Every VPS review on ProBlogGuru is based on a server we paid for, deployed ourselves and measured with the same set of tools. This page explains exactly what we measure, how we measure it, how the scores on our review pages are calculated, and where the limits of our testing are. If a number appears on a review page, the date it was recorded and the plan it was recorded on are shown next to it.
What we test on every provider
We deploy the provider’s entry-level plan (typically 1 vCPU, 1–2 GB RAM, 20–25 GB SSD) because that is the plan most of our readers start on. We do not test on reviewer-upgraded or provider-supplied instances; every account is opened with our own details and our own payment method, and we pay the normal price or use the same free trial any new customer can use.
Unless a review says otherwise, the server runs Ubuntu [24.04 LTS] with a default WordPress install on [LiteSpeed / NGINX] and no page cache, so we are measuring the server, not a caching plugin.
Test locations. For global providers we use the New York region. For Europe-focused reviews we use Frankfurt, and for country-specific pages (for example, Turkey or Hungary VPS), we use the provider’s data center in or nearest to that country and say which one on the page.
The five things we measure
1. Time to First Byte (TTFB)
How long the server takes to start responding to a request. We measure it from [tool, e.g. KeyCDN Performance Test / GTmetrix] from the test location nearest the server, run it [three] times, and report the median. TTFB is the number that most directly reflects the server itself.
2. Page load: First Contentful Paint (FCP) and Largest Contentful Paint (LCP)
Measured on the same default WordPress page with [GTmetrix / PageSpeed Insights lab test]. These tell you how the server feels to a visitor. They are lab numbers, not real-user field data, and we say so on every review.
3. Uptime
We monitor the test server with [UptimeRobot / HetrixTools] at [5-minute] intervals for a minimum of 30 days before publishing. The uptime percentage on a review is the figure the monitor reported for that window, and the screenshot of the monitor is embedded on the page. We do not round 99.95% up to 100%.
4. Load handling
We run a [Loader.io] test that ramps to [10,000] concurrent connections over [one minute] and record the error rate and the average response time under load. This is a short stress test, not a simulation of a month of real traffic.
5. Setup, dashboard, and support
We record how long deployment took, whether the control panel let us do the basics (rebuild, snapshot, firewall, console) without opening a ticket, and how the provider’s support responded to one real, non-urgent question we asked during the trial. These are our observations, and they are described in words on the review rather than converted into a precise number.
How the review scores are calculated
Review pages show six category scores out of 10 and an overall score. Here is what each category means and how the overall score is weighted.
| Category | What it’s based on | Weight |
|---|---|---|
| Performance | TTFB, FCP, LCP and load-test results from the tests above | 30% |
| Reliability | Measured uptime over the monitoring window, plus published SLA and redundancy | 20% |
| Features | OS choice, one-click apps, snapshots, firewall, scaling, Windows availability | 15% |
| User-friendliness | Deployment time and what we could do from the dashboard without support | 15% |
| Support | Channels available, response time and quality of the answer to our test question | 10% |
| Pricing | Entry price, billing model (hourly / monthly / prepaid), trial or credit availability | 10% |
Performance and reliability scores are set from measured numbers. Features, user-friendliness, support and pricing are judgement scores, and two members of the team set them independently before we publish; where we disagreed, the review says so. A score is never changed because of a provider’s commission rate, and providers do not see scores before publication.
Dates, updates, and re-testing
- Every benchmark on a review page carries the date it was run and the plan it was run on.
- The “Updated” date on a review changes only when we changed something in the content, and the review states what changed.
- We re-run the performance tests when we update a review, and at least [once a year] for providers in our main comparison lists. If a number on a page is more than [12 months] old, the page says so.
- If a provider changes its trial terms, pricing or credit amount, we update the review as soon as we can verify the change on the provider’s own site. If you spot something outdated, tell us via the contact page and we’ll check it.
What we don’t do
- We don’t publish provider marketing figures as test results. If a number came from a provider’s website rather than our own test, it’s labelled as such.
- We don’t test every plan a provider sells. Our numbers apply to the plan named on the review.
- We don’t measure from your location. Latency to your users depends on where they are; use the data center location we name to judge whether it suits your audience.
- We don’t run long-term production workloads on the test servers. Our tests are point-in-time measurements plus a 30-day uptime window.
- We don’t accept payment for placement in our comparison lists or for a higher score.
Affiliate links and independence
ProBlogGuru is reader-supported. Some links on our reviews are affiliate links, and if you sign up through them, we may earn a commission at no extra cost to you. This is how we pay for the servers we test. It does not change what we measure, how we score, or which provider we rank first, and we also review and link to providers that pay us nothing. Our full policy is on the affiliate disclosure page.
Corrections
If we get something wrong, we fix it and note the correction on the page. Contact us through the contact page with the URL and what you believe is incorrect.
Who runs the tests?
Ravi Kumar deploys and benchmarks the servers and writes the reviews. Narendra Singh reviews the server configuration and the results before publication. Questions about a specific test can be sent to either of us via the contact page.