Corrections

This log covers errors in the data or the method — a wrong price, a wrong rule, a missing country. Typos and styling fixes are not listed. Nothing is removed from this page once published.

To report something: use the contact form. Include the country, the provider, and what you saw — see Contact & corrections for what helps most.

2026-09-15

We said the price index is not changed after publication. It is updated every week.

The September price index said it was a snapshot that is “not silently updated”, and the edition list said a figure cited there “will still say the same thing next year”. Neither was true: each weekly price check rebuilt the edition with new figures. It was first published on 8 September; its figures now date from 14 September. Each rebuild also replaced the CSV with a new file and removed the old one. The pages now say that an edition is updated weekly during its month, and that anyone citing a figure should keep the CSV they took it from. Found while adding structured data that records each edition's publication and update dates.

The reports had the same problem. The reports index said each page “is dated and not silently revised”, and that nothing is edited after publication without a correction notice. Both reports are in fact recomputed with every weekly price check, and their pages did not say when their prices were read. Each report now says it is recomputed weekly and gives the dates of the prices behind its current figures.

Checking this also showed that a finished edition would not have survived: the build only produced the current month, so the first October rebuild would have removed the September edition and its data file. Past editions are now kept and no longer changed once their month is over.

We said prices were collected by hand. Most are read by software.

The homepage, every country page, the methodology page and the data licence said competitor prices were read “by hand”, and that there was “no scraper”. That was true when the site launched and stopped being true when we automated Airalo and Holafly — the wording was never updated, even as this log began describing the Airalo crawler. Airalo and Holafly are read by software; Saily in a browser session we run ourselves; Easy Travel Data from its catalogue. The pages now say so.

The date was also wrong, in the direction that matters. Each page showed one “prices checked” date, taken from the newest batch. Providers are read on different days, so that date made some prices look fresher than they were: the Japan page said 14 September while its Saily prices were read on the 7th. Pages now show the range, and each country page lists the date per provider. No price changed. Both problems were pointed out by an outside review of this site.

Airalo in Canada was listed at $32 for a week. The cheapest week is $14.

The crawler that fixed our Airalo figures on 8 September opens each country page and clicks through its three tabs. On Canada's page it came back with the Unlimited tab only — six plans instead of eighteen — and nothing flagged it, because a country with some plans looks like a country that was read. So for Canada we showed Airalo's cheapest qualifying week as unlimited data for 7 days at $32.00, when Airalo also sells 5 GB for 7 days at $14.00.

That changes who is cheapest in Canada: Airalo, not Saily, as this site had said. As with the September error, it overstated a competitor's price. It was caught by crawling every country twice, a week apart, and comparing how many plans each crawl found per country: 189 of 191 countries matched exactly; Canada and Argentina did not. Canada's second crawl was the complete one, confirmed by re-reading the page twice more. Argentina's was a one-off incomplete read and changes nothing here. From now on the Airalo crawl runs every week and a country whose plan count drops is not accepted until it has been re-read.

2026-09-14

We said nobody sells an eSIM for Italy. Easy Travel Data has sold five since before this site launched.

Our own homepage, the methodology page and the No local eSIM report all used Italy as the first example of a country with no single-country plan — “ask which eSIM to buy for Italy and nobody sells one.” That was wrong. Easy Travel Data has sold Italy 2 GB, 7 GB, 12 GB, 25 GB and 50 GB since 30 August. Each of those plans also lists San Marino and Vatican City, the two states inside Italy, and we counted any plan with three country codes as a three-country pack.

It is the same mistake as the United States plan we once labelled “3 countries” (below, 2026-09-07), and it survived because the fix for that one was a list of exceptions copied into this site by hand, with one entry. The list is gone. Whether a plan belongs to one country is now read from the same rule Easy Travel Data’s own shop uses, so the two sites can no longer disagree about it. Italy now counts as a country with plans of its own and has its own page; the report’s counts moved by one (78 → 79 with a plan of their own). Germany and Spain, the other two examples, were checked and remain correct: the closest plans on sale are genuine two-country packs, Germany with Austria and Spain with Portugal. No price on this site changed — the cheapest week in Italy is still a 35-country Europe pack.

2026-09-08

We read one tab of Airalo's price list and treated it as the whole list

Airalo splits a country's plans across tabs — Data, Fixed data and Unlimited — and our crawler only ever read whichever tab happened to load first. Where that tab was the unlimited one, we recorded an unlimited plan as Airalo's cheapest qualifying offer while a much cheaper fixed-gigabyte plan sat one click away, unread.

This is the largest error on this page. It affected 135 of the 191 countries we hold Airalo prices for. Airalo's qualifying price fell in every one of them, by a median of 2.40×: Bosnia and Herzegovina from $72.50 (unlimited, 30 days) to $12.00 (5 GB, 7 days), Bulgaria from $26.50 to $6.00, Israel from $26.50 to $7.00, Poland from $27.00 to $7.50. The plan count we hold for Airalo went from 1,171 to 2,666.

Every part of that error made this site's operator look better than it is. Recomputed against the full list on 8 September, Easy Travel Data was cheapest in 118 countries rather than 140, and was beaten in 74 rather than 52 — we had understated our own losses by 42%. Airalo won 65 countries rather than 23. The median qualifying week across the map fell from $14.99 to $13.00. Every affected page, the September price index, the Where we lose report and both CSV files were rebuilt from the full data.

Those four numbers are the state on 8 September, not today's. Easy Travel Data repriced its entire catalogue on 12 September, which moved most of its prices up, so the win and loss counts have moved with them. We are leaving the September figures above as written, because they are the record of what the error cost and correcting a correction into invisibility defeats the point. The current counts are computed on every build and live on the Where we lose report.

Found by a colleague opening Airalo's live pages and comparing them against ours, not by our own checks — which is the part worth dwelling on. Our regression tests were green throughout, because a regression test compares today's numbers with yesterday's and can only see data you already have. Missing data is invisible to it. The crawler now clicks through every tab on the page and fails loudly if a country returns fewer plans than before; The merge step now refuses any update in which a country's plan count falls, for all three providers — we checked that rule against the bad data and it flags 135 countries, where the guard we already had saw nothing at all, because the country count never moved. We also re-read the other two providers for the same tabbed structure: Saily lists every tier on one page, and Holafly sells only unlimited plans, so neither hides prices behind a tab.

Our own row skipped the rule every competitor had to pass

This page's comparison uses one rule: each provider's cheapest plan of at least seven days and at least 5 GB. Competitor rows were filtered by that rule in code. Ours was not. Where Easy Travel Data had no qualifying plan, the site fell back to its cheapest plan of any size and put that in the table anyway — so in Maldives and Timor-Leste a 1 GB plan sat in a column next to competitors' 20 GB and unlimited ones, and looked cheaper for it.

Two countries out of 197, and no competitor was ever affected — the bug could only ever favour us. Our row now passes the same filter. Where we have nothing that qualifies, the page says so and names the plan we do sell, instead of quietly entering a smaller one.

A second, smaller thing surfaced with it: where two of our plans cost the same, the site was picking one arbitrarily rather than the better one. Japan and China now show 5 GB for 30 days instead of the same-priced 5 GB for 15 days.

This is the same mistake as the very first entry in this log, in a different place: a rule applied to everyone except the party writing it. Found by a colleague reading the live pages line by line before promoting them — not by our own checks, which is the more useful half of this note. The September price index was recomputed and carries a correction notice.

Hawaii was counted as a country, and 43 pack labels were off by one

The supplier's coverage field mixes ISO country codes with the odd sub-division code: US-HI, the state of Hawaii, appears in 90 products alongside real country codes. We were splitting that field without filtering, so Hawaii appeared in this site's country list as if it were a country — making the headline "198 countries" wrong — and every "covers N countries" label that included it was one too high, on 43 countries.

The count is now 197. No price changed and no country moved between the single-country and regional-pack categories; this was a counting error, not a pricing one. The raw supplier field is untouched — only our reading of it changed.

Found by the team maintaining Easy Travel Data's own catalogue, who hit the same field from a different direction. It is the third time a code that is not a country has been read as one here; the earlier two were Puerto Rico and the US Virgin Islands, listed below.

We stated an analytics retention period we had not checked

The privacy page said Google Analytics was "set to the shortest retention Google offers". Nobody had verified that against the actual property setting — it was written from a reasonable assumption, which is not the same thing. The sentence now says what we can stand behind, and the figure will be published once it has been read off the property.

Same day it was written, before anyone could rely on it. Listing it anyway: the standard on this page is what we published, not whether it caused harm.

The privacy page said this site sets no cookies. It now does.

Google Analytics was added on this date, so the claim "no cookies at all, no third-party analytics" stopped being true. The privacy page was rewritten the same day to say exactly what is set, and analytics is off until you accept it in the EU, EEA, UK and Switzerland. The site's own counter is unchanged and still has no cookies and no identifiers.

Listing this is the point of the page: a statement that quietly stops being true is worse than one that was wrong to begin with, because nobody goes back to check it.

2026-09-07

Our own row was picked by a different rule than everyone else's

The site said each provider's row was its cheapest plan of at least 7 days and 5 GB. Easy Travel Data's row was not: it used a fixed 10 GB / 30-day plan instead. All 20 compared countries were affected. The stated rule now matches the applied rule for all four providers. Found by an outside audit, not by us.

Saily prices for the United States and China were Japan's prices

Tier for tier, the recorded figures for those two countries were identical to Japan's — they had been copied rather than read. Re-read first-hand: United States 5 GB $10.99 → $13.99, China 5 GB $10.99 → $15.99. 20 of 103 comparable tiers were wrong, across China, the United Arab Emirates, the United States, the United Kingdom and South Korea. The error made a competitor look cheaper than it is.

Holafly rows described as "unlimited, 1 day" did not exist

That $3.90 figure appears only as a "from" price in the page's structured data. Holafly's own price table starts at 3 days. Those rows have been removed rather than re-dated.

Thirteen countries were shown as having no 10 GB option

They had one — through a regional pack the site was not looking at. Canada ($31.99), Greece ($13.99), Vietnam ($12.99), Cambodia, Bulgaria, Israel, Montenegro, Oman, Serbia, Seychelles, Tajikistan, Gabon and Kosovo were all shown as blanks on the map.

Twelve countries were priced above the cheapest plan that covers them

The map preferred a single-country plan even where a regional pack was cheaper at the same specification. United States $13.99 → $12.99, United Kingdom $16.99 → $13.99, Taiwan $18.99 → $14.99, Colombia $62.99 → $31.99, Egypt $41.99 → $24.99, Mexico $41.99 → $29.99, and six others. Every one of these overstated a price.

A United States plan was labelled "3 countries"

The plan covers the United States, Puerto Rico and the US Virgin Islands — US territories on US networks, sold as "USA". The label is gone on the United States page and kept on Puerto Rico's, where it is accurate.

Maldives and Timor-Leste were missing entirely

Both have purchasable single-country plans and should have been on the map from the start. The country list is now built from a fixed reference of 198 destinations rather than from whatever the price query happened to return.

The stated map rule did not match the applied rule

The legend said the map used "a single-country plan where one exists, otherwise the cheapest regional pack". The code took the cheapest of either. The legend now says what the code does.

The homepage counted "compared across 4 providers" wrongly

That figure included countries checked against fewer than four. It is now split: the number price-checked against at least one rival, and the number checked against all four.