We checked all 135 vacation listings live on As-Home in August 2026. Seven carry a map pin that is wrong, missing or unplaceable. Four display a location but sit in no region filter at all. The location a guest reads on a listing card comes from a different database column than the one our filters actually query.
Two location systems on one listing card, and they do not agree
Method: full census of every approved, publicly visible vacation listing, taken from the platform database on 7 August 2026 and reconciled against the live search API. Every unit renders. Source: As-home platform data, n = 135.
What we checked, and what we could not check
This is a census, not a study. Every approved vacation listing, all 135, read from the database rather than the page: location fields, coordinates, photos, amenities, descriptions, policies.
Two limits first. It says nothing about which listings get booked: eight weeks of trading on a platform that opened in June 2026 will not support that claim. And it is one platform. Egypt holds far more inventory than we do.
There is a third limit that we walked into ourselves. An earlier version of this audit was built by reading the public collection page rather than the database. It produced a confident conclusion that turned out to be wrong. A rendered page is evidence about the rendering. It is not evidence about the data underneath, and the gap between those two things is what this article turned out to be about.
The location a guest reads is not the location we search on
A listing card shows one line of location text. That line comes from a free-text field called city, typed by whoever created the listing, printed exactly as entered with no validation, no normalisation and no spell check.
The search filter does not query that field. It queries three other columns: region, destination and compound. There is no city filter anywhere in the search path.
So a guest reads one column and searches another, and nothing in the system reconciles them. Across 135 listings there are 26 distinct labels a card can display, against 19 locations the filter recognises.
Where the two systems disagree, and how many units sit in the gap
| The card reads | The filter files it under | Units |
|---|---|---|
| hurghada | Red Sea | 39 |
| Marsa Matrouh, Matruh, Matrouh | North Coast | 12 |
| Suez | Ain Sokhna | 4 |
| Alexandria | North Coast | 3 |
| Ras Sedr, Ras Sudr | Sinai | 2 |
| Ain Sokhna | North Coast | 1 |
Method: distinct values of the free-text city field, grouped against the normalised region each unit files under. Labels differing only in spelling or capitalisation are counted separately, because the system treats them separately. Source: As-home platform data, 7 August 2026.
Read the highlighted row. Four units display Suez, the governorate, which is not wrong. They file under Ain Sokhna, which is where a guest searches. Both are defensible and they point at different words.
The largest group on the platform, 39 Hurghada units, renders in lowercase next to properly capitalised neighbours, because nothing checks. Marsa Matrouh is typed three different ways. Ras Sedr is typed two.
None of that breaks the filter. It does something subtler: it teaches a guest that the location text means something, when it is the one part of the card the system ignores.
Four listings that look located and cannot be found
Four units carry no region and no destination at all. Their cards read Alexandria, Ain Sokhna, South Sinai and North Coast, so they look perfectly placed. They are unreachable from every regional browse page on the site.
This is the worst version of the split. The label creates an expectation that the filter cannot honour, and neither the host nor the guest has any way to see it. A host looking at their own listing sees a location. A guest browsing North Coast never sees the listing.
One chalet says Sokhna. Its pin is on the wrong sea.
Region and destination are not typed by anyone. They are computed from the map pin by point-in-polygon lookup. That is the right design, and it has one consequence: the classifier cannot detect a bad pin. It faithfully converts one into a confident, wrong answer.
One listing is titled for Porto Bay in Ain Sokhna. Its city field says Ain Sokhna. Its pin sits on the Mediterranean near El Alamein, roughly 400 km away on a different sea. The title said Sokhna, the city said Sokhna, and both were overruled by a coordinate nobody checked. The system then computed a kilo marker from that pin and stored it. Every derived value on that listing is confidently wrong.
A second listing states Kilo 67 in its title. Its pin implies Kilo 12.7, some 54 km away, dropping it inside the Alexandria polygon instead of North Coast.
That second one has a detail worth sitting with. The classifier does read a stated kilo out of the title, and that stated value overrides everything else. But it only runs that check when the unit has already been classified as North Coast. Because the bad pin resolved to Alexandria, the branch never ran, the host's own explicit Kilo 67 was discarded, and the kilo was stored as empty. The one piece of evidence that would have caught the bad pin was thrown away because the bad pin had already been believed.
Seven of 135 units have some coordinate problem: two demonstrably wrong pins, two valid pins that fall outside every polygon we hold, and three with no usable coordinates. That is 5.2% of live inventory, and all seven are ours to fix.
What else the audit found
Location is not the only field where the default quietly wins.
Almost nobody sets a cancellation policy, because almost nobody has to
| Field | Set | Left at default or empty |
|---|---|---|
| Cancellation policy | 1 | 134 |
| Amenities | 101 | 34 |
| Bedroom count | 106 | 29 |
| Five or more photos | 124 | 11 |
| Region | 131 | 4 |
Method: field-level completeness across every approved live vacation unit. Amenities counted as distinct canonical amenities after synonym resolution, so a unit listing both Kitchen and Equipped Kitchen counts once. Source: As-home platform data, 7 August 2026, n = 135.
One listing out of 135 has a cancellation policy other than the default. That is not 134 hosts choosing to offer nothing. It is a form field that nobody is required to touch, which is a product problem wearing a host problem's clothes.
Thirty-four listings advertise no amenities at all, a quarter of live inventory. Almost every filter a guest uses on the vacation homes page is an amenity filter, so those 34 drop out of all of them. The median listing carries six.
A quarter of live inventory advertises nothing at all
Method: distinct canonical amenities after synonym resolution, so a listing recording both Kitchen and Equipped Kitchen counts one. Deduplication changed the count on exactly one unit across 727 raw assignments. Source: As-home platform data, 7 August 2026, n = 135.
Two listings have no photographs and still appear in results. Eleven have fewer than five. The median is 15, and the most complete carries 67.
One number we are deliberately not publishing as a finding: only 26 of 135 listings record WiFi. That is not a claim about Egyptian vacation homes. It is a claim about our intake form, and treating it as the former would be dishonest.
What this does not show
Nothing here measures bookings. Correlating any of these fields against booking outcomes needs a clean quarter of trading, and our owner approval flow was failing for part of the period, so a comparison run today would measure our own bug rather than host behaviour.
It also cannot say which listings a guest wanted to find. A unit missing from the North Coast filter is a loss only if someone looked for it, and we cannot see the search that never returned it.
What to change this week
Most of the above is ours. Four things are worth ten minutes if you host with us:
- Open your unit page and check the map pin, not the address text. The pin decides which filters you appear in. The text decides nothing.
- If your title states a kilo marker, check it survived. If the pin was wrong, the kilo was thrown away with it.
- Tick every amenity you have. Each one you skip removes you from a filter a guest is using right now.
- Set a cancellation policy on purpose rather than leaving the default, so a guest comparing two units knows which one is telling them something.
On our side: the card will render the location the filter actually uses, the four unclassified units will be placed, and the kilo check will run regardless of which region a pin resolved to. We will publish the count again when it is done.
Thinking about where to list your unit?
Guests pay by card in Egyptian pounds against a price with tax already in it, and you are not fielding deposit negotiations over WhatsApp.
List your property on As-homeQuestions hosts ask
Why is my vacation rental not showing up in filtered search?
On As-home, filtered search runs on region and destination, which are computed from your map pin, not from the location text on your listing. If the pin is wrong or missing, the listing can display a correct-looking location and still be absent from that area's browse page. Check the pin first.
Does the location text on a listing affect search results?
Not on As-home. The city line shown on a listing card is free text and is not queried by the location filter. It matters for how a guest reads the card once they see it, and for typed free-text searches that match no known place, but it does not decide which filtered results you appear in.
How many photos should a vacation rental listing have?
The median As-home listing carries 15 and the most complete carries 67. Two live listings carry none and still appear in results. There is no threshold below which a listing is hidden, so the practical answer is that photo count is entirely a persuasion question rather than a visibility one.
How many amenities should I list?
Record every amenity the unit genuinely has. The median As-home listing carries six distinct amenities and 34 of 135 carry none at all. Almost every guest-facing filter on the vacation homes page is an amenity filter, so an unrecorded amenity is a filter you silently drop out of.
Does instant booking get more bookings?
We are not publishing a number on this yet. Forty-five of 135 live listings have instant booking on, but our owner approval flow was failing for part of June and July 2026, so any comparison across that period would measure our own fault rather than host behaviour. We will publish it after a clean quarter.



