How SWEA Rankings Work
SWEA rankings aren't selected by editors or purchased by restaurants. Every position is calculated from qualified diner reviews, weighted for quality, recency, confidence, and authenticity.
The core principle: moderate legitimacy, not outcomes.
Our job as administrators is to ensure reviews are real. The algorithm decides the rankings. No restaurant can buy a position. No editor can override the math.
The Scoring Formula
Every review produces an individual score based on five dimensions. The weighted combination becomes that review's contribution to the restaurant's SWEA Score.
Individual Review Score
Example: A diner rates Food 96, Service 90, Atmosphere 94, Value 86, and selects Eat Here (100).
96×0.40 + 90×0.20 + 94×0.15 + 86×0.15 + 100×0.10
= 38.4 + 18.0 + 14.1 + 12.9 + 10.0
= 93.4 individual score
But that review does not simply move the restaurant to 93.4. The system then applies safeguards for sample size, recency, reviewer credibility, and fraud detection.
Bayesian Confidence Adjustment
A restaurant with 3 reviews averaging 99 should not beat a restaurant with 2,800 reviews averaging 95. That's the most important protection in the system.
Every restaurant starts with a neutral baseline — the DFW-wide average of approximately 72. As legitimate reviews accumulate, the baseline's influence disappears.
Adjusted Score =
(Restaurant Reviews × Restaurant Average
+ Baseline Weight × DFW Average)
÷ (Restaurant Reviews + Baseline Weight)
Restaurant A
3 reviews · 99 average
88.7
Provisional SWEA
Restaurant B
2,800 reviews · 95 average
94.9
Full SWEA
As Restaurant A earns more legitimate reviews, the baseline influence disappears and its true score emerges. This makes the ranking statistically much harder to game.
Ranking Tiers
Not every restaurant with reviews enters the leaderboard. Minimum sample sizes prevent a restaurant from rallying five friends and becoming #1.
The 100-review requirement for Top 25 eligibility is the strongest anti-manipulation protection. It is practically impossible to fake 100 legitimate, varied reviews.
Confidence Labels
Two restaurants can have similar SWEA Scores but very different levels of statistical confidence. We show this publicly so you can make informed decisions.
94.72
Based on 3,842 qualified reviews
95.31
Based on 21 reviews
95.31 does not necessarily mean it is definitively better than 94.72. The confidence label tells you how much to trust the number.
Recency Weighting
Restaurants change. Chef changes. GM changes. Service improves. Kitchen falls apart. A review from last night should matter more than one from four years ago.
Review Age → Weight
This gives you the live-market effect without allowing one bad dinner to destroy a restaurant. Older reviews still contribute historical signal — they just carry less weight.
Reviewer Trust Multiplier
Not every account should have identical influence. A brand-new account posting one suspicious review should carry less confidence than a user with 60 legitimate restaurant ratings.
Helpful vote bonus: reviewers who consistently receive helpful votes earn up to +0.10 additional multiplier. This rewards diners who write useful, detailed reviews.
Wealth, follower count, and paid membership never increase scoring influence. The system rewards legitimate participation only.
Controlled Volatility
One review should not send a restaurant from #7 to #29 — unless it barely has enough reviews to rank. The Bayesian weighting naturally controls this.
Restaurant with 5,000 reviews
A normal new review moves the score by approximately
0.01–0.08 SWEA
Restaurant with 40 reviews
A normal new review moves the score by approximately
0.2–1.0 SWEA
This feels market-like and is mathematically defensible. Large-sample restaurants move gradually. Small-sample restaurants move more, but remain marked Provisional.
Fraud Detection & Review Bombing
The algorithm detects and quarantines suspicious activity before it affects rankings. Suspicious reviews are marked Pending Validation and do not affect SWEA until cleared.
The system monitors for:
- —Abnormal rating bursts — 70 reviews in two hours
- —New-account clusters submitting similar scores
- —Same-network or device patterns
- —Extremely similar review text across submissions
- —Coordinated voting patterns
- —Sudden geographic anomalies
- —Restaurant social-media campaigns driving unnatural spikes
We do not publish every fraud-detection rule because bad actors would exploit it. But the core scoring methodology is fully public — this page.
Editorial Reviews vs. Community Rankings
SWEA's editorial opinions do not determine the live public rankings. The leaderboard is powered entirely by diner reviews.
The editorial review reflects one critic's experience on one visit. The community score reflects hundreds of real diners over time. Both are valuable. Neither overrides the other.
Immutable Score History
Every score change is recorded permanently. Nobody can credibly claim we're quietly changing the numbers.
The score history chart on every restaurant page shows this record publicly. Rankings are frozen by nothing — if legitimate reviews move a restaurant to #18, it goes to #18.
Internal Precision
Rankings use exact scores internally to avoid arbitrary ties.
Restaurant A at 94.716284 ranks above Restaurant B at 94.715991 — even though both display as 94.72. This eliminates ties and ensures rankings are always deterministic.
Ready to move the rankings?
Every legitimate review recalculates the leaderboard in real time.