597 Small-Business Websites: Houston vs. San Salvador (2026 Study)

Short version: on September 14, 2026, we sent automated HTTP requests to 597 small-business websites listed on OpenStreetMap in Houston, Texas and San Salvador, El Salvador. Of those, 469 returned a 2xx or 3xx response and formed the base for the home-page measurements below. Almost all declared a mobile viewport and loaded over HTTPS. The bigger gaps were in contact and search signals: about half lacked either an HTML form element or a WhatsApp signal on the home page, a third of Houston sites and six in ten San Salvador sites were missing an H1 or a meta description, and roughly one in ten of all 597 listed domains returned no HTTP response.

Key findings

• Non-responding sites are common. 7% of the 350 Houston domains and 13% of the 247 San Salvador domains returned no HTTP response after an HTTPS request and an HTTP fallback, each with a 30-second timeout. The test did not determine whether the cause was DNS, hosting, an expired domain or a temporary outage.
• Many home pages lack an obvious contact signal in the HTML. 48% of Houston sites and 35% of San Salvador sites had neither a <form> element nor a detected WhatsApp signal on the home page. A known booking-platform signal appeared on 18% of Houston sites and 3% of San Salvador sites.
• WhatsApp signals are much more common in San Salvador. The HTML contained a WhatsApp link or reference on 47% of San Salvador sites, against 12% in Houston. A <form> element appeared on 47% in both cities; the test did not classify the form type.
• Analytics code is common, while Tag Manager is less common. We detected Google Analytics 4 signals on 77% of Houston sites and 72% of San Salvador sites; Google Tag Manager appeared on 34% and 26%. This study inspected code presence only and did not test whether forms, calls or WhatsApp clicks were configured as conversion events.
• Basic on-page SEO is still missing. 36% of Houston home pages and 61% of San Salvador home pages lacked an H1 or a meta description. Structured data (JSON-LD) was present on 64% of Houston sites and 44% of San Salvador sites. The test detected markup but did not identify which plugin or system generated it.
• Speed is no longer the differentiator. Median time to first byte is 0.51 s in Houston and 0.64 s in San Salvador; the slowest quarter of San Salvador sites take more than 1.19 s just to start responding. Only 4% (Houston) and 9% (San Salvador) take more than 3 s to deliver the HTML.
• WordPress dominates, especially in El Salvador. 39% of San Salvador sites run WordPress (24% with Elementor); in Houston it is 28% (15% Elementor), followed by Squarespace (9%) and Shopify (8%). Roughly one in ten sites in both cities still shows a copyright year of 2023 or earlier.

The numbers, side by side

Percentages in the comparison table use the 469 sites that returned a 2xx or 3xx response: 276 in Houston and 193 in San Salvador. The final row uses all 597 tested domains.

What this means if you run a small business

The technical floor has been reached. HTTPS, a mobile layout and a fast server are now table stakes that almost everyone has; nobody wins a customer with them anymore. The gap is in what happens after the page loads.

In Houston, the site that lets a visitor book, request a quote or message the business from the home page is in the minority. Half of your competitors make the visitor hunt for a phone number. A form that fires a conversion event in Analytics, a booking link and a clear service page structure are the cheapest advantages available.

In San Salvador, WhatsApp is the strongest contact signal we detected, while GA4 appeared on roughly three quarters of the sites and Tag Manager on one in four. That creates an opportunity to verify whether WhatsApp and form interactions are measured, but this study did not inspect event configuration. The other gap is basic search structure: six in ten home pages had no H1 or no description.

In both cities, one in ten sites announces a copyright year of 2023 or earlier, and a visible share of listed sites are simply dead. A visitor who finds a stale or broken site does not call; they go to the next result.

A five-point check for your own site

1. Open your home page on a phone. Can a visitor contact you in one tap, by form, WhatsApp or booking, without scrolling to the footer?
2. Does that contact register as a conversion in Google Analytics 4? Verify the event in DebugView or a test report; the presence or absence of Tag Manager alone does not answer this.
3. View the page source. Is there one H1 that says what you do and where, and a meta description written for a person?
4. Is the copyright year current, and does every page in your map listing and social profiles actually load?
5. Does the site answer the questions a customer asks before buying: prices or ranges, service area, hours, what happens after they contact you?

Methodology and limits

We queried OpenStreetMap for businesses tagged with a website in the Houston urban area and the San Salvador metropolitan area, removed national and international chains, franchises, banks and social-media links, and kept one entry per domain. The source OSM exports should be retained with the dataset for full reproducibility. From Houston we selected 350 domains using Python with random.seed(7); from San Salvador we kept all 247 unique domains. On September 14, 2026, a curl request with a Chrome-like User-Agent fetched each home page once, followed redirects and recorded status, time to first byte, total transfer time, HTML size and simple HTML signals. If HTTPS returned no response, the script tried HTTP; “no response” means neither request completed within its 30-second timeout. Of 597 domains, 469 returned a 2xx or 3xx response and were used for the comparison percentages; 71 returned 4xx/5xx and 57 returned no response. A viewport meta tag does not prove a good mobile design. A <form> element was not classified by purpose; WhatsApp was detected by link or text patterns; booking was detected from a limited provider list. Results are per home page and one request, not a Lighthouse or Core Web Vitals audit, and do not assess design or content quality. The sample is biased toward businesses maintaining a map listing. Download the measured dataset (CSV, 597 rows). The original OSM exports used to build the candidate list are not included, which limits full source-list reproduction.

If you want a plain-language review of where your site stands against these numbers, send us the URL. For the work itself: web design, WordPress development and AI automation.