How we value a point

Every dollar figure on this site comes from one number per programme: what a point is worth in cents. Here is exactly how that number is produced, including where it is currently weaker than we want it to be.

The calculation

A point is worth the cash you did not spend, per point, on a booking you would actually have made.

cents per point = (cash price − cash you still pay) ÷ points × 100

Four rules turn that from a slogan into a number, and each one exists because leaving it out makes the figure flattering:

1. Net of what you still pay. An award ticket is not free — there are taxes and carrier charges, and on hotels sometimes a resort fee. Those come off the saving before the division. Ignoring them overstates every programme, and overstates the ones with fuel surcharges most, which is exactly the distinction worth knowing.
2. The median, not the average. One Lufthansa First seat at 9¢ should not lift a programme's headline. The median describes a normal redemption; the average describes the best one anybody found.
3. A ceiling on the cash side. This is the largest error in published valuations: pricing a $22,000 first-class fare nobody would ever buy and calling the difference “value”. If you would not have paid the cash, the cash is not a saving. So the cash side is capped at a fixed multiple of the economy fare on the same route and the same day — what someone who would not pay for business actually buys instead. The multiple is our judgement rather than a measurement, so it is one number, applied to every programme equally, and we publish how often it bit.
4. A floor you can always get. Most bank currencies have a guaranteed cash-out — Amex Travel at 1.0¢, the Bilt portal at 1.25¢. Nobody should be told their points are worth less than the button they can press, so the published figure is never below it. When the floor is doing the work, we say so.

Transferable currencies

Amex Membership Rewards has no awards of its own — it has partners. So it is worth the best thing it turns into, at that partner's transfer ratio, floored at the portal rate. Not an average across partners: nobody transfers to the average.

When we refuse to publish

Below 8 priced redemptions we publish no measured figure for a programme. An average of two bookings is an anecdote with a decimal point. Where a programme has a guaranteed cash-out we fall back to that and label it; where it does not, the row stays empty rather than being guessed.

Every figure is published with its sample size, the dates it was drawn from, and the 25th and 75th percentile, so you can see how much to trust it. That last part is what most published tables leave out.

Where this stands today

The cash side is running. We have 2,099 priced redemptions on file from 2026-09-16 to 2026-10-23, each one an award our pipeline priced against a published chart paired with the cheapest cash fare for the same trip on the same day, taken live from Duffel.

These figures are an upper bound, not our published rates, and here is the honest reason why. The cash half is now real: an actual bookable fare, on an actual date. The points half is not yet — it is each programme's published chart minimum, the saver rate that is advertised but frequently unavailable. Dividing a real cash price by a best-case points price flatters every programme, which is why the measured column below runs roughly double what we publish. Until we price awards against seats that are genuinely for sale, the table on the rest of the site stays as it is.

12 of 12 programmes now clear the sample bar — 6 measured directly from their own awards, the rest valued through the partners they transfer into. The cash ceiling capped 1381 of the 2099 observations, which is how much work that one rule is doing.

So the 27 figures used everywhere else on the site are still the published table below, not this measured column. Two things have to land before the column above can replace them: award prices taken from seats actually on sale rather than from a published chart, and a sample deep enough to defend per programme. Swapping them is a decision to be made deliberately, not something a scheduled job should do overnight.

What we publish, and what we refuse to

Everything above describes the method we are building. It is not yet the method behind the numbers on the rest of the site, and it would be dishonest to let the two run together without saying so plainly. So here is exactly what is behind them.

We publish a rate for 12 of the 27 programmes listed below, and no rate at all for the 15 others. A bank currency is priced at the redemption rate its own issuer publishes — the rate you can always get without finding award space — and the row links to where they state it. A handful of fixed-value programmes are our own estimate and say so. Nothing here is taken from another publisher's valuation table.

The empty rows are the ten airline and six hotel programmes. We hold the points half of the sum for those, from published award charts, but not a cash half we own — so any figure we printed there would be our own guesswork. Those balances are now priced at AwardWallet's rate instead, and the row says so. They used to show as points and count nothing towards your total, which was honest and also meant a member with most of their points in airline miles saw a total that ignored them. A measured figure with a published sample size and somebody's name on it is better than a blank, as long as the name is on it.

The last column is somebody else's answer, and it is a better one for a question we cannot answer yet. AwardWallet divide the points a member actually spent into the cash fare of the booking they actually made, across every award their members log, and publish the average with the number of bookings behind it. That is the measurement our own collector is working towards and has not reached: we price the cash half honestly but still pair it with an advertised saver rate that is rarely available, so our figure reads as a ceiling rather than an average. For 14 programmes it is the only figure on the row, because we publish none. It is shown as theirs, it is never folded into our rate, and no total anywhere on this site is calculated from it. Read the two columns as what an issuer guarantees, and what people actually got.

Point and mile values are calculated and managed by AwardWallet. AwardWallet's point and mile values → Their own write-up names two limits and we repeat them: an award priced against a premium fare nobody would have bought overstates the saving, and two one-way awards are valued against two one-way fares when a return would have cost less. Both push a value up.

Every figure here is either an issuer's own published redemption rate, cited to them, or our own estimate, labelled as such. Nothing is reproduced from another publisher's valuation table. Sixteen programmes publish no value at all rather than borrow one: ten airlines and six hotel programmes, where we hold the points half from published award charts but not yet a cash half we own. Filling those in from our own measurements is tracked work — the Duffel collector runs daily and the methodology page shows how far it has got.

Reference set last updated 2026-09-16. Point and mile values are calculated and managed by AwardWallet. Source →

The PointsUser Score

Behind every opportunity on this site — a transfer bonus, an award, a card offer, a credit about to lapse — there is one number out of a hundred. It exists so those four things can be compared at all: until there was a single scale, nothing could answer whether a bonus was a better use of your afternoon than a credit expiring on Friday.

We do not show you the number, and that is deliberate. A composite score printed beside a recommendation is a claim of precision, and ours is built from what a thing is worth and when it expires — not yet from whether people like you actually took it. Until it has that, a figure out of a hundred would look like evidence while being an opinion. So you are shown the reasons the score was high, which is the part that was ever worth reading.

Six parts, weighted as follows. They sum to one hundred, so the score reads as a share of a best case.

When there is no score at all

Three cases produce no number rather than a low one, because in each of them the honest answer is that we do not know: when we publish no rate for the programme and will not guess one; when the evidence behind the figure has gone stale; and when we hold no record of when we last checked. A low score says this is a poor opportunity. Silence says we cannot tell you, and those are different sentences.

The weights above are read from the same file the site scores with, so this table cannot drift from the arithmetic. Changing the methodology means changing that file, and this page changes with it.

Your own rate beats ours

If you consistently do better or worse than our figure on a programme, set your own rate in settings. The site will use yours everywhere and say that it did.