Methodology
& Glossary

Figures on this tracker fall into three tiers of evidential strength. They are not interchangeable, and the single most common analytical error in drone-war reporting is summing across them. A photo-confirmed loss and a claimed strike are different kinds of object.

Where both a claimed and a confirmed figure exist for the same thing, this tracker aims to show both rather than reconcile them. Averaging across tiers produces a number belonging to no methodology at all.

Neutralised, not intercepted

Air defence rates on this site report neutralisation: weapons kinetically destroyed plus those suppressed by electronic warfare and lost before reaching their target. This matches how the Ukrainian Air Force reports its own headline figure, and it is the basis on which comparable trackers publish.

The distinction is not cosmetic. Counting only kinetic destruction understates the all-time rate by roughly ten percentage points, because electronic warfare accounts for a substantial and varying share of drones brought down. A rate computed on one basis cannot be compared against a rate computed on the other.

Damaged and destroyed are separate columns

USF strike figures report targets hit and, within that, targets destroyed. The two are nested, not additive: a category showing 3,675 hit and 834 destroyed means 3,675 targets were struck, of which 834 were assessed as destroyed. Summing the columns double-counts.

Monthly versus cumulative

Series labelled monthly contain that month alone. Series labelled cumulative contain everything to date. Where a source publishes a running total under a monthly heading, it is relabelled here rather than silently converted, because the conversion is not always recoverable.

Floors versus exact counts

Where a source publishes "more than 10,000" or "almost 20,000", the figure is recorded as a floor and marked as such — on charts, by a hollow rather than solid bar. Treating a floor as an exact count creates false precision and, worse, false month-to-month movement when a floor in one month is compared against an exact figure in the next.

Gaps are left empty

Months with no published figure are omitted, not interpolated. A missing point is information; an invented one is not.

Event logs are not samples

The USV and UGV logs record notable, well-sourced individual operations — firsts, capability milestones, unusually well-documented actions. They are curated, not systematic. Counting entries per month measures reporting attention, not operational tempo, and the two move independently.

Long-running open-source datasets accumulate distortions that look like findings. These are the ones known to affect data on this site. Each has produced an apparent trend that is not real.

Reporting convention changes

Between roughly October 2024 and July 2025 the Ukrainian Air Force reported electronic warfare suppression in a field separate from kinetic destruction, then merged the two. Any series built on the destroyed field alone shows a collapse in interception through that window and a recovery afterwards. Neither happened. The apparent trend is entirely an artefact of which fields were populated when.

Taxonomy revisions

USF revised its target categories during 2026. Categories were renamed, merged and added — copter drones became multirotor drones, robotic complexes became UGVs, and roughly a dozen new headings appeared. Category-level series do not run continuously across that break, and matching old to new categories involves judgement that would be invisible in a chart.

Retroactive correction

Territorial assessments are revised as control is clarified, often weeks later. A published monthly figure may incorporate corrections belonging to earlier months, which means the headline number for a month is not the same thing as that month's own change. Where the difference is known, it is stated alongside the figure.

Aggregation of unlike categories

Some air defence records log multiple weapon types in a single row — a strike package combining surface-to-air missiles and cruise missiles, for instance. Any filter that classifies by weapon type will attribute the whole row to whichever type it matches. Type-specific rates drawn from small samples should be read with this in mind, particularly where launch volumes are low.

Novelty bias

Unusual events are over-reported when a capability is new and under-reported once it becomes routine. The first ground robot casualty evacuation was widely covered; the ten-thousandth was not. Apparent declines in novel capabilities frequently reflect the fading of news value rather than the fading of the capability.

Confirmation as a function of drone coverage

Drone footage is now the primary confirmation mechanism for battlefield claims on both sides. This creates a circularity: strikes conducted without drone observation are less likely to be confirmed and therefore less likely to be counted at all. High drone attribution in casualty data partly reflects what can be filmed, not only what drones did.

Primary sources behind the datasets on this site, with what each is good for and where it stops.

• Ukrainian Air Force daily reports, compiled by Petro Ivaniuk. Attack-wave level records of weapons launched, destroyed and EW-suppressed, from September 2022. The backbone of the air defence tab.
• USF Pidrakhuyka — Unmanned Systems Forces strike reporting, verified in the DELTA combat system. Category-level detail on targets hit and destroyed. Covers USF-branch units only, not all Ukrainian unmanned units, which is why its personnel figures sit below General Staff totals.
• Oryx and WarSpotting (Jakub Janovsky) — visually confirmed equipment losses. The strictest standard available and correspondingly the most conservative.
• Mediazona / BBC News Russian — Russian military deaths confirmed by name through obituaries, social media and official records. A hard floor; the real total is substantially higher.
• DeepState, ISW and Black Bird Group — territorial assessment. Read net change rather than gross gains, and expect revision.
• Ukrainian MoD — UGV mission counts from DELTA, published monthly.
• Ukrainian General Staff — daily aggregate loss reporting. Broad in scope, unaudited, and the standard denominator for attribution ratios.

Everything on the tracker is free to read, and every figure carries its source. All charts, tables and filters are open to anyone, with no account and no paywall. That is not changing.

Bulk exports of the underlying datasets are available on request. They used to download directly from each tab. They now come by email instead, for two reasons: the licence attached to this data is routinely misread, and knowing who is using the datasets makes it possible to tell people when figures are corrected — which, as the sections above make clear, happens.

What is available

• Air defence — attack-wave level records from September 2022: date, time window, weapon type, launched, kinetically destroyed, EW-suppressed.
• USV strike log — curated operations with target, platform, location, damage assessment and source citation.
• UGV operations log — as above, both sides, plus the monthly DELTA mission series.
• UAV kill board — USF Pidrakhuyka annual and monthly figures, headline and category level.
• Territorial change — net monthly series with sources.

How to request

Email with your name, your institution or publication, which datasets you need, and what you intend to do with them. A couple of sentences is enough. Requests from researchers, journalists, students and analysts are granted as a matter of course, normally within a few days.

[email protected]

Requests to reproduce charts, cite figures, or quote the tracker in published work do not need any of this. Cite it and go ahead.

Terms as used on this site and in the sources it draws on. Usage in the field is not always consistent; where a term is contested, that is noted.