26 September 2026· ≈ 6 min read
The article of 7 September stays as it was published, on the pilot data. Here the same questions are answered with the first regular measurement, of 25 September: more shops, more cities, and a rule for when an example has enough sample to be quoted.
In this article
What changed in the measurement
The pilot counted only the pork wrap that carried a name: 1,113 of its 1,287 rows. The regular measurement looks at the price we publish for each shop, the pork wrap and, where that is missing, the cheapest other wrap, together with the name of that same item. That gives a name on 1,654 of 1,654 rows.
The definition changed because this is the price that enters the city table. To separate what comes from the definition and what from the new sample, the old definition is also computed on the new measurement, and gives 1,598 named rows. Where the two definitions disagree, we say so.
Cities with a sample of 4 shops or more went from 51 to 61. We use the same threshold here for something new: to compare the prices of the two names within a city, each name needs 4 shops.
The border held
In Thessaloniki 115 of 127 shops price from an item called a sandwich, 91%. In the pilot it was 91 of 98, 93%.
Sandwich is the majority name in 13 cities: Serres 100%, Giannitsa 100%, Thessaloniki 91%, Kozani 90%, Veria 86%, Katerini 84%, Halkidiki 81%, Florina 75%, Orestiada 75%, Kastoria 70%, Xanthi 68%, Larisa 62%, Kavala 57%. On 7 September we wrote "the twelve", and there were 12; now there are 13. New on the list: Kozani 9 of 10, Orestiada 3 of 4. Off the list: Alexandroupoli, from 5 of 9 to 5 of 13.
Orestiada has exactly 4 shops, the minimum for a city to enter the table. It is on the list, but on its thinnest sample.
Part of the change is the definition. With the pilot's definition, on the same new data, there are 12 majority cities: Xanthi falls below half. So 13 against 12 is not only new knowledge about the north; it is also which price we look at.
The Pindus, with one correction
The map now has 61 cities instead of 51. Each of the 10 new positions was checked against two independent sources, Wikidata and OpenStreetMap. Every majority city lies north of 39 degrees latitude, in Macedonia, Thrace and Thessaly.
On 7 September we wrote that Ioannina sits at zero. That was already wrong then: the pilot gave 1 shop of 34. The regular measurement gives 3 of 49, 6%.
Igoumenitsa, Arta and Corfu stay at zero. The word crosses the Pindus in a few menus, but as the usual name it does not.
Cities with 4+ shops where no shop uses it: 33, against 30 in the pilot. Off zero: Chania 1 of 37, Patras 1 of 47, Ptolemaida 1 of 11, Santorini 1 of 17. New at zero: 7, all of them cities without enough sample in the pilot.
In the north it is the same product
Where sandwich is the majority and both names appear, the median ratio of sandwich to pita is 1.000, across 11 cities. In the pilot it was 1.024, across 10. In Thessaloniki 5.00 € for the sandwich and 5.00 € for the pita.
That ratio, though, admits a city even when very few shops use one of the names: Kozani, 1 shop with pita against 9 with sandwich. With the rule of 4 shops per name, 5 cities remain (Thessaloniki, Katerini, Xanthi, Larisa and Kavala), with a median of 1.023 and a range from 1.000 to 1.075. The sandwich comes out equal or slightly dearer, by at most 7.5%: far from what a bigger product in bread gives, as Athens shows below.
As on 7 September, this does not compare two items inside one shop, but the median of the shops that call it a pita with the median of those that call it a sandwich, within the same city.
In the south, one example survives the rule
On 7 September we showed 4 cities where the word is rare and the sandwich costs at least 20% more than the pita: Lefkada, Heraklion, Ioannina and Athens. With the same criteria the new measurement gives 5: Lefkada 1 of 10, Ioannina 3 of 49, Santorini 1 of 17, Chania 1 of 37, Athens 10 of 634.
In 4 of them, though, the sandwich rests on fewer than 4 shops: Lefkada, Ioannina, Santorini and Chania. The median of so few menus is not a city price, so under the rule of this article we do not quote them as examples.
In Santorini the sandwich is a single item, at 12.70 €, 2.62 times the city's pita: one menu, not a market. In Ioannina the 3 are unlike each other. The dearest is a "SANDWICHARA" with chicken in a handmade bun. In another shop the pork pita was not recognised as pork, because its name spells the Greek Ch with a Latin X, so the shop's price came from the sandwich. The Ioannina median, 5.60 €, is exactly that shop. Missing that Latin X is an error of our own classifier and has been logged for a fix.
One city passes the rule: Athens. In Athens 10 of 634 shops price from a sandwich, at 6.65 € against 4.20 € for the pita. In the pilot it was 5 of 452, at 6.50 € against 4.20 €. There is a note here too: two of the 10 are a "gyro pork chop in a sandwich", a different dish that the classifier does not separate. Without them enough shops remain for the rule, and the median stays far above the pita.
Under the same rule, of the 4 examples of 7 September, only one would pass: Athens. The finding that in the south the word means a different, dearer product stands, but it rests on fewer cities than we wrote then.
What would break if the word came out
Without the word sandwich, 311 of 1,654 shops would lose their price, 19%. In the pilot it was 209 of 1,113, 19%. The figure is an upper bound: some shops also have another wrap without the word and would keep a price from it.
Thessaloniki would fall from 127 shops to 12. In the pilot it fell from 98 to 7. Its factor against Athens would stay at 1.19, while in the pilot it went from 1.19 to 1.24. The damage is not in the factor: it is that the country's second city would be measured on a handful of shops.
That is why the word stays on the list, and so does the safeguard that keeps the two products apart: in each shop we take the cheapest matching pork price. Where a shop has both a pork pita and a sandwich, the pita wins. Ioannina shows where this breaks: when the pita is not recognised as pork, it never enters the comparison.
See the same numbers live in the interactive dashboard.
The figures in this article are not written by hand: they are computed from the data series and change when it does. Price and wage coverage: August 2026. Dataset version 2026.09.26-mean-wage-source, last updated 2026-09-26.