About Loyalty.travel

Loyalty intelligence, built on evidence

Loyalty.travel exists to give sophisticated travellers an honest, source-anchored view of every major airline, hotel and rental loyalty program. We don't take compensation from programs to influence scores — the methodology below is the entire product.

Methodology

How we score loyalty programs

Every program is graded on a 0–100 composite assembled from eight weighted dimensions. The pipeline is reproducible, source-anchored, and refreshed on a quarterly cadence — or sooner whenever a program announces a material change.

1 · Web-grounded research

A Perplexity sonar-pro query is restricted to a whitelist of authoritative loyalty publications and consumer-research panels, with a 12-month recency filter. The model returns a structured brief covering valuations, award availability, elite perks, devaluations and CX signals — every fact carries a citation URL.

2 · AI synthesis

Anthropic Claude (Sonnet) ingests the research brief and is forced through a strict JSON tool-call schema. It cannot return prose — only an integer 0–100 per dimension plus a 1–2 sentence rationale anchored to the cited evidence.

3 · Deterministic composite

The composite is a fixed weighted sum (see weights below) computed in code, not by the AI. The same inputs always produce the same composite, so the AI never moves the headline number on its own.

The eight dimensions

weights sum to 100%

Redemption Value
weight 20%
What it measures. Cents-per-point realised on real bookings, sweet spots, fixed vs dynamic charts, fuel-surcharge policy on partner awards.
Benchmark. 85+ = >1.5 cpp on a fixed chart with no YQ. <40 = revenue-based at <0.5 cpp.
Earning Power
weight 15%
What it measures. Base earn rates, status bonuses, co-brand credit-card multipliers and transferable-points partners (Chase, Amex, Capital One, Bilt).
Benchmark. 85+ = multiple major transfer paths plus strong co-brands. <40 = stays/flights only with no card ecosystem.
Award Availability
weight 15%
What it measures. Saver-award openness, partner-award access through tools like AwardWallet, blackout reputation and phantom-availability complaints.
Benchmark. 85+ = consistently open saver space. <40 = published space rarely bookable.
Elite Benefits
weight 15%
What it measures. Upgrade hit-rates, lounge access, bonus earning, soft benefits, status-match policies and top-tier invitation perks.
Benchmark. 85+ = AA Concierge Key / UA Global Services-tier perks. <40 = badge with little real value.
Program Stability
weight 10%
What it measures. Devaluation history, predictability of pricing, frequency of unannounced changes over the last 5 years.
Benchmark. 85+ = stable 5+ years with advance notice. <40 = frequent stealth devaluations.
Partner Network
weight 10%
What it measures. Alliance and non-alliance partners, transfer-in partners, hotel/airline cross-redemption breadth.
Benchmark. 85+ = global alliance + multiple bank transfer partners. <40 = isolated single-brand program.
Customer Experience
weight 10%
What it measures. J.D. Power, Skytrax, Forbes Travel Guide ratings, IT reliability, customer-service responsiveness and call-centre wait times.
Benchmark. 85+ = top-quartile satisfaction across multiple panels. <40 = persistent CX complaints.
Transparency & Flexibility
weight 5%
What it measures. Published award charts, points-expiry policy, family pooling, redeposit/cancellation fees and account-freeze rules.
Benchmark. 85+ = public chart, free redeposits, generous expiry. <40 = opaque pricing and punitive fees.

Source whitelist

Research is restricted to recognised authorities on points, miles, hotel loyalty and airline service quality. Citations for every program are listed in its detail panel.

The Points GuyFrequent MilerOne Mile at a TimeView from the WingNerdWalletAwardWalletUpgraded PointsJ.D. PowerSkytraxForbes Travel Guide

Credibility guardrails

• Composite is computed in deterministic code, not by the AI.
• Each dimension score must be justified by a citation in the rationale.
• Every run is persisted to a score history table so devaluations and improvements are visible over time.
• Scores refresh quarterly on a schedule, plus on-demand whenever a program announces a material change.
• We surface the model name and computation date on every program detail view.

Loyalty.travel is independent — we accept no compensation from loyalty programs to influence scores. See the disclaimer for full terms.