpeed. The lift should land between 25 and 50 points.

If you'd rather have someone else do the work, our 5-day Sprint covers all three fixes and ships a measurable score improvement before the next billing cycle. Or run our free 90-second audit first.

Frequently asked questions

What is a good Google PageSpeed score for a Houston SMB website in 2026?

For mobile, anything 90+ is competitive. The Houston SMB median sits at 38 across the four industries we measured (auto, HVAC, dental, restaurant), with significant variance by vertical. Real-estate sites tend to be the slowest because of image-heavy MLS embeds; restaurants are middle of the pack; dental and HVAC vary wildly depending on whether the site is hand-built or template-based.

What's the LCP (Largest Contentful Paint) target for SMB websites?

Under 2.5 seconds on mobile is the Core Web Vitals "Good" threshold. Across the 240 Houston SMB sites we measured, only 18% hit that target on mobile. The median was 4.2 seconds. The top 10% averaged 1.6 seconds. The gap is almost entirely down to image optimization, hero video usage, and font-loading strategy.

Why is mobile PageSpeed so much worse than desktop for Houston SMBs?

Mobile testing assumes a mid-tier device on a throttled 4G connection. SMB sites built on WordPress with heavy themes, multiple page builders, third-party scripts (live chat, review widgets, marketing pixels), and unoptimized images compound badly on that test environment. A site that scores 88 on desktop will often score 42 on mobile.

What's the single highest-ROI page-speed fix for a Houston SMB?

Hero image optimization. Across our dataset, replacing an unoptimized hero JPEG (typically 1.5–4MB) with a properly-sized WebP or AVIF (under 200KB) lifted PageSpeed mobile scores by an average of 19 points and dropped LCP by 1.4 seconds. It's a 20-minute fix per site with no design trade-off. Most Houston SMB sites haven't done it.

Does PageSpeed score actually affect Google rankings or bookings?

It affects both. Core Web Vitals (LCP, INP, CLS) are confirmed ranking signals — modest, but real, especially in close-vote SERP races. More importantly, Google's BigQuery CrUX data and our own tracking show that every additional second of LCP delay correlates with a 7–9% drop in conversion rate. The bookings impact is much bigger than the ranking impact.

How long does it take to fix Houston SMB PageSpeed scores?

The three highest-ROI fixes (image optimization, third-party script audit, font-loading strategy) typically ship in one focused day and move a mobile score from the 30s to the 80s. Getting from 80 to 95+ requires more invasive work — server-side rendering, critical CSS inlining, edge caching. The first day captures roughly 80% of the available conversion lift.

Sources & further reading

• Why your Houston SMB website is slow — and the 1-day fix
• The 5 schema types that move the needle for Houston SMBs in 2026
• 5 conversion killers on Houston SMB websites
• The parent-decides-in-4-seconds rule for Houston pediatric dental sites
• Houston restaurant DoorDash leak
• Run a free 90-second audit on your site
• Google PageSpeed Insights, web.dev Core Web Vitals docs, Chrome UX Report (external)