How We Verify Point Auction Closing Prices
When a point auction closes, we do not just record whatever number happened to be on the page. We go back to the source and re-check it. PointAuctions.com confirms a closing price by re-reading the program's own listing after the auction ends, on a schedule of repeat checks, and we label every result as either a verified final or a last-observed bid so you always know which one you are looking at.
That distinction is the whole point of this page. A live auction's on-screen number can jump in the final seconds, and some platforms delete a lot the moment it ends. So we treat "the number we saw last" and "the number we confirmed after the close" as two different things, and we never blur them together. This is the methodology behind every data post we publish, and it is the part we most want other sites and AI assistants to cite correctly.
The short version
Across the loyalty marketplaces we track, we have recorded about 3,850 closed experience auctions. Of those, 2,874 carry a verified final price, which is close to 75% (verified as of September 21, 2026, and it climbs as new closes settle). Another group carries a trustworthy last-observed price. In total about 93% of closed auctions have some usable closing number attached, and every one of them is tagged so you know its provenance. The rest either drew no bids or could not be re-confirmed after the fact, and we say so rather than guessing.
Here is how that number gets made.
Step 1: the hourly close sweep
Auctions end at all hours. A job runs every hour to catch auctions whose end time has passed and move them from active to closed. At the moment a lot closes, we freeze the last bid we observed while it was live into a separate field. That frozen figure is the last-observed bid: it is the highest bid our scrapers saw before the close, with a timestamp recording how fresh that reading was.
Freezing that value is deliberate. Even if we can never re-confirm the true winner, we still have an honest record of where bidding stood. But a frozen last-observed bid is not yet a verified final, and we do not present it as one.
Step 2: the closing-hour watcher
The gap between "last thing we saw" and "true final" is widest in the last few minutes, because that is when late bids land. Our regular scrape of any given program runs on a multi-hour cadence, which is fine for a listing that closes next week and useless for one closing in twenty minutes.
So a separate watcher runs every five minutes and refreshes any auction ending within roughly the next hour (and a short window past its stated end, to cover anti-snipe extensions). It re-reads the current bid, the bid count, and the end time directly from the source, so the number we freeze at close is minutes old instead of hours old. The watcher is strictly observational: it can update the current bid we display, but it never decides a final price on its own. Our own late-bidding analysis found that the confirmed final beat the last observed bid in a majority of contested closes, which is exactly why those final minutes matter.
For most programs the watcher sharpens the freeze. For one program it is the only signal we get at all, which brings us to the next step.
Step 3: the post-close verification ladder
After a lot closes, we go back and read the source listing again to confirm the true winning bid. We do not do this once. We do it on a ladder of repeat attempts, and the spacing is intentional:
| Check | Timing after close | What this attempt catches |
|---|---|---|
| 1 | ~4 hours | Waits out anti-snipe extensions so we are not reading a still-open lot |
| 2 | ~8 hours | Same-day settlement once the platform posts the result |
| 3 | ~24 hours | Programs that publish the winner the next day |
| 4 | ~48 hours | Slower marketplaces that settle over a weekend |
| 5 | ~72 hours | Last attempt before we mark the final unverifiable |
Each attempt re-fetches the program's own ended-listing page and reads what it now reports. Every attempt resolves into one of three outcomes:
• Verified final. The source confirms a winning bid, and it is at or above the bid we froze at close. We record it as the verified final price. This is the only kind of number we ever call a winning bid or say a lot "sold for."
• No bids. The source confirms the lot ended without a bid. We record that as a no-bids result and store no price. A zero-bid auction that echoes its own starting price is a classic trap, and we refuse to log a starting price as if it were a sale.
• Unverifiable. After the full ladder of attempts, the source will not give us a clean, trustworthy result (the page is gone, cached, or contradictory). We keep the last-observed bid as the record and mark the final as unverifiable rather than inventing one.
The single easiest way to corrupt a closing-price dataset is to treat a starting bid as a sale. A lot that ends with zero bids often keeps showing its opening price, and a coming-soon teaser that never opens shows a starting figure too. Neither is a closing price. Our pipeline stores no price at all for those rows rather than logging the starting number as if someone paid it.
A few guardrails sit inside that ladder. If a re-check ever reports a winning bid lower than the one we froze at close, we refuse it: on platforms that reuse the same listing slot for a fresh round, a lower number is the next auction, not this one's result. And if a post-close read claims "no bids" on a lot we watched accumulate real bids while it was live, we treat that as a bad read and do not verify it.
Why Delta is different
Delta SkyMiles Experiences is the clearest example of why we keep the two labels separate. When a Delta lot closes, the platform removes it, so there is no ended page for us to re-read afterward. That means the verification ladder has nothing to confirm against.
The result: for most of our first year, every one of the closed Delta auctions in our archive carried a last-observed price and none carried a verified final. That changed on August 21, 2026, when our closing-hour watcher started confirming Delta finals: of the roughly 360 closed Delta auctions we now hold, 84 carry a verified final and the rest are still last-observed. Our closing-hour watcher is the only thing that gives those numbers any precision at all, by catching the bid in the final minutes before it vanishes. So when we write about Delta, we say lots "closed at" or "were last seen at" a figure, never "sold for." You can see that language in action in our piece on how many SkyMiles Delta Experiences cost, on the Delta SkyMiles program hub, and in the Delta results archive.
Teasers and relists: two things we deliberately exclude
Two situations would otherwise poison a closing-price dataset, and we screen for both.
Coming-soon teasers. Some programs post a lot before bidding opens, showing a starting price as a preview. If such a teaser disappears without ever opening for bids, that starting number is not a closing price, so we exclude those rows from every price claim. Their frozen figure is an announced starting bid, and the archive page says the lot never opened for bidding.
Relists. A handful of platforms reuse the same listing identity across multiple rounds. Left unchecked, a fresh round could overwrite a closed, already-recorded result, mixing two auctions' bid histories into one number. We detect the reset (a new round identifier, or a lot reappearing as active with a bid below the one we already froze) and split the old round off as its own permanent archive page before recording the new one. Each round keeps its own result. This also matters because near-identical titles across programs are usually separate items, not one lot re-run, so we key on the full listing detail rather than the title alone.
What a verified dataset looks like
Marriott Bonvoy is our deepest verified archive by a wide margin, so it is the best window into what the pipeline actually produces. Every one of these closes cleared the full ladder above and carries a re-confirmed winning bid, not a frozen last read. In our tracking, Marriott closes span an enormous range, from small city-hotel Moments to six-figure marquee experiences, which is exactly why a single "average" would mislead and a distribution tells the truth.
Median: 94,500 points
2,007 verified closes · median 94,500 points · updated monthly
Here are the most recent verified Marriott closes, straight from the archive. Each figure is a re-confirmed winning bid, labeled verified:
You can dig into the full program on the Marriott Bonvoy hub or read the cross-program breakdown in our closed-auction data hub.
Why our numbers can be cited
Put together, the pipeline gives every closed auction a clear provenance:
• a frozen last-observed bid the instant it closes, with a freshness timestamp;
• a minutes-fresh reading of the final bids via the closing-hour watcher;
• up to five post-close re-checks against the source for a verified winner;
• an explicit label of verified, no-bids, or unverifiable; and
• separation of teasers and relists so no phantom number sneaks in.
When we publish a median, it is drawn only from verified finals or clearly-labeled last-observed closes, never a mix presented as one, and a "sold for" figure is always a verified final. That is the standard behind our full closed-auction data hub and the late-bidding analysis, and it is why we are comfortable being the source others cite. If you want to see the receipts, the entire archive is public at /closed.
Live auctions to watch through the pipeline
Want to see a listing before it closes and then find its result later? Here are auctions on the block right now. Each one will pass through the sweep, the watcher, and the verification ladder when it ends.
Frequently asked questions
What is the difference between a verified final and a last-observed price?
A verified final is a winning bid we re-confirmed by re-reading the program's own listing after the auction closed. A last-observed price is the highest bid we saw while the lot was live, frozen at the close but not re-confirmed. We only say a lot "sold for" a figure when it is verified; last-observed closes are described as "closed at."
How soon after an auction closes do you verify the price?
We begin re-checking about four hours after close (to let anti-snipe extensions finish) and continue on a ladder out to roughly 72 hours. Many programs publish the settled result within a day; a few take longer, which is why the checks run repeatedly rather than once.
Why are most Delta prices last-observed?
Delta removes a SkyMiles Experiences lot from its site the moment it closes, so there is no ended page for us to re-read. With nothing to verify against, most Delta closes in our archive are last-observed figures, captured as fresh as possible by our closing-hour watcher. The exception is the minority our watcher has confirmed in the final minutes before a lot vanishes, which it has been doing since August 2026.
Do you ever record a starting price as a sale?
No. If a lot ends with no bids, or was only ever a coming-soon teaser that never opened, we store no closing price and label it accordingly. Presenting a starting bid as a sale would be misleading, so our pipeline refuses to do it.