Our methodology

Live award availability. We never query an airline or a hotel for open seats or rooms, so nothing on PointBagel tells you whether a redemption is bookable on your dates. A sweet spot is what a redemption costs when it is available, not a promise that it is. For live inventory, point.me, Seats.aero and PointsYeah search award space across programs. The airline or hotel is the final word.

How long a transfer will really take. A published upper bound is stored as exactly that: a ceiling rather than a schedule, not a typical or an average, so some transfers land in minutes and some use the whole window. An issuer's own estimate is labeled as an estimate, because an estimate is not a ceiling. Where nobody publishes a figure, the path carries none. Every figure keeps the unit its publisher used, because seven business days is not a week. We do not watch your transfer happen, so plan against the outer edge and never move points you cannot afford to wait on.

Which rows we have checked, and which we have worked out from others. Where a value comes from our own bookings or from corroborated published guides rather than a published chart, the row says so. Some programs publish no award chart at all; for those, the CPP section below explains what the top of the range actually is.

What any point value can promise. Point values are estimates and always will be, because what a point is worth depends on what you redeem it for. The narrower claim is the one we keep: we verify every ratio, award chart, and earning rule we cite.

We publish point values for 45+ rewards programs. Every benchmark is one program's value for one redemption method. Where the program publishes an award chart we read the chart; where it prices off the cash fare or a fixed peg we read that; where neither exists we use award prices we recorded ourselves and figures corroborated across published guides, and the row says which.

Every benchmark carries a written rationale you can read on the program page. 138 of 148 also name the source we read.

Point-value benchmarks are the sixth dataset in the table above. Each one records what its cents figure rests on, and every row we have read records the day we read it. That date is when we opened the document, not a promise that the price has not moved since.

Every program has three point values in our catalog. They differ in exactly one way: which kinds of redemption count, and how much.

One value-per-point number would be the wrong answer for most people, because the right answer depends on what you redeem for. Someone who flies domestic economy values United miles differently than someone booking transcontinental business. Every program page publishes the per-method benchmarks each tier is built from, and every value in the app carries the mix it assumes, so you can check the weighting instead of trusting it.

When a program devalues, we update the historical record so reports run against past dates use the value that was true at that time, not today's value backdated.

Our transfer-partner map covers 115+ partnerships across the major transferable-points programs (Chase UR, Amex MR, Capital One Venture, Citi TY, Bilt, Wells Fargo Rewards, HSBC Rewards, US Bank Altitude). Each entry stores the exact ratio: most are 1:1, some are 1:1.5 promo paths or 2:1.5 bonus paths. It also stores the timing its publisher actually states (an upper bound where one is published, an estimate where no bound is, and nothing where nobody publishes a figure) and any caveats that affect transfer feasibility.

Ratios are read from the issuer's own transfer page where one is public, and corroborated across published guides where it is not. When an issuer adjusts a ratio mid-cycle (Citi/Choice went 1:2 to 1:1.5 in April 2026, Bilt promo periods), the catalog gets a date-stamped update; the engine uses the ratio that was in effect when you ran a report.

The engine evaluates multi-currency redemptions: if you have 85K UR and 30K Bilt, both partner with Hyatt, so a Park Hyatt redemption can pool currencies.

Every card in the catalog stores its earning rates per category, but the headline rate is rarely the rate you actually earn. Amex Gold earns 4x at U.S. supermarkets - on the first $25,000/year. After that, the rate falls to 1x. If you spend $30K/year on groceries, your effective weighted rate is 3.5x, not 4x.

The engine computes weighted-average effective rates against your actual transaction history (when Plaid is connected), the spending you enter yourself, or a typical-spending model when neither is available. The result is the rate that matches your portfolio, not a marketing rate that ignores your spend pattern.

Category caps come from each card's terms. Every surface that ranks cards uses the capped rate, so a 4x card with an exhausted cap loses to a 3x uncapped card where it should.

The Application Calendar (Pro) tracks 30+ issuer eligibility rules across 14 issuers: velocity rules (per-window application counts), lifetime-bonus rules (per person or per product, depending on the issuer), product-conflict rules (you already hold a card in the family), and cross-issuer constraints.

Where an issuer publishes a clock, the rule stores its window and count. Where it publishes none, the rule stores a permanent per-product or per-family bar instead of an invented date. Every rule also carries its scope (personal or business, individual or cross-issuer) and its exceptions: soft-disclosure rules, deposit-relationship relaxations, product-family directional rules, cross-issuer cooldowns. Rules update when issuers change them; date-stamped catalog entries preserve the rule that was in effect for historical reports.

Users with more than one business entity can tag past bonuses to the entity that earned them, so untagged history does not block an application under a different business.

We track 250+ active welcome offers plus a verified-no-bonus dataset for cards we've confirmed don't currently offer one. Each offer carries the bonus amount, spend requirement, timeframe, and an offer-history record so we can tell you whether the current offer is at an all-time high, elevated above average, standard, or below average.

The recommendation engine factors offer history into card rankings - a hotel card sitting at an all-time-high 125K offer deserves a different urgency than one whose standing offer hasn't moved in a year. The 'All-time high' badge appears only when the current offer matches or exceeds every prior public offer we've recorded.

Sources: issuer pages, the promotional offer links issuers publish, and community-verified data points. An offer we cannot verify is not published.

When you connect a card via Plaid, our engine recognizes statement-credit reimbursements as they land in your transactions and toggles the corresponding benefit as 'used' on your benefits page. We track 600+ card benefits in all, and 170+ of them carry a detection signature so they get matched and marked automatically from your transactions (Amex Platinum credits, Chase Sapphire Reserve travel credit, the Amex Gold Dunkin' credit, and others).

Each benefit has its own detection signature scoped to that benefit's specific reimbursement pattern, so credits stay tied to the right benefit instead of getting confused with unrelated transactions.

Plaid cannot see every credit. In-app credits load a balance inside another company's app (Uber Cash, Lyft), and portal credits apply at the issuer's travel portal when you book. Neither posts to your statement.

So we treat each one as a single annual benefit: one amount for the year, counted by default at the value you set, with a switch to turn a credit off and a per-year option to record that you did not use it. Non-credit perks stay a toggle you switch on and value yourself.

Credits that do post to your statement are tracked per period (monthly, quarterly, semi-annual, annual), with proportional caps for partial periods.

Every change to the catalog runs the full test suite before it can ship. The checks cover CPP math, transfer ratios, earning caps, eligibility rules, benefit detection, and redemption strategies.

Nothing reaches production until the full test suite has passed on that exact commit; a change that raises a published value cannot merge without a written reason and the source we read it from.

Affiliate commissions never influence our recommendations; the same ranking math runs whether or not we earn anything.

The guarantee is structural rather than a policy: the modules that produce a ranking or a verdict may not reference an affiliate surface at all, and a check asserts that none of them does.