people killed — the price of Russia’s war against Ukraine since February 2022: 73,080 km² of seized land, — dead per square kilometre, one for every — hectares.
These are only deaths registered by Russian civil registries. The wounded, the missing and foreign mercenaries are not here — their deaths never enter the Rosstat register. Total combat losses are therefore far higher.
The map is coloured by place of residence: where people were taken from, not where they were killed. Click a region — the statistics below are rebuilt for it; click an address — lines are drawn from it to the places of death.
All breakdowns are scaled to —: — records without a stated cause are added to men, spread over front-line territories, and distributed in proportion to confirmed records elsewhere.
The place of death is known for most records and is filled in more often than the place of residence. Kursk and Belgorod oblasts are the fighting on Russian territory in 2024–2025. Records with no place of death usually carry no cause either.
the lines on the map run from the address of residence to the region of death: thickness follows the number of deaths
All breakdowns are scaled to —: — records without a stated cause are added to men, spread over front-line territories, and distributed in proportion to confirmed records elsewhere.
A record takes months to reach the database, so we watch the flow of new records rather than the date of death: over the last six months — — a month. By the end of the year the database will hold — military deaths — —. The scenarios below are for the whole calendar year, including records already received.
That is about the register. The dead are already more numerous: deaths of recent months are only partly recorded, and by the registration-lag curve — records have not yet arrived. With this correction, the dead as of — already number about —, and by the end of the year —: the register will lag by the same amount in December.
The share of records reaching the register N days after death is measured on the latest fully observed cohort (deaths 15–9 months ago): —. Registration keeps slowing (46 % of early-2024 deaths arrived within 30 days, 21 % of late-2025 deaths), so the correction is a lower bound. Records without a date of death (a fifth of new arrivals) are not corrected. The last two months are not divided by their share; the average expected total of the three months before them is used instead.
We count by the month the record appeared in the database, not by the date of death. Most records without a cause code carry no date of death, while the date of arrival exists for all of them and does not depend on the registration lag. Within each month every layer is included: combat deaths with code Y36 or Y89.1 or with combat wording, non-combat deaths of service members, and an estimate of combat deaths among records without a cause — from the excess of men against ordinary mortality of the same month, separately for front-line and rear territories. The last month is cut off by the extract date (—) and scaled up in proportion to the days.
Why not by date of death: one month after the death 26 % of the records are registered, after two 39 %, after six months 69 %, and in full only after a year. That is why any breakdown by date of death over the last year looks like a decline and is not one.
Code Y36 does not distinguish service members from civilians: the number also includes residents of shelled towns, children and the elderly among them. But the profile is men of conscription and contract age.
the age is exact, taken from the certificate: completed years on the day of death; it is not known for every record, so the sum is smaller than the overall number
* the last year is still arriving: its records keep coming, so the column is understated — see the forecast above
All breakdowns are scaled to —: — records without a stated cause are added to men, spread over front-line territories, and distributed in proportion to confirmed records elsewhere.
What is counted are death records, not rows: one record has 2.55 versions on average, and counting rows would inflate the number two and a half times. Each layer has its own documentary basis; the estimate is added separately and is never mixed into the count.
The first two rows are counts: codes Y36 and Y89.1 and combat wording in the certificate. The third and fourth contain an estimate, obtained from the excess of men among records with no cause stated, and it is not an exact number.
Digits are searched as a number: spaces and dashes do not matter, the last six digits are enough. Letters are searched in the place of death and place of birth, Russian word forms included: «Волчанск», «Купянский район». The Rosstat extract holds no names.
The same extract holds every death in Russia. The cause is taken from the code assigned by Rosstat, and where there is none, from the main-cause code of the certificate, then reduced to a three-character ICD-10 category. Click a cause — you will see its map and age profile.
clicking a row picks the cause: the map and the charts below are rebuilt