Renewal transparency
Is the renewal price published on the pricing page? This is the site's core thesis.
How: renewal price published → 25 · not published → 0 + warning on the card
HomeOn this page
Methodology — how we score
The number in a plan's ring comes from five measured inputs; each input is verifiable at its source, and the weighting is the editorial judgement we publish openly. Every weight and threshold on this page is read straight from the scoring engine — if the code changes, the page changes.
Today: preliminary scoreFrom verified plan data only. No speed or uptime.
Measurement network: day 17 / 30From 4 locations every 5 minutes. Nothing enters the score before 30 days.
Full score: once measuredThe same ring becomes "score"; order may change, the rules don't.
In short: Scores are 0–100. The ring is green at 70+, amber at 45+, red below. Coupons, commission and paid plans can never touch the score. Unknowns aren't punished; they get a warning on the card.
Today's preliminary score — five components, 100 points
The ring on cards is a preliminary score: computed from verified plan data, no measurements. The five component weights total 100. The ring in the plan page header is different: that one is the weighted engine score (see “Full score” below). The two are named separately.
2525201515
Renewal transparencyRenewal termsPrice positionSpec completenessTerm flexibility
Renewal terms
How close the renewal multiplier is to 1.0 — is the intro price a trap?
How: ≤ ×1.1 → 25 · ≤ ×1.5 → 20 · ≤ ×2 → 16 · ≤ ×3 → 10 · ≤ ×4 → 6 · above → 2 · unknown → 8
Price position
Where the renewal sits against the category median (robust to outliers).
How: toward free → 20 · at the median → 12 · at double → 4 · above → 2 (continuous curve) · no context → neutral 10
Spec completeness
How many of six key spec fields (sites, storage, panel, backups, SSL…) are known.
How: 6 fields, 2.5 points each · NO spec data at all → neutral 8 — missing data is not a penalty
Term flexibility
Monthly billing availability, years paid upfront, absence of promo games.
How: monthly billing → 7 · shortest term ≤ 1 year → 4 · first price at least a fifth of renewal → 4
How a plan's score comes together
The same calculation runs on every surface: list, plan page, provider profile, comparison and the decision engine read one source. A real example:
Renewal transparency25/25
Renewal terms2/25
Price position13/20
Spec completeness10/15
Term flexibility4/15
25 + 2 + 13 + 10 + 4 = 54 — the "How the score was built" link on its card shows these exact bars; if the plan changes, this example changes too.
Personal score in the decision engine
Your five answers never change the data score; they add a 0–20 fit on top. The same plan can rank differently for two people while its "data score" stays identical everywhere.
Personal score=0.8 × data score+fit (0–20)
Region /5Support language /5Panel /5Priority /5
Fit contains no measurements; items without data get 0, never invented. Example: data 54, assume fit 12 → personal 0.8 × 54 + 12 = 55.
What never enters the score
Speed and uptime — being collected; nothing publishes before the 30-day window fills. No fixed middle value.
Datacenter location — proximity is good or bad depending on preference; it can't enter a neutral score — it stays a filter.
Coupons and commission — a coupon is the card's second line, never the order; affiliate data never reaches the engine.
Full score — once measurements land
When the window fills, the same ring switches to the five-part full score. Weights follow your chosen priority; the four canonical profiles:
Balanced default
38%22%18%12%10%
Lowest cost
34%32%14%12%8%
Most reliable
34%16%20%16%14%
Fastest
38%14%26%12%10%
Fit for youTrue costMeasured performanceTrustFeatures
Performance measurement is not part of the score yet: the independent measurement network is not in place. That component is dropped and its weight is redistributed across the remaining ones with their proportions intact — no placeholder value is substituted. The component returns on its own once measurement begins. Since 2026-09-18 the same rule applies to the feature component: if we could read fewer than half of a rubric’s features for a plan, that component is not produced either and its weight is redistributed. This is currently the case for shared, WordPress, reseller and domain types; the covered weight is recorded.
True cost is not the advertised price
The cost component measures what you'll pay over your chosen horizon, not the intro price. The "~3-year total" line on plan pages is this sum.
Intro term+Renewal terms+Mandatory setup+Mandatory add-on+Estimated tax=3-year true costCoupon excluded
Coupons are never subtracted from the total; they're shown separately. No company can rank higher by adding a coupon. If the shortest term exceeds 3 years, the total is that upfront amount — you can't buy less.
Measured performance
Performance is measured, not claimed. Sub-components and weights:
An unmeasured component never gets a fixed number; the pool median is used and the plan is marked "collecting measurements". A fixed middle value would rank measured-but-poor plans below never-measured ones.
The measurement source changes trust, not the score
The same uptime value earns the same points whatever the channel. What the source changes is data confidence and badge rights:
SourceConfidenceBadge
Anonymously purchased accountfullIndependently measured
Provider-sponsored account-5Independently measured
Provider-hosted agent-12no badge
External status claim-25no badge
Where this measurement stops: if our own account sits behind a CDN, the request is answered at the edge and never reaches the server — it would read "up" even with the server completely down. So we point the check at the hostname the provider itself publishes for that server. What you get is therefore the reachability of that server, not of our own account's vhost. You can probe the same hostname and audit our figure: BurtiNET → www.hesapis.com, Namecheap → business145.web-hosting.com
We only publish offers from sources whose prices we can verify automatically: if a firm has no pricing API, its offers do not appear in our lists. The reason is not that such prices are wrong, but that their staleness is invisible — an API price refreshes every night and raises an alarm when it goes quiet, while a hand-entered price ages for months unnoticed. The rule empties some categories; we state it rather than hide it.
Providers can buy a subscription from us; it does not affect score or ranking. How to verify: Premium and independence.
Trust score
72 points of trust come from fixed components; the remaining 28 are shared between user reviews and editorial assessment by review count. No thresholds, no jumps.
0 reviewseditorial extra 28 · users 0
8 reviewseditorial extra 14 · users 14
40 reviewseditorial extra 5 · users 23
A 9/10 from three reviews doesn't weigh like a 9/10 from three hundred: the user share is 28 × n / (n + 8).
Penalty items — capped
Risks lower the score but can't take it over: each item has its own cap, and the total can't exceed 30.
−30total penalty cap. Unknowns aren't punished: if the renewal price is unverified the penalty is zero and a high-visibility warning appears on the card.
What a plan is compared against
Same product typeShared and dedicated are never normalized in one pool.
Same price tier and regionMedian and position are computed within the tier.
Fewer than 5 plans → no rankingSaying "not enough data" beats producing a weak ranking.
We never fill blanks with guesses for an unmeasured plan. Where data is missing it says "collecting measurements".
Changelog
Every change that touches scores or rankings sits here with its date.
2026-09-18 — Feature component: a feature we could not read is no longer counted as “absent in this plan”. In addition, if fewer than half of a rubric’s features are known, the component is not produced at all and its weight is redistributed — so thin data cannot produce a high score.score + order
2026-09-18 — Availability calculation fixed: probes a provider blocks were already outside the denominator, but a day blocked in full was written as 0%. A day that could not be measured no longer yields a percentage; 237 past days were corrected. Provider-profile averages are now weighted as well.score changed
2026-09-12 — The ring in the plan page header moved from the data score to the engine score: listing cards show the pre-score, the plan page shows the weighted engine score. The two are named separately.score changed
2026-09-09 (late) — Fit component added to the decision engine: personal score = 0.8 × data score + a 0–20 fit derived from answers; different region/priority answers now produce different rankings.score + order
2026-09-09 — Decision-engine results are ordered by pre-score; one plan per provider is shown, and reason sentences are built from provider profile data (datacenter, panel, support language).order changed
2026-09-07 — The price-position component moved from stepped thresholds to a continuous curve: a small price difference now yields a small score difference, no bracket jumps.score changed
2026-09-05 — The pre-score was bound to a single source: list, plan page, provider profile, comparison and the decision engine all read the same calculation; the same plan gets the same score on every surface.score + order
Frequently asked
Can a company pay to raise its score?
No. Paid provider plans only buy data and tools: own measurements, correction requests, a badge. Weights, penalties and data confidence are plan-independent.
Why does the renewal price weigh so much?
Because most of what you pay goes to renewals: on average two-thirds of a 3-year total is renewal terms. Any ranking that spotlights the intro price misleads you in year two.
Will my score change when measurements start?
It can. The preliminary score uses five data components; the full score adds measurements. At the switch, cards relabel to "score" and a dated entry lands in the changelog. The rule stays: what isn't measured isn't guessed.
Prefer numbers over rules?
On any plan card, the "How the score was built" link shows these five bars for that plan.