How we rank

Our ranking systems

Each system below is a category in the Top100 Venue Score. Together they answer how strong a venue is as a wedding venue on evidence we can verify consistently. They do not answer whether the venue fits one couple. That later question is a Match Score.

How a score is produced

1. 01

   Collect evidence

   Official pages, registers, maps, approved reviews

2. 02

   Accept facts

   Unknown stays unknown. No guesses from venue type

3. 03

   Run systems

   Thirteen category systems, each on its own facts

4. 04

   Check scoreability

   Enough of the model must be known

5. 05

   Score or wait

   A number, or insufficient evidence

6. 06

   Freeze an edition

   Official lists are historical records

The mathematics

The Top100 Venue Score is a deterministic calculation. The same accepted facts and the same scoring version always produce the same result. There is no editor boost and no commercial multiplier.

Category percentages are versioned in configuration and are not listed here. The structure below is what those percentages feed.

1. Layer 1

   Raw fact

   A stored value with a data state: known present, known absent, unknown, or not applicable.

2. Layer 2

   Metric score

   Each known applicable fact is mapped onto 0 to 100 with a named normaliser.

3. Layer 3

   Category score

   A weighted mean of known applicable metrics in that system, only if the category is scoreable.

4. Layer 4

   Overall score

   A weighted mean of scoreable categories, only if the whole venue is scoreable. Otherwise no number.

Normalisers

true maps to 100. false maps to 0. This is used only where present is unambiguously better for that metric.

• Known present: 100
• Known absent: 0
• Unknown: omitted, not zero

Counts, distances, capacities, ratios, times, and review volume use piecewise linear curves. Between two published anchors (x1, s1) and (x2, s2), the score is s1 plus the fraction of the way from x1 to x2, times the difference in scores. Values outside the end anchors are clamped. Several curves plateau: extra size or extra urban convenience stops adding points once the venue is adequate.

Review shrinkage

Public star averages mislead when sample sizes differ. We first shrink each venue's mean rating toward a market prior, then convert that adjusted rating to 0 to 100, then combine it with a separate volume score that has diminishing returns.

AdjustedRating = (v / (v + m)) * R + (m / (v + m)) * C

If the approved source has almost no rating spread (for example nearly every venue displays the scale maximum, or C equals the scale maximum), we do not score reputation at all. Volume is evidence depth, not a contest. Extra reviews after a few hundred add almost nothing. Reputation is mostly adjusted rating quality, with a smaller share for volume.

Category combination

CategoryScore = sum(metric_score * metric_weight) / sum(metric_weight for known, applicable metrics)

• Unknown metrics are omitted. They are not scored as zero.
• Not applicable metrics are omitted from both the score and the confidence denominator.
• If verified applicable weight is below 50%, the category has no numeric score.
• From 50% to just under 70%, research may see a provisional figure. It does not enter the overall Top100 Venue Score.
• At 70% or more verified applicable weight, the category is scoreable.

Toy example. Three metrics in one category, internal weights 40, 40, and 20. Only the first is known, and it scores 100. Without a coverage gate the category would also be 100. That is why TVS-1.2.0 requires 70% verified weight before a category number is used.

Overall combination

Top100Score = sum(category_score * category_weight) / sum(category_weight for scoreable categories)

• An overall score exists only if scoreable categories cover at least 75% of total category weight.
• Ceremony, reception, guest facilities, and accessibility must all be scoreable.
• Reputation must come from an approved source with real rating spread.
• Official ranking also requires at least 85 overall confidence.
• If those gates fail, the public result is insufficient evidence. Not a 99 with a footnote.

Internal ranking order uses unrounded scores. The public figure is one decimal place. Neighbouring ranks inside a venue-specific stability half-width are not treated as a meaningful gap. That half-width is at least 0.25 points, and it increases when scoreable coverage or confidence is lower: 0.01 times (100 minus scoreable category weight) plus 0.02 times (100 minus capability confidence).

When we will not show a score

Unknown is not a no. It is also not a free pass to look perfect on whatever happens to be published. If too little of the model is verified, we will not display an overall number.

What the 0 to 100 scale is meant to mean

Official order uses full-precision scores. If two venues sit inside a published stability range, we will not imply that one place is meaningfully better. Thin evidence widens that range.

What venues can do to rank higher

The useful work is to make true capabilities visible and verifiable. The useless work is to negotiate the methodology or purchase attention.

What does not help

• Paying Top100, advertising with us, or asking for a boost
• Buying reviews, incentivising five-star copy, or flooding a directory
• Renaming the same room three ways so it counts as three spaces
• Claiming every guest is welcome instead of stating access facts
• Asking us to treat a missing fact as yes because similar venues have it
• Adding 200 extra bedrooms, or a 500-person room, as if bigger were automatically better
• Objecting to a weight. That is methodology feedback, not a correction

Methodology updates

When ranking systems change, we publish a new version identifier. Official editions stay frozen on the version used at publication. We keep earlier versions so a later method can be compared on the same accepted facts, not so a famous venue can be moved by taste.

1. August 2026TVS-1.2.0

   Scoreability and honest numbers

   No overall score until enough of the model is verified. Reputation is not scored from a source where almost every venue looks the same. Practical guest capacity is treated as adequacy, not a size contest. Neighbouring ranks use a stability range that widens when evidence is thin.

2. August 2026TVS-1.1.0

   Curves and duplication

   Stopped treating larger hotels as automatically better on bedrooms and capacity. Flattened transport after reasonably accessible. Reduced raw building age versus official heritage. Required facts instead of venue-type guesses. Required transparency on the venue's own site.

3. August 2026TVS-1.0.0

   First rehearsal architecture

   Thirteen category systems and 95 metrics. Unknown is not zero. Confidence is separate from score. First run exposed review-source and size distortions. Kept as a snapshot so later versions can be compared on the same facts.

Before we freeze a public edition we still need an approved review source with real rating spread, a deep 20 to 30 venue fact set across all metrics, and a 300 to 500 venue England universe that is not defined by one directory. We compare versions on the same facts. We do not move a famous name by hand.

Corrections

If a published fact is wrong, send the venue name, the fact, the correction, and a source we can verify to hello@top100weddingvenues.com with the subject Correction. There is no paid appeal.