Skip to main content

Transparency

How scoring works at RackList

Every host on RackList carries a score between 0 and 100. This page lists every rule that feeds into it, the points each rule grants, and the safeguards that keep the score honest.

The principle

One number, five axes

Before the rule details, here is the structure of the score in three markers.

  • 0 - 100

    Score range

    Higher is better. Always clamped, never out of range.

  • 5axes

    Composition

    Five axes graded out of 20, adjusted with bonuses and malus.

  • 100%

    Transparency

    Every threshold shown is read from the real calculator code.

The engine

The formula

overall = seniority + reviews + compliance + responsiveness + value
        + sum(bonuses)
        - sum(maluses)
overall = clamp(overall, 0, 100)

Each axis is bounded to [0, 20], and the final score is clamped to [0, 100]. The clamp catches corrupted upstream data (a 6.0 average rating, a 150% response rate) before it can leak into the displayed number.

The breakdown

The five axes

Each axis is graded out of 20 and works on its own data. They are summed without weighting, because we want every axis to matter equally to the final score.

Seniority (0 - 20)

Three dates, ranked by how trustworthy they are. Tenure on RackList weighs the most: it is earned over time and cannot be back-dated. The legal entity's age comes next, clearly lower, because an old company is not necessarily an old host. The owner-declared project launch date is worth only a token point.

0 - 20
Rule Condition Points
On RackList for 5+ years On RackList for 5 years or more +14
On RackList for 3 - 5 years On RackList for 3 years or more +11
On RackList for 2 - 3 years On RackList for 2 years or more +8
On RackList for 1 - 2 years On RackList for 1 year or more +5
On RackList for under 1 year First approved publication less than a year ago +2
Legal entity 15+ years old Legal entity founded 15 years ago or more +5
Legal entity 10 - 14 years old Legal entity founded 10 years ago or more +4
Legal entity 5 - 9 years old Legal entity founded 5 years ago or more +3
Legal entity 2 - 4 years old Legal entity founded 2 years ago or more +2
Legal entity 1 - 2 years old Legal entity founded 1 year ago or more +1
Project launched 2+ years ago Project launched 2 years ago or more +1

Reviews (0 - 20, up to 60 for an unclaimed listing)

Linear projection of the average rating onto 0-20, after weighting by sample size: with few reviews the rating is pulled towards a neutral starting rating, and the real rating is only approached as reviews accumulate. Recent reviews count more than old ones. On an unclaimed listing the scale extends to 0-60, with a proportionally heavier review requirement, see below.

0 - 20
Rule Condition Points
Average projection score = (rating - 1) / 4 × 20 with the rating clamped to [1, 5] formula
Weighting by sample size A rating is a sample, and a small sample proves nothing. The rating used is therefore pulled towards a neutral starting rating of 3 out of 5, the more strongly the fewer reviews there are. The real rating is only reached gradually as reviews accumulate: no threshold to cross, and every further review counts slightly less than the one before it formula
Weight of the starting rating, claimed listing The starting rating weighs as much as 10 real reviews, so it takes 10 of them before it counts for no more than half +10
Weight of the starting rating, unclaimed listing The starting rating weighs as much as 30 real reviews. An unclaimed listing is given a ceiling three times higher on this axis while nobody has proven they are responsible for it, so that ceiling costs three times as many reviews to approach +30
The starting rating never earns points The weighting only ever pulls downwards. A listing with poor reviews keeps the rating its reviews give it and is never lifted back towards the starting rating formula
Time decay Each review is weighted by exp(-age_days / 365); 6-month-old review ≈ 0.5, 5-year-old review ≈ 0 weighted
Unclaimed listing rebalancing An unclaimed listing can never score on responsiveness or value (both require owner access): those 40 points carry over to the reviews axis instead, which extends from 0-20 to 0-60 +60

Compliance (0 - 20)

Eight checklist items on profile completeness and legal disclosure. Each item is small on purpose: no single field grants a runaway advantage.

0 - 20
Rule Condition Points
Claimed listing A verified host owner has claimed the listing +5
Legal identifier verified An operator confirmed the SIREN/SIRET against the registry +3
Legal identifier declared A SIREN or SIRET is set +2
Description (≥ 50 chars) Listing description has at least 50 characters +2
Founded date founded_at is set +2
At least one category Host is linked to at least one category +2
Logo A logo file is attached +1
Website A website URL is set +1
Legal form The legal form field is filled (SAS, SARL, Ltd, etc.) +1
Language A primary language is set on the listing +1

Responsiveness (0 - 20)

Rate of approved reviews that received a host response, weighted by reply speed. Stale replies count for less, neglecting reviews scores zero.

0 - 20
Rule Condition Points
Formula score = clamp(rate, 0..100) / 100 × 20 × speed_factor formula
Fast factor (×1.0) Median first-reply time ≤ 24 hours +1
Medium factor (×0.8) Median first-reply time between 24 hours and 72 hours +0.8
Slow factor (×0.6) Median first-reply time between 72 hours and 7 days +0.6
Very slow factor (×0.4) Median first-reply time greater than 7 days +0.4
Unknown timing fallback (×0.8) Rate present but no timing data - never grants the full multiplier +0.8

Value (0 - 20)

Catalogue depth (product count), price transparency (share of products with a public price) and category diversity. Each capped so stuffing cannot run away.

0 - 20
Rule Condition Points
Catalogue tier 10+ Host lists 10 or more active products +14
Catalogue tier 7-9 Host lists between 7 and 9 active products +12
Catalogue tier 4-6 Host lists between 4 and 6 active products +8
Catalogue tier 1-3 Host lists between 1 and 3 active products +4
Price transparency Up to +4 proportional to the share of active products with a public price +4
Category diversity (3+) Active products span at least 3 distinct categories +2
Category diversity (2) Active products span exactly 2 distinct categories +1

Adjustments

Bonuses and penalties

Once the five axes are summed, the score is nudged up or down based on targeted signals.

Bonuses (+)

Points added when the host goes beyond the baseline.

Rule Condition Points
Verified legal identifier Operator confirmed the host's SIREN/SIRET against the official registry +3
Claimed by owner The listing has been claimed by its verified owner +2
Complete profile Compliance axis maxed at 20/20 +2
Rich catalogue 10 or more active products AND at least one approved review +2
Verified ASN Operator confirmed the host's declared ASN(s) against PeeringDB/RIPEstat +2

Maluses (-)

Points removed when the host falls short on signals readers rely on.

Rule Condition Points
No products listed No active product is attached to the host -5
No legal identifier disclosed Neither SIREN nor SIRET is set - the comparator cannot tell who legally stands behind the listing -4
Severely incomplete profile Compliance axis below 5/20 -3
Aggressive pricing More than half of the host priced products sit below 30% of their category median -2
No description The description field is empty -2

Anti-gaming

Anti-manipulation safeguards

The rules above are useless if a single dishonest host can game them, or if a host with no idea its listing exists ends up penalised with no way out. The score carries seven guard rails baked into the calculator code.

  1. Sample-size confidence cap

    A single five-star review cannot top the reviews axis. The axis is capped at 8/20 below 3 reviews and at 16/20 below 10. A host needs at least 10 independent approved reviews to reach the maximum.

  2. Time-weighted reviews

    Each approved review is weighted by exp(-age_days / 365). A host coasting on five-year-old praise cannot mask a recent drop in quality.

  3. Reply-speed multiplier

    A 100% response rate alone does not give 20/20. The rate is multiplied by a speed factor that decays from ×1.0 (under 24h) to ×0.4 (over 7 days), so a quarterly cron of "thanks!" replies scores 8/20, not 20.

  4. Mandatory re-moderation

    Whenever a host edits its SIREN, SIRET or declared ASNs, its listing automatically returns to pending verification and its verified marks are removed. A team member reviews it again before any new publication.

  5. Catalogue stuffing block

    The rich-catalog bonus only fires when the host has both 10+ products AND at least one approved review. Listing throwaway products without user validation no longer grants any point.

  6. Aggressive pricing detection

    When a majority of a host's priced products fall below 30% of their category median, the aggressive-pricing malus fires. A single entry-tier offer never triggers it. We look for a recurring pattern, not a one-off deal.

  7. Equal footing for unclaimed listings

    A host with no idea its listing exists, or one that chooses not to claim it, should not be structurally disadvantaged on the platform. Since it can never score a single point on responsiveness or value (both require owner access), those 40 points carry over to the reviews axis instead. An unclaimed listing is judged on the one signal it can actually produce: real user reviews, not a number capped by default.

Boundaries

What the algorithm never does

A score is also defined by what it refuses to consider. The rules below are out of scope on purpose, and they will stay out unless the policy changes publicly.

  • No money buys score. There is no "boost my listing" feature and no plan modifies the algorithm. Paid features (when they exist) buy visibility clearly labelled as sponsored, never points.
  • No black-box model. The whole algorithm fits in five PHP files. The page you are reading is generated from them. Anyone can audit the formula.
  • No silent change. Every threshold update lands in a commit on the public repository, and this page picks it up on the next request. We cannot quietly retune the score.
  • No discrimination by entity type. We label whether a host is a company, a sole trader or a non-profit for readers to factor in, but we do not give points based on which one.

Track hosting changes

Weekly price and score changes in your categories.

Your interests

At least one category.

I agree to receive RackList's weekly digest for my chosen categories (price changes and score movements). I can unsubscribe at any time via the link in every digest email.

Double opt-in confirms your choice by email. See our privacy policy (GDPR art. 6.1.a - consent).