We re-judged 2,811 rejected venues. 606 earned their way back.

July 12, 2026 · The Retreat Register

This week the register put 2,811 of its own rejections back on trial. 606 won.¹ That is a 21.6% overturn rate, which means the register’s early quality gate had been throwing away roughly one real venue for every four pieces of junk it correctly cut. We found the flaw, measured it, fixed it, and re-tried every case. Here is the story, because a register that hides its corrections does not deserve the name.

The flaw: judging venues by their front door

The register’s early checks read a venue’s homepage and asked one question. Does this place host groups, retreats and meetings? That is a reasonable question to ask a homepage. It stops being reasonable the moment you look at how upscale properties actually build their websites.

The homepage sells the dream. The lake at dusk, the spa, the restaurant. The group business lives on an inner page, at /meetings or /groups or /events, that the front door barely mentions. A resort that does eight million dollars a year in conference business will often put none of it above the fold, because the room it is trying to fill on the homepage is a honeymoon suite.

So the gate was quietly failing exactly the properties a corporate planner wants most. The polished resorts, too elegant to shout “CORPORATE MEETINGS!” on their homepage, were being read as places that do not take groups. The blunter a venue was about wanting your company’s business, the more likely it was to survive the check. That is the opposite of a quality filter. It is a filter for loudness.

Measure first, then fix

Before anything got rebuilt, a random sample of rejections was pulled and re-read by hand. About 1 in 20 looked wrongly cut. That is a small percentage and a large absolute number against a rejection pile 2,811 deep.

Two numbers on this page describe the same problem and they do not match, so here is the honest version. The 1 in 20 came from a hand check of just twenty venues. The full re-check read more of each venue’s website, including the meetings, groups and events pages the first screening never opened. Reading further found more mistakes than the hand sample did, which is why the finished rate landed at 21.6% rather than 5%. Checking all 2,811 that way brought 606 venues back, kept 2,193 out, and set 12 aside as obvious junk.¹

The hand sample was never the answer. It was the evidence that justified spending the compute to re-try every case.

The verdicts

Those three outcomes add to 2,811 exactly, and the 12 matters. A summary that quotes only the 606 and the 2,193 is 12 short of its own total, which is the kind of small arithmetic gap that tells a careful reader a number was copied rather than checked.¹

The whole pile was re-tried in twelve minutes.

Being restored to the pipeline is not the same as being published. On July 12, 2026, the day of the re-judging, the register held 1,374 curated corporate retreat venues.² The 606 rejoined the queue of candidates behind that number, not the published list in front of it. Rescued does not mean waved through.

What 2,193 rejections upheld actually tells a venue owner

The 2,193 that stayed out share one property. There was no group evidence anywhere on their site. Not on the homepage, and not on any inner page the rebuilt judge could reach.

If you run a property that takes corporate groups and you cannot find a page on your own website that says so, that is the finding to act on. A single page at /meetings or /groups, naming your meeting rooms, your layouts and your overnight capacity, is what separated the 606 from the 2,193. It is also the page a planner looks for first, which is why the register was wrong to ignore it and why the fix mattered.

What the rescue did to the size of the register

The re-judging was one of two changes made on the same day. The second was a fix to brand families: hotel groups that run many properties on one website are now tracked per property, “a flaw that had been hiding whole resort markets, including most of Hawaii’s.”³

Together they moved the register faster than anything before or since. On July 12, 2026 it held 1,374 curated venues.² Five days later, on July 17, 2026, it held 2,281,⁴ which is a 66% increase in five days. That is the 606 rejoining the pipeline, the brand-family fix unhiding whole resort markets, and the ordinary daily sweeps, all landing inside one week.

The other direction, and the day the register got smaller

An audit that only adds things back is not an audit. It is a growth story with better manners.

The day after the register hit 2,281 venues, it published 2,125.⁵ That is 156 venues gone in one day, and it is not a typo in either page. The build log entry for July 18, 2026 reads “Quality passes removed venues, and those cuts are part of how we check.” Two separate pages, written for two separate purposes, agree that the register got smaller on purpose that day. Louisiana alone went from 92 listings to 80, and California from 168 to 162.⁴⁵

That was not the first cut, and it was not the largest. On July 11, 2026, “a full read of every published row, every name, every website, cut 274 entries no automated filter had caught: budget hotel chains and dozens of template-name fakes pointing at unrelated websites.”⁶ The day before that, 95 convention and expo facilities were moved off the public list to a review shelf, described in the log as “real businesses, wrong category.”⁷ Two days later, 88 of those borderline properties were still on the shelf, waiting for a decision from a person.⁸

The fashion retailer and the online dictionary that messy public data once pushed into the catalog as “venues” are still out.⁹ Six hundred and six real properties are back in the pipeline where they belong. Both halves of that sentence are the product.

Why we are telling you this

Every directory you have ever used has a rejection pile, and almost none of them audit it.

Errors only get fixed in the direction users complain about, which is junk that slipped in. The silent error, meaning good venues that never appeared at all, stays invisible forever, because nobody complains about a listing they never saw. You cannot file a bug report about a resort you did not know existed.

A register has to audit both directions. Junk out, and wrongly-cut venues back in. The numbers on this page are the record of one day when both happened, and the numbers on July 18 are the record of a day when the register chose to be 156 venues smaller than it had been the night before.

What this pass did not do

Three honest limits, so nothing here is read as more than it is.

The re-judging happened once, on July 12, 2026. It was not a schedule and nothing on the site re-runs it on a timer. The evidence that quality passes recur is the build log itself, which records cuts on July 18, July 22 and July 24, 2026, and nothing beyond those dated entries is being claimed here.

The 606 were restored to the pipeline, not to the published list, and the register has never published how many of them ended up live.

The last exact total the register published was 2,152 curated retreat venues on July 19, 2026.¹⁰ Every count on this page carries the date it was measured, and there is no newer one to give you.

Have you read: 9 in 10 retreat venues won't show you a price

Figures from The Retreat Register review pipeline, July 12, 2026: 2,811 rejected properties re-judged, 606 restored to the pipeline, 2,193 rejections upheld. How the register is built: the methodology.