Best Ecommerce Hosting: Size for Black Friday, Not Tuesday

Ecommerce hosting doesn’t fail on a Tuesday afternoon. It fails during the handful of hours a year when your traffic is ten or fifty times normal — and those are the hours that carry the revenue.

Which means most of what you’ll read about store hosting is measuring the wrong thing.

The technical core: checkout can’t be cached

This is the fact everything else follows from, and it’s architectural rather than a benchmark result.

Page caching serves a stored copy of a page to many visitors. That works beautifully for product listings, category pages and content — and it’s what every host’s speed test measures.

Cart and checkout pages cannot work that way. They’re per-session and stateful: this visitor’s basket, this visitor’s shipping options, this visitor’s payment step. Serving a cached copy would show one customer another’s cart. So every WordPress host bypasses page caching on cart and checkout by design.

The consequences are the whole article:

Every cached-TTFB benchmark you’ve seen is irrelevant to checkout. A host can post extraordinary numbers on a cached homepage and still collapse at 200 concurrent checkouts, because those requests go all the way through PHP to the database, every time.

Under load, three things decide whether checkout survives:

• PHP workers. Each concurrent uncacheable request occupies one. Run out and requests queue, then time out. This is the single most important spec for a store, and it’s the one hosts publish least clearly.
• Database performance. Every checkout is multiple writes — cart, order, inventory, session. Shared database resources fail here first.
• Object caching (Redis). Since you can’t page-cache checkout, object caching is what reduces the database work per request. For a store this isn’t an optimisation, it’s the mechanism.

So the question to ask a prospective host isn’t “how fast is your CDN.” It’s: how many PHP workers, is Redis available, and what happens when workers are exhausted — queue or error?

When the expensive plan is the cheap one

The instinct is to compare $25 against $115 and take the $25. Here’s the arithmetic that reverses it.

Upgrading from a $25 plan to a $115 plan costs $1,080 a year.

An hour of downtime costs you an hour of revenue — but not an average hour. Peak hours are what fail, and peak hours are worth multiples of average ones.

At $30,000/month GMV, an average hour is $41.10 of revenue:

So the “one hour of downtime pays for a year of hosting” framing is true, but only above a threshold. At $30k/month GMV you need a genuinely extreme peak hour. Run it the other way and the threshold is clear:

Above roughly $40,000/month GMV, a single Black Friday hour of downtime costs more than a year of the plan upgrade. At $100,000/month it’s $2,740 for that hour against $1,080 for the year — no longer a close decision. At $300,000/month it’s $8,219.

Below about $40k/month, the cheap plan is genuinely the rational choice, and anyone telling you otherwise is selling. Size up when your peak-hour revenue crosses that line, not before.

The hosts

Ordered as we rank them for stores specifically.

Nexcess is built around WooCommerce and Magento rather than adapted to them, with store-specific caching that understands which paths must bypass the page cache. If you want a host whose product decisions were made with carts in mind, start here.

Cloudways is the value pick and the most controllable — you rent a defined server, so PHP workers and memory are yours rather than an allowance, and Redis and Varnish are available on the platform. $11/month for 2GB is remarkable for a small store, and you can scale the server for a sale week and scale back after, which is exactly the right shape for this workload.

Kinsta is the premium managed option, from $35 for a single site, with a $0.50 per 1,000 visit overage — the gentlest in managed WordPress — and bandwidth-based plans available since November 2025 if bot traffic inflates your counts.

SiteGround is the common shared-hosting starting point, and its GoGeek tier adds staging and more resources. Know the renewal: $7.99 becomes $44.99 after twelve months, the steepest step we’ve measured.

Liquid Web, which owns Nexcess, is the fully-managed end — the answer when an outage is expensive enough that you want someone contractually accountable rather than a status page.

WP Engine is strong on agency and staging workflow, with the caveat that its overage is $2.00 per 1,000 visits — four times Kinsta’s — which is exactly the wrong shape for a store that spikes.

Rocket.net doesn’t meter visits at all, billing disk and bandwidth instead, from $30/month. For a store facing an unpredictable sale spike, not being metered on traffic is worth more than a slightly lower rate.

Hostinger is the right answer for a small store rather than a busy one — $2.99/mo on a 48-month term renewing at $10.99, with SSL, CDN and backups bundled. Below the GMV threshold above, this is where the arithmetic actually points.

What to do before your next sale

Practical, and worth more than switching host:

1. Find out your PHP worker count. Ask support directly. If they won’t say, that tells you something.
2. Enable object caching (Redis). On a store this is the highest-value change available, because the checkout path can’t be page-cached.
3. Load-test your own store, not a benchmark. Tools that simulate concurrent checkouts exist and a test against your actual theme and plugins is worth more than any comparison table, including this one.
4. Scale up for the sale week. On Cloudways or any VPS this is a slider. Doing it for one week a year costs almost nothing and removes the entire risk.
5. Check what happens at the limit. A host that queues requests degrades; a host that errors loses orders. Ask which yours does.

How we checked this

The load test this article’s brief called for is not here. It asked for a real test of a stock WooCommerce install at 50, 200 and 500 concurrent checkouts across three managed hosts and a tuned VPS, with published checkout error rates. We did not run it — we don’t hold accounts across these providers and won’t publish performance figures we didn’t measure. That’s the central limitation of this piece, and it’s why the article is built on architecture and arithmetic instead.

What is solid: cart and checkout pages being excluded from page caching is a design property of every WordPress caching implementation, not a measurement — they’re per-session and stateful, so a cached copy would serve one customer another’s basket. Everything in the first section follows from that, including why cached-TTFB benchmarks don’t describe checkout behaviour.

The revenue arithmetic is ours: $30,000 monthly GMV divided across 730 hours, multiplied by peak concentration factors, against a $90/month plan difference over twelve months. The peak multiples are illustrative — 1×, 5×, 20× and 50× — because real peak concentration varies enormously by category and promotion. The $40,000/month threshold is where a 20× peak hour crosses the $1,080 annual upgrade cost, and it moves if your peak is peakier or your upgrade cheaper.

Host pricing is each provider’s published figures read in August 2026 across this series: Cloudways at $11 for 2GB, Kinsta from $35 with a $0.50 per 1,000 visit overage, SiteGround’s GoGeek at $7.99 renewing at $44.99, WP Engine’s $2.00 per 1,000 overage, Rocket.net from $30 unmetered, and Hostinger at $2.99 renewing at $10.99. We have not verified Nexcess’s or Liquid Web’s plan pricing for this article, so both are described rather than priced.

What we did not do: no load tests, no PHP worker counts (hosts rarely publish them, which is itself the finding), and no evaluation of any host’s store-specific caching in practice. The recommendation ordering reflects product positioning and the pricing structures we did verify, not measured checkout performance.

The host links above are affiliate links. The article’s conclusion — that below roughly $40k/month GMV the cheap plan is the rational choice — points at the cheapest options we list.

FAQ

What’s the best hosting for an ecommerce site?

Nexcess or Liquid Web if outages are expensive, Cloudways for the best control-to-price ratio, Rocket.net if you spike unpredictably and don’t want traffic metered. Hostinger below about $40k/month GMV, where the cheap plan is genuinely correct.

Why don’t page speed benchmarks predict checkout performance?

Because cart and checkout can’t be page-cached — they’re per-session, so every request runs through PHP and the database. Benchmarks measure cached pages, which is the opposite situation.

What spec matters most for a WooCommerce store?

PHP worker count, then database performance, then object caching. Workers are what run out under concurrent checkouts, and most hosts don’t publish the number — ask.

Should I pay for expensive ecommerce hosting?

Above roughly $40,000/month GMV, yes — a single peak hour of downtime then costs more than a year of the upgrade. Below it, the cheap plan is the rational choice.

Does Redis actually matter for a store?

More than for a content site. Since checkout can’t be page-cached, object caching is the main mechanism for reducing per-request database work. On a store it isn’t an optimisation, it’s the plan.

How do I prepare for Black Friday?

Find your PHP worker count, enable object caching, load-test your own store rather than trusting benchmarks, and scale the server up for the sale week. Scaling for one week a year costs very little.

Is unmetered traffic worth paying for?

For a store with unpredictable spikes, often yes. A host charging $2.00 per 1,000 visits over your allowance is the wrong shape for a business whose good days are its busiest days.