How to Choose WordPress Hosting: Shared, Managed, VPS or Cloud?

Learning how to choose WordPress hosting properly is the highest-leverage technical decision most site owners ever make, and almost nobody makes it deliberately. People pick the host their first developer used, or the one with the loudest discount, and then spend years wondering why the site feels slow, why it goes down on the one day a campaign runs, and why every support ticket takes three days to answer. Hosting is the floor under everything else: caching, optimisation and clean code can only work as well as the server underneath them lets them.

I inherit sites in every hosting situation you can imagine – overloaded shared accounts, expensive managed plans running brochure sites, unmanaged VPS boxes nobody has patched since 2022 – because agencies and small businesses in Germany, the UK, the US and Australia hand me their WordPress problems. This guide is what I have learned from 12+ years and 500+ projects: the four hosting types honestly compared, the specs that actually matter and the marketing fluff that does not, the benchmarks and support tests you can run yourself before paying, realistic pricing by site type, and the red flags that predict pain. No affiliate links, no “top 10” list – brand names appear only as honest examples.

Table of contents

• Why hosting is your highest-leverage decision
• The four types of WordPress hosting, honestly compared
• Specs that matter vs marketing fluff
• Benchmarks you can run before you buy
• The support test: open a ticket before you pay
• Do not host your email with your web host
• Moving between hosts without drama
• Pricing reality by site type
• Red flags that predict a bad host
• Server location, CDN and international audiences
• The decision table: match your site to a hosting type
• Frequently asked questions

Why hosting is your highest-leverage decision

Three things about your website are decided almost entirely by your host, and no plugin can fully compensate for any of them.

TTFB: the head start every page load gets

Time to first byte is how long the server takes to start answering. On a good host, an uncached WordPress page returns its first byte in 200-400ms; on an overloaded shared server I regularly measure 800-1,500ms – before a single image, stylesheet or script has loaded. Everything else in performance work stacks on top of that number. I have moved identical sites – same theme, same plugins, same content – from a crowded shared account to a decent managed plan and watched TTFB drop from 1.1 seconds to 250ms with zero code changes. That is the kind of win that months of image optimisation cannot buy, and it is why hosting sits at the top of my complete WordPress speed guide.

Uptime: the outages you never hear about until they matter

“99.9% uptime” sounds close to perfect but still allows almost nine hours of downtime a year – and cheap hosts measure it generously. The pattern I see on bottom-tier hosting is not one dramatic outage but hundreds of 30-second stalls at busy hours, when the server your site shares with 500 others runs out of breath. Your monitoring never catches it; your visitors feel it constantly.

Support: the difference between a bad hour and a bad week

When the site is down at 9am and you do not know why, the host’s support quality is suddenly the only spec that matters. Good hosts put an engineer on chat who can read logs; bad hosts put a script-reader on chat who asks you to clear your cache. You can test this before buying – I show you how below – and you should, because it is the spec that no comparison table ever prints.

The four types of WordPress hosting, honestly compared

Shared hosting: cheap because you are sharing the kitchen

Shared hosting (Hostinger, SiteGround, Namecheap and hundreds of others, typically EUR 3-12 per month) puts your site on a server with dozens to hundreds of other accounts. The honest version: for a low-traffic brochure site with good caching, decent shared hosting is genuinely fine, and I tell clients that even when it earns me nothing. The real cost of the cheapest tier is overcrowding – your TTFB depends on what your neighbours are doing, CPU limits throttle you during traffic spikes, and support is stretched thin. The better shared hosts limit accounts per server and give you staging and server-level caching; the worst oversell aggressively and quietly restrict you with CPU-seconds and inode limits buried in the fine print.

Managed WordPress hosting: what you are actually paying for

Managed WordPress hosting (Kinsta, WP Engine, Rocket.net, SiteGround’s higher tiers, typically EUR 20-60 per month for one site) charges four to ten times shared prices. What the money buys, when the host is good: server-level page caching (so cached pages skip PHP entirely), one-click staging environments, automatic daily backups with one-click restore, PHP workers reserved for you rather than shared, security hardening and malware cleanup included, and support staff who actually know WordPress. For a business where the website generates revenue – leads, bookings, sales – this is usually the right answer, because one avoided outage or one afternoon of not debugging a broken update pays the annual difference. For a five-page site nobody updates, it is overkill, and I say so.

VPS: control and responsibility in the same box

A virtual private server (Hetzner, DigitalOcean, Vultr – EUR 5-40 per month for the server itself) gives you dedicated resources: your own RAM, your own CPU share, nobody else’s WooCommerce sale slowing your checkout. The catch is the word “unmanaged”: you are now the system administrator, responsible for security patches, firewall rules, PHP upgrades, backups and 3am outages. Panels like RunCloud, Ploi or GridPane (EUR 8-30 per month on top) automate most of it and are what I use when a client needs VPS economics without a sysadmin. Managed VPS from a traditional host splits the difference but often costs as much as managed WordPress hosting while delivering less WordPress-specific help. My rule: choose a VPS when you need the resources or the control – many sites on one server, unusual software, strict data-location requirements – not to save money, because your time is the hidden line item.

Cloud and platform hosting: for specific problems, not for everyone

Above these sit platform setups: autoscaling cloud (AWS, Google Cloud, or Cloudways as a friendlier layer over them), and edge-heavy architectures where Cloudflare Enterprise or a headless front end absorbs most of the traffic before it reaches WordPress. These solve real problems – six-figure monthly traffic, global audiences, traffic spikes measured in multiples – and I cover them in my guide to scaling WordPress for high-traffic websites. But for a normal business site they add cost and complexity without adding anything a visitor would notice. If a salesperson is pitching you “enterprise cloud” for a 10,000-visits-a-month site, they are solving their commission, not your problem.

Specs that matter vs marketing fluff

PHP workers, explained plainly

A PHP worker is one queue position for uncached requests. Think of a bakery with two staff: two customers get served at once, everyone else waits in line. Two PHP workers means two uncached page loads, checkouts, form submissions or admin actions processed simultaneously – the rest queue, and visitors feel the queue as slowness or 502 errors. Cached pages skip the queue entirely, which is why worker counts matter little for a static brochure site and enormously for WooCommerce, where carts and checkouts can never be cached. Rough guidance from my projects: 2 workers suit a small cached site, 4-6 a busy business site, and a store doing real volume wants 8+ and an honest conversation with the host about limits. Many hosts hide their worker count – asking for it is itself a good support test.

The hardware and stack items worth checking

• NVMe storage. Database-heavy operations (search, WooCommerce filtering, wp-admin) are noticeably snappier on NVMe than on SATA SSDs. Most decent hosts have it now; a host still advertising “SSD” as a feature in 2026 is behind.
• RAM and CPU that are actually yours. On shared plans, ask what the limits are – the honest hosts publish CPU-seconds and process limits; the dishonest ones enforce them silently.
• HTTP/3 and modern TLS. HTTP/3 shaves real latency on mobile connections. It is a five-second check at http3check.net and a good proxy for whether the host keeps its stack current.
• Object cache availability. Redis or Memcached speeds up every uncached request – crucial for WooCommerce, membership sites and busy wp-admin. Some hosts include Redis free, some charge EUR 100 a month for it (genuinely), some do not offer it at all. Ask before you sign.
• PHP version policy. A good host offers PHP 8.2-8.4 today and lets you switch per site. A host still defaulting to PHP 7.4 is telling you how they treat security in general.
• Backup policy in detail. Daily automatic backups, retained 14-30 days, restorable by you with one click – not “backups available on request from support within 48 hours”.

The fluff to ignore

“Unlimited storage”, “unlimited bandwidth” and “unlimited websites” are marketing, not engineering – the limits moved into the fine print as CPU throttling, inode caps and fair-use clauses, where they are enforced exactly when your site gets busy, which is the worst possible moment. “Free domain for the first year” is a EUR 12 gift attached to a multi-year commitment. “99.99% uptime guarantee” usually pays out pennies of credit, not compensation. And a “free SSL certificate” has been table stakes since Let’s Encrypt made certificates free for everyone in 2016. Judge hosts on the specs above, not on the adjectives.

Benchmarks you can run before you buy

Test TTFB from where your visitors are

Every serious host has demo sites, trials or a money-back window – use it. Put a default WordPress install (or your actual site, migrated to a trial) on the plan you are considering and measure TTFB from your market’s location, not from wherever you happen to sit. KeyCDN’s free performance tool tests from 14 locations at once; WebPageTest lets you pick a city and a real device. If your customers are in Frankfurt, the number from Frankfurt is the only one that matters. What good looks like: under 400ms uncached, under 100ms for a cache hit. A simple repeatable check from any terminal:

curl -o /dev/null -s -w "TTFB: %{time_starttransfer}s\n" https://example.com/
# run it 5 times; the pattern matters more than one number

A five-minute load test with k6

TTFB with one visitor tells you nothing about TTFB with fifty. Grafana k6 is a free load-testing tool; this tiny script sends 20 concurrent visitors at a page for two minutes – enough to expose an oversold server without abusing anyone (run it against your own trial site only, and keep it modest):

import http from 'k6/http';
import { sleep } from 'k6';

export const options = { vus: 20, duration: '2m' };

export default function () {
  http.get('https://your-trial-site.com/');
  sleep(1);
}

Watch the p95 response time in the summary. A healthy host stays flat; an oversold one climbs from 300ms to 3 seconds as the workers saturate. I run exactly this before recommending any host to a client, and it has vetoed more “top-rated” hosts than I can count. Measure the same things again after you migrate – before/after numbers are the only honest proof the move was worth it, and the same metrics feed straight into Core Web Vitals, where TTFB is the first slice of LCP.

The support test: open a ticket before you pay

Before you give any host money, open a presale ticket or chat with a technical question – not “what does it cost” but something an engineer must answer. My standard three: “How many PHP workers does this plan include, and what happens when they are exhausted?”, “Is Redis object caching available, and at what cost?”, and “Can I switch PHP versions per site, and which versions do you offer today?”

What a good answer looks like

A good host answers in specific numbers within a few hours: “4 workers, requests beyond that queue for up to 60 seconds then return 502; Redis is included on this plan; PHP 8.1 through 8.4, switchable in the dashboard.” A bad host answers in an evasive paragraph about their “optimised infrastructure”, pastes a knowledge-base link that does not answer the question, or takes three days. Whatever they do with a presale question – when they are trying to win you – is the best treatment you will ever receive. Support quality degrades after the sale; it never improves.

Do the same test on response channels: is there live chat with technical staff, or only a ticket queue? Is phone support real or a sales line? For a revenue site, I weight support quality above a 50ms TTFB difference, because the worst hosting moments of your year will be conversations, not page loads.

Do not host your email with your web host

Almost every shared host bundles “free email”, and I advise every client to leave it switched off. Three reasons, all learned the hard way on inherited projects.

• Deliverability. Shared server IPs send everyone’s mail, including the spammers two accounts over. When the IP lands on a blocklist, your invoices and quotes go to spam folders, and you find out from an annoyed customer weeks later. Dedicated email providers guard their IP reputation obsessively; web hosts do not.
• Coupling. If your email lives with your web host, a hosting migration becomes an email migration, and a hosting outage becomes an email outage. The single worst inherited situation I see is a business locked into a terrible host because moving means touching ten mailboxes.
• Quality. Webmail on a cPanel server is a 2009 experience with 2009 spam filtering.

The fix costs little: Google Workspace or Microsoft 365 at roughly EUR 6 per user per month, or Zoho Mail from about EUR 1 – each with proper SPF, DKIM and DMARC support, which modern inbox providers effectively require anyway. Keep the website’s own outgoing mail (form notifications, WooCommerce receipts) off the web server too, via an SMTP or API service like Brevo, Postmark or SES. Decoupling email from hosting is an hour of DNS work that buys you permanent freedom to change hosts.

Moving between hosts without drama

The fear of migration keeps more businesses on bad hosting than any contract does, and it is mostly unfounded. A standard WordPress migration – files, database, DNS – is a few hours of careful work, and many managed hosts will do it free when you join them. The parts that actually need attention: keeping the old site live until DNS has fully propagated, lowering DNS TTL a day in advance, preserving every URL (or redirecting changed ones), re-pointing any hard-coded domain references, and testing forms, checkout and email sending on the new server before the switch. Do it properly and Google never notices – I have written up the full process in how to migrate a website without losing SEO rankings, and when clients want it done for them, it is exactly what my WordPress migration service covers: staged copy, checks, DNS switch, post-launch monitoring, zero downtime in almost every case. The practical takeaway for choosing a host: never stay somewhere bad because leaving feels risky. Leaving is a Tuesday afternoon.

Pricing reality by site type

Real monthly numbers from projects I run or maintain, in 2026 money:

• Brochure site (5-15 pages, under 5,000 visits/month): EUR 5-12 on quality shared hosting is honestly enough, provided caching is configured. Paying EUR 40 here buys convenience, not speed a visitor would notice.
• Business site (lead generation, bookings, 5,000-50,000 visits/month): EUR 20-45 on managed WordPress hosting or a well-run VPS. This is the tier where staging, daily backups and competent support start paying for themselves in avoided incidents.
• WooCommerce store: EUR 30-80 minimum, scaling with order volume – stores need PHP workers, Redis and headroom because checkout traffic cannot be cached. Under-hosting a store is the most expensive false economy in WordPress; a checkout that stalls during your busiest hour costs more than a year of proper hosting.
• High-traffic site (100,000+ visits/month, spiky campaigns): EUR 100-500+ on upper managed tiers, tuned VPS clusters or cloud setups, usually with a CDN doing heavy lifting.

When to upgrade

Upgrade when you see the symptoms, not the marketing: uncached TTFB creeping past 600ms at busy hours, 502/503 errors during traffic peaks, CPU-limit warnings from the host, wp-admin turning treacly, or checkout slowing when concurrency rises. Upgrade one step at a time and re-measure – most sites need one tier more, not four.

Red flags that predict a bad host

• The renewal price trap. EUR 2.99 for the first term, EUR 11.99 on renewal, and the real price is in grey text under the buy button. Judge every host on its renewal price – that is what you will actually pay for years.
• Forced multi-year terms. The headline price requiring 48 months upfront is a bet that you will not want to leave – made by people who know their own service.
• No staging environment. In 2026, updating plugins directly on a live business site is unnecessary risk. No staging on a EUR 25+ plan means the platform is old.
• Anonymous cPanel resellers. Hundreds of “hosts” rent a server, install WHM and resell slices with zero engineering behind them. If you cannot find who runs the infrastructure, where the data centres are and who answers when the server dies, walk away.
• Hidden inode and CPU-second limits doing the throttling that “unlimited” implies never happens.
• Backups you cannot restore yourself, exit fees, charges for a migration away, or refusal to give full file and database access – all signs a host retains customers by friction rather than quality.
• A conveyor-belt brand. When one corporation owns twenty hosting brands, support and infrastructure converge on the cheapest common denominator regardless of which logo you bought.

Server location, CDN and international audiences

Physics is part of the spec sheet. A request from Munich to a Frankfurt server makes the round trip in about 10-20ms; the same request to a US-east server takes 90-120ms, multiplied across every uncached resource. Rule one, always: put the origin server in or near the country where most of your customers are. A German business on a cheap US-only host is donating a tenth of a second to every competitor with a Frankfurt server.

A CDN changes the maths for static assets – images, CSS, JavaScript get served from an edge node near the visitor – and Cloudflare’s free tier does this well enough that I put it in front of most client sites. What a basic CDN does not fix is the uncached HTML itself: first visits, logged-in pages, carts and checkouts still travel to the origin. Options that push HTML to the edge (Cloudflare APO at USD 5 per month, or hosts with built-in edge caching) help genuinely international audiences. My practical rule: one main market – host there and add a free CDN; visitors spread across continents – host near the biggest market, cache HTML at the edge, and only consider multi-region hosting when the numbers prove the need. Host location choice matters more than any CDN toggle, and it costs nothing extra to get right.

The decision table: match your site to a hosting type

Collapse everything above into one honest lookup:

When hosting is not your problem

One honest caveat before you spend money on a migration: a bloated site is slow on excellent hosting too. If your pages carry 4MB of images, a page builder emitting thousands of DOM nodes and thirty render-blocking scripts, a better server improves TTFB and leaves the other four seconds untouched. I test this before recommending anything – a quick TTFB check tells me whether the server or the site is the bottleneck. Good TTFB and a slow site means the work is optimisation, not hosting, and that is a different job: my WordPress speed optimisation service exists precisely for the sites already sitting on decent servers. Fix the right layer first; sometimes that costs you nothing but an honest measurement.

And whichever host you choose, someone still has to run the site on it – updates, backups tested by actually restoring them, uptime and security monitoring. Managed hosting covers the server side of that; the WordPress side is what a maintenance and security plan covers, and the two together are what “my website just works” actually costs.

Need help choosing or moving hosts?

I do this for clients regularly: benchmark the current host, recommend the right tier for the site and budget with no affiliate interest in the answer, and handle the migration end to end – or fix the site itself when the measurement shows hosting was never the problem. Have a look at my portfolio, then send me a message with your current host and your traffic numbers – I reply within 24 hours with an honest recommendation and a fixed price.

Frequently asked questions

How do I choose WordPress hosting?

Match the hosting type to the site: quality shared hosting for brochure sites, managed WordPress hosting for business sites and stores, a VPS when you need control or run many sites. Then verify with a TTFB test from your market’s location and a presale support ticket before paying.

What is the best WordPress hosting?

There is no universal best – a host that is perfect for a Frankfurt brochure site is wrong for a US WooCommerce store. The best host for you has servers near your audience, honest specs (PHP workers, Redis, staging, backups) and support that answers technical questions in numbers.

Is shared hosting good enough for WordPress?

For a low-traffic brochure site with caching configured, yes – decent shared hosting at EUR 5-12 a month is genuinely fine. It stops being fine when the site earns revenue, runs WooCommerce or sees traffic spikes.

What are PHP workers and how many do I need?

PHP workers are the number of uncached requests a server processes simultaneously; everything beyond that queues. A small cached site is fine with 2, a busy business site wants 4-6, and a WooCommerce store with real order volume needs 8 or more because checkouts cannot be cached.

Should I host email with my web host?

No. Shared server IPs hurt deliverability, and coupling email to hosting makes every migration and outage twice as painful. Use Google Workspace, Microsoft 365 or Zoho, and send site emails through an SMTP service.

How much should WordPress hosting cost?

Realistically: EUR 5-12 for a brochure site, EUR 20-45 for a business site, EUR 30-80+ for a WooCommerce store, EUR 100+ for high-traffic sites. Always judge by the renewal price, not the first-term discount.

Will changing hosts hurt my SEO?

Not if the migration is done properly – same URLs, correct redirects, no extended downtime. Search engines do not care where a site is hosted, only that it stays reachable and fast; a faster host usually helps rankings slightly through better Core Web Vitals.