Methodology
A score you can recompute.
Every hosting site prints a badge — 9.6 Excellent, 9.3 Very Good — and not one of them will tell you what makes it 9.6 rather than 9.4. It is a vibe with a decimal point, and its real job is to make a commission feel like a measurement. Ours is a function. Here it is.
The four inputs
Four-year cost
40%What actually leaves your account over four years, promotional term and renewal included. Scored on a log scale so a premium host is ranked, not annihilated.
Freedom to leave
20%How many months you must pay for up front to get the advertised price. Cancel-any-month scores 10; a four-year prepayment scores 0.
Renewal honesty
20%How far the price jumps when the promotional term ends. A host whose price never changes scores 10. A host that quadruples scores near zero.
What's included
20%How much of the job is done for you: management, backups, staging, SSL, migration, 24/7 support, CDN, a domain. Eight things, one point each.
score = 0.4 × cost + 0.2 × freedom + 0.2 × renewal + 0.2 × included
What each input actually measures
Four-year cost (40%)
We total what leaves your account over 48 months, promotional term and renewal included, from the vendor's own published prices. Four years because it is long enough that every promotional rate on the market has expired — comparing month-one prices is precisely how buyers get caught.
It is scored on a log scale, not a linear one. Linear normalisation would hand the most expensive host a flat zero and squash every meaningful difference beneath it; on a log scale the gap between $182 and $528 stays legible alongside the gap between $528 and $1,680.
Freedom to leave (20%)
How many months you must pay for in advance to get the advertised rate. Cancel-any-month scores 10. Hostinger's 48-month prepayment scores 0 — not because prepaying is immoral, but because handing over four years of money before you know whether the product suits you is a real cost, and a score that ignores it is measuring the advert rather than the purchase.
Renewal honesty (20%)
How far the price jumps when the promotional term ends, as a ratio. A host whose price never changes scores 10. SiteGround, which renews at 501% of its promotional rate, scores 1.7. This is the number the industry works hardest not to print, so we weight it as heavily as the price itself.
What's included (20%)
Eight jobs somebody has to do. One point for each one they do for you:
- Server managed for you
- Backups included (not an add-on)
- Staging environment
- Free SSL
- Free migration
- 24/7 support
- CDN included
- Free domain
A worked example — check us
Cloudways is currently top of the ranking. Here is exactly how that number is produced:
Cloudways
$528 over four years
| Input | Points | Weight | Contribution |
|---|---|---|---|
| Four-year cost | 5.7 / 10 | × 0.4 | 2.27 |
| Freedom to leave | 10.0 / 10 | × 0.2 | 2.00 |
| Renewal honesty | 10.0 / 10 | × 0.2 | 2.00 |
| What's included | 6.3 / 10 | × 0.2 | 1.25 |
| Total | 7.5 | ||
Bands
The word beside the number is fixed, not chosen: 8.0+ Excellent, 7.0+ Very good, 6.0+ Good, 5.0+ Fair, below 5.0 Weak. No host gets a kinder adjective for paying us.
What this score is not
It measures value — what it costs to keep a site online for four years, and what you have to give up to get that price. It does not measure how good a product is in the abstract. Kinsta scores in the middle and is a genuinely excellent piece of engineering; it is simply not what most people should buy with their own money. If we called this an overall quality score we would be lying, so we don't.
It also cannot capture things that are not in the price list: how good the support actually is on a bad day, whether the control panel makes you want to throw your laptop. Where that matters, we say so in words, on the head-to-head pages, rather than pretending we quantified it.
Why you can trust the ranking
Not because we promise to be good. Because the formula is public and the inputs are the vendors' own published prices, so rigging a ranking would require visibly changing the formula — in public, in this page's git history.
The evidence is already on the homepage.DigitalOcean pays us nothing, ranks fifth, and sits above 5 hosts that pay us well.And Hostinger, the cheapest host on the site, ranks seventh — because cheapness bought with a four-year lock-in and a 348% renewal is not the same thing as value. A formula built to serve commissions would not produce either of those results.
If you think a weight is wrong — that cost should be 50%, that lock-in matters less than we say — that is a completely legitimate argument, and it is one you can only have with us because we published the weights.Tell us.