How to Benchmark SiteGround’s Speed and TTFB in 2026 (Full Testing Protocol)

Disclosure: This post contains affiliate links; we may earn a commission at no extra cost to you.

SiteGround has a credible performance toolkit: Google Cloud infrastructure, dynamic and static caching, CDN options, the SG Optimizer plugin, and tier-dependent resources. Those features make it benchmark-worthy, but they are not measurements. No verified raw test output was supplied for this article, so no load time, TTFB, or uptime figure below is invented.

Instead, this is a reproducible SiteGround benchmark specification. It shows how to measure StartUp and GrowBig fairly, separate cache effects from hosting capacity, and interpret results using established web-performance thresholds.

Questions the benchmark must answer

A useful test should determine:

1. How fast is a cold WordPress response?
2. How much does SiteGround Dynamic Cache improve repeat delivery?
3. How consistent are median and slow-tail results?
4. What happens when several uncached requests arrive together?
5. How does server location affect distant visitors?
6. Do real-user Core Web Vitals confirm laboratory findings?

One PageSpeed score cannot answer all six.

Record the environment

Document the exact plan, purchase date, data-center region, PHP and WordPress versions, theme, plugins, database size, Site Tools settings, SG Optimizer configuration, CDN status, and test dates.

StartUp and GrowBig should not be described as one generic “SiteGround server.” GrowBig commonly supplies more resources plus staging and additional backup features. If comparing tiers, clone the same fixture and change no unrelated setting.

Create the test fixture

Use a representative WordPress installation containing:

• A 1,500-word article
• Responsive hero image and thumbnails
• Category archive and site search
• Contact form
• One font family
• A fixed navigation menu
• A defined plugin list
• A WooCommerce catalog for dynamic testing

Record total bytes and request count for every benchmark page. Export the database and media so another host can receive an identical copy.

An empty default theme tests little beyond the platform’s fastest possible path.

Benchmark four cache states

Run four labeled conditions:

1. Dynamic Cache off
2. Dynamic Cache on but cold
3. Dynamic Cache on and warm
4. CDN or edge delivery enabled and warm

Purge caches only before the cold series. For warm tests, request the page and verify a cache hit through response headers or SiteGround’s tools. Do not assume installation of SG Optimizer proves the response was cached.

Exclude cart, checkout, account, and personalized pages from full-page cache. Test those separately.

Measure TTFB correctly

Collect at least 30 samples per location and cache state across multiple periods. Use the same URL, protocol, redirect path, and test agent.

Report median, p75, p95, minimum, maximum, and failures. Google offers a rough field guideline of 0.8 seconds or less for good TTFB, 0.8–1.8 seconds as needing improvement, and above 1.8 seconds as poor. TTFB is not itself a Core Web Vital, but it precedes rendering.

Separate DNS, TCP, TLS, and time-to-first-byte timings when the tool supports them. A slow distant connection is not automatically slow PHP.

Run browser measurements

Use WebPageTest or another controlled browser platform from at least three regions: near the origin, across the same continent, and in a distant target market. Fix device, browser, network profile, and run count.

Capture:

• Largest Contentful Paint
• Interaction to Next Paint when field data exists
• Cumulative Layout Shift
• First Contentful Paint
• Total Blocking Time in lab tests
• Page weight and requests
• Waterfall and filmstrip

Google’s good field thresholds are LCP at 2.5 seconds or less, INP at 200 milliseconds or less, and CLS at 0.1 or less at the 75th percentile.

PageSpeed Insights field data represents a trailing real-user window when sufficient Chrome data exists. Lighthouse lab data is simulated and varies; do not mix the two as if they were the same dataset.

Test dynamic work

Cached articles reveal delivery efficiency, not database capacity. Test WordPress search, login, dashboard actions, uncached query pages, and sandbox WooCommerce cart or checkout.

Begin with one user, then ramp gradually through 5, 10, and 20 concurrent virtual users. Use realistic pauses and stop if error rate or latency rises sharply. Run only against an authorized staging clone.

For each stage, publish throughput, median, p95, p99, HTTP errors, timeouts, and any visible resource metrics. A fast error response is still a failed request.

Compare StartUp and GrowBig fairly

Deploy the identical fixture to both tiers in the same region. Match software versions and recommended platform settings. Repeat tests at similar times across several days.

Test StartUp with its available features and GrowBig with its normal supported stack. Do not cripple GrowBig to create artificial equality; instead, disclose that its resources and cache features differ.

The useful output is the marginal benefit: how much stability, dynamic capacity, staging value, and backup capability the higher renewal buys.

Isolate SG Optimizer changes

Create a baseline, then enable one feature group at a time:

1. Dynamic caching
2. Browser caching and compression
3. Front-end minification
4. Media optimization and lazy loading
5. CDN integration

Retest after each change. Combining every toggle at once makes attribution impossible and can introduce visual or JavaScript problems.

Inspect pages after minification and defer settings. A higher score is not a win if menus or forms break.

Uptime monitoring

Monitor from multiple external regions for at least 30 days. Use a public page and, where possible, a lightweight endpoint that distinguishes application failures from DNS or TLS problems.

Record HTTP 5xx, timeouts, TLS failures, DNS failures, maintenance, and plugin-caused errors separately. State the check interval and observation period.

A month with no recorded outage does not prove annual 100% uptime. Report only the observed window.

Results template

Replace “Pending measurement” only with raw, timestamped evidence. Publish the test manifest and exclusions beside the table.

Common misleading practices

Do not compare SiteGround’s cached homepage with another host’s uncached checkout. Do not select the fastest run, omit failed requests, change themes, or test origins on different continents without disclosure.

Avoid translating SiteGround’s estimated visit figures into guaranteed concurrency. Visits per month, requests per second, and simultaneous dynamic users are different quantities.

Optimize from evidence

If TTFB improves dramatically when warm, caching is working but origin processing deserves attention. If TTFB is good and LCP is slow, optimize images, CSS, fonts, and scripts. If p95 latency climbs under uncached load, inspect plugins, queries, external APIs, and plan capacity.

Choose a nearby data center, use supported PHP, compress images, limit plugins, and reduce third-party code. Retest with the original protocol after each material change.

SiteGround versus Bluehost

SiteGround emphasizes daily backups, Site Tools, SG Optimizer, and tiered cache resources. Bluehost emphasizes guided WordPress onboarding, an eligible first-year domain, CDN and caching features, and a broad beginner workflow.

For a valid speed comparison, clone the fixture, align origin geography and software, use each platform’s supported cache stack, and run the same tools and schedule. Compare long-term price and recovery tools alongside milliseconds.

Final verdict

SiteGround offers a strong performance stack, especially for owners who use Dynamic Cache, SG Optimizer, and a nearby origin. A responsible 2026 benchmark conclusion still requires raw results from a disclosed plan.

The protocol above measures typical speed, slow-tail behavior, dynamic saturation, field experience, and uptime without confusing features with facts. Until those runs exist, “Pending measurement” is the only honest result.

Frequently asked questions

Is SiteGround fast?
It has capable infrastructure and caching, but finished-site speed depends on plan, location, configuration, and code.

Does GrowBig outperform StartUp?
It has additional resources and features, but the benefit should be measured with an identical fixture.

Should Dynamic Cache stay enabled?
Usually for public WordPress pages. Exclude personalized and transactional routes appropriately.

How many runs should a test use?
Use at least 30 timing samples per condition plus repeated load profiles and multi-week uptime monitoring.