SIPRI Arms Transfers Database: Intelligence Source Guide

SIPRI records every international transfer of major conventional weapons since 1950, deal by deal, in a common unit of military capability rather than money. It is the baseline against which every claim about who arms whom is tested, and its central measure is the thing analysts most often misread.

At a glance

Source	SIPRI Arms Transfers Database
Category	Conflict, Crime & Human Security › Military, Weapons & CBRN
Homepage	https://www.sipri.org/databases/armstransfers
Machine interface	`https://armstransfers.sipri.org/ArmsTransfer/`
Format	`HTML`
Access	Open — no account required
Disciplines	Weapons Intelligence, Economic Intelligence
Mission domains	Weapons Trafficking, Military & Defense

Global arms transfer records. — as catalogued in the platform’s own source registry.

The SIPRI Arms Transfers Database is a hand-curated register of international transfers of major conventional arms, maintained by the Stockholm International Peace Research Institute and covering the period from 1950 to the most recent completed calendar year. The unit of record is a deal: a supplier, a recipient, a weapon designation and description, the number ordered, the number delivered, the year the order was placed and the years over which deliveries occurred, plus a free-text comment field that frequently carries the most important information on the row. Weapon categories are restricted to major conventional systems – aircraft, air defence systems, anti-submarine warfare weapons, armoured vehicles, artillery, engines, missiles, sensors, satellites, ships, and a residual category covering items such as turrets, gun systems and air refuelling equipment. Each delivered item is assigned a trend-indicator value, SIPRI's own unit representing the military resources embodied in the weapon, derived from known production costs of comparable systems. The database is queried through a web interface with filters on supplier, recipient, year range, weapon category and status, and results can be exported. There is no public programmatic API, and the underlying research is manual: SIPRI staff read national reports, company statements, parliamentary records, specialist press and official registers and adjudicate what counts.

The job this database does that nothing else does is make arms transfers comparable across countries, decades and currencies. Financial trade statistics measure what was paid, which is distorted by offsets, grants, aid packages, second-hand pricing, political discounts and the sheer opacity of defence contracting; SIPRI's trend-indicator value instead measures what was delivered in terms of military capability, using a fixed valuation for a given system type regardless of what changed hands financially. That is what makes a fifty-year time series meaningful and what makes a comparison between a grant transfer and a commercial sale legitimate. For WEAPINT and ECONINT work the database is the structural backbone: it establishes normal supply relationships, which is the only way to recognise an abnormal one. Sanctions and embargo analysis in particular depends on it, because the question is almost never whether a transfer occurred but whether it fits the historical pattern of that supplier-recipient pair, and answering that requires a consistent multi-decade baseline. It also functions as an industrial-base map, linking weapon designations to producing states and, through SIPRI's companion arms-industry work, to the companies behind them.

Who publishes it, and why that matters

SIPRI is an independent international institute established in Sweden, funded principally by a Swedish government grant with additional project funding from other governments, foundations and international organisations. That structure has two consequences worth stating. It has produced remarkable institutional continuity – this database has been maintained for decades and the annual publication rhythm has held through changes of leadership and funding environment – and it makes SIPRI a European-anchored institution whose research priorities and access to sources reflect that position, something to note rather than to hold against it. The output is free at the point of use, which is deliberate policy rather than a business model, and the institute's incentive is scholarly reputation. The practical implication for reliability is good: SIPRI is conservative, publishes its sources and methods openly, revises past entries when better information emerges, and does not have a commercial reason to inflate anything. The practical implication for longevity is that the resource depends on sustained public funding for a research institute, which is a more stable arrangement than most open sources enjoy but not an unconditional one.

Provenance is the first question to ask of any dataset and the one most often skipped. Who collects it, what their incentive is, whether they publish a methodology, and whether they correct the record when they get something wrong all bear directly on how much weight a finding drawn from it can carry.

What a record actually contains

The fields you will be working with, what each one means, and whether it is something you can pivot on. Read the meanings carefully — more analysis is wrecked by misreading a field than by failing to find one, and a field that looks like an observation is often an inference.

Field	Type	What it means	Pivot value
`supplier`	string	The state, or in a limited set of cases the non-state entity, from whose territory or under whose authority the transfer originated. For licensed production and re-exports the attribution logic is set out in the sources-and-methods documentation and is not always the intuitive one.	National arms export reports, the supplier's declarations to international registers, and the producing companies behind the designation.
`recipient`	string	The receiving state or entity. This is the party that took delivery, which is not necessarily the operating force and not necessarily the final end user in transhipment and re-export chains.	Country dashboards, embargo status, and the recipient's own reporting to international transparency instruments.
`weapon_designation`	string	The specific system as designated by its producer or operator. This is the join key to almost everything else you will want to do, and it is written in producer conventions rather than a controlled vocabulary.	Manufacturer and country of origin, technical references, and observed field use in conflict documentation.
`weapon_description`	string	The category and functional description of the system – fighter aircraft, self-propelled gun, surface-to-air missile system, naval gun, fire-control radar and so on. Coarser than the designation and more reliable for aggregation.	Category-level aggregation across suppliers, which is how capability-class trends are actually measured.
`number_ordered`	int	The quantity in the deal as SIPRI understands it, frequently an estimate and frequently revised. An order is a contractual intention, not a delivery, and a substantial share of ordered quantities never arrive.	The delivery figures for the same deal, and the comment field, which usually explains any gap.
`number_delivered`	int	The quantity actually transferred, distributed across delivery years. This is the field that drives the trend-indicator values and the one that should anchor any analysis of what a force actually has.	Order-of-battle and inventory references, and visual confirmation of the type in service.
`year_of_order`	int	The year the deal was agreed, as best established. Contract signature, licence approval and political agreement can be years apart and the database records one date, so treat this as approximate for policy-timing arguments.	Export licence records and parliamentary approval timelines in the supplier state, where those are published.
`years_of_delivery`	array	The span over which deliveries occurred, with per-year quantities. Major platform deals routinely deliver across five to fifteen years, which is why annual transfer figures are lagging indicators of political decisions.	Time-series construction; alignment with conflict onset and with military expenditure series.
`status`	enum	Whether the deal is a delivered transfer, an outstanding order or a cancelled or suspended arrangement. Filtering on this incorrectly is the most common way analysts produce inflated figures.	Outstanding-order backlogs, which are the closest thing to a forward-looking indicator the database offers.
`tiv_unit`	int	The trend-indicator value assigned to one unit of the system. It is a measure of military resources, benchmarked on known production costs of comparable equipment, and it is not a price, not a contract value and not denominated in any currency.	None financial. Use it only for comparison within the SIPRI system; joining it to monetary trade data is invalid.
`tiv_delivered`	int	The trend-indicator value of the items delivered in a given year for a given deal. Summing this across deals is how SIPRI's headline supplier and recipient rankings are produced.	Five-year aggregate comparisons, which SIPRI itself uses because single-year figures are dominated by delivery lumpiness.
`comments`	string	Free text carrying second-hand status, licensed production arrangements, aid and grant characteristics, uncertainty about quantities, refurbishment, and the identity of intermediaries. Analysts who filter this column out lose the qualifications that make the row interpretable.	Everything. This is where transhipment countries, brokering arrangements and end-user ambiguities are recorded.
`licensed_production_flag`	enum	Whether the transfer involves production under licence in the recipient country rather than delivery of finished units, recorded in the deal description and comments. It changes the meaning of both quantity and timing completely.	Industrial base analysis and technology transfer questions, which are distinct from equipment counts.

Coverage — and what is not in it

The database covers international transfers of major conventional weapons from 1950 to the most recently completed year, and updates once annually, conventionally in March, accompanied by a fact sheet on trends in international arms transfers. Coverage is global in ambition and even in practice for the large industrial suppliers, whose exports are relatively well documented through national reports, parliamentary scrutiny and specialist press. It is thinner for suppliers with limited transparency and for recipients whose acquisitions are politically sensitive. Both state and, in a limited set of documented cases, non-state recipients appear. What it does not cover is at least as important as what it does: small arms and light weapons, most ammunition, most components and subsystems, dual-use goods, services such as training and maintenance, and technology transfer that does not attach to a physical system are all outside scope by design. Second-hand transfers, refurbished equipment, aid and grant deliveries and licensed production are all inside scope but valued and coded by specific rules that materially affect what the numbers mean. Because the data is revised retroactively as better information emerges, the entire historical series can change between editions, so the version and access date you used is part of your finding rather than an administrative detail.

Known blind spots

Absence of evidence here is not evidence of absence. These are the conditions under which SIPRI Arms Transfers Database will not show you something that is nevertheless real:

Small arms, light weapons and ammunition are excluded almost entirely, so the database is silent on the category of weapon that does most of the killing in most contemporary conflicts and on the flows that dominate illicit trafficking.
Illicit and covert transfers appear only where they have become documented enough for SIPRI to record them, which typically means years after the fact if ever, so recent covert supply relationships are structurally invisible.
Components, subsystems and dual-use items fall outside scope, which means a supplier can be materially critical to a weapon programme through microelectronics, optics, machine tools or propellants and not appear as a supplier at all.
Services, maintenance, overhaul, training and sustainment contracts are not counted, and for many recipients these are worth more and matter more operationally than the platform deliveries that are counted.
The trend-indicator value measures capability delivered, not money, so it cannot answer questions about revenue, market share, industrial employment or affordability, and any analysis that needs those must use financial data with all its distortions.
Annual figures are dominated by delivery lumpiness. A single large ship or aircraft batch arriving can double a recipient's annual figure with no change in policy, which is why SIPRI presents five-year aggregates and why single-year comparisons are usually noise.
The recipient is the receiving state, not necessarily the end user. Transhipment, re-export, third-party transfers and supply to allied forces operating on a recipient's territory are handled in comments where known and are invisible where not.
Retroactive revision means historical figures are not stable. A conclusion you drew and published from one edition may not be reproducible from the next, and the database offers no versioned public archive of superseded editions.
The order year records one date for a process that spans political agreement, contract signature and export licensing, so it should not be used to make fine-grained arguments about when a government decided something.

Write the blind spot into the product. A statement that something “was not observed in SIPRI Arms Transfers Database” is defensible; a statement that it “did not happen” is not, and the difference is what survives cross-examination.

Access, licensing and what you may do with it

Access model: Open — no account required

Access is free and requires no account for basic querying. In practice you use the web application, which lets you build a query by supplier, recipient, year range, weapon category and deal status, and returns a results table you can export. There is no public REST API and no bulk dump of the whole database, so systematic collection means driving the query interface programmatically or, more sensibly, pulling a small number of broad exports once per annual release cycle and doing your filtering locally. Because the update is annual, this is one of the few sources where a once-a-year collection job is not merely adequate but correct. SIPRI publishes its sources and methods openly and the documentation should be read before the first query rather than after the first surprising result, because most of the questions analysts raise about apparently anomalous rows are answered there. Register the exact query parameters and the access date with every export you keep.

Licence

SIPRI makes the database freely available for research, education and non-commercial use with attribution, and requires permission for reproduction and for commercial exploitation. The precise terms are published on the SIPRI site and have been adjusted over time, so confirm the current statement before building anything commercial on top of it rather than relying on a summary in a guide like this one. Citation practice matters more here than with most sources: because the data is revised retroactively, a citation that does not include the database name and the date of access is not reproducible. Extracting figures for internal analysis is uncontroversial; republishing substantial extracts, embedding the dataset in a commercial product, or presenting derived rankings as your own work is not, and SIPRI is a small institute whose funding case rests partly on visible use of its outputs, so accurate attribution is also a matter of not degrading the commons you are drawing on.

Rate limits and fair use

There are no published limits and no reason to approach any. The data changes once a year. A collection design that hits the query interface repeatedly is both antisocial and pointless: pull broad exports at the annual release, cache them, and serve all internal queries from your own copy. If you must drive the interface programmatically, serialise requests, identify yourself in the user agent, and schedule the work outside European business hours. Treat the annual fact sheet publication as your collection trigger rather than polling for change.

Licensing changes, and it changes without warning. A dataset that was free for research this year may not be free for commercial or evidential use next year. Confirm the current terms before you build a dependency on it, and record the terms you relied on alongside the data — the licence in force at the time of collection is part of the provenance.

Collecting it

How SIPRI Arms Transfers Database is actually pulled, in the order you would set it up. Prefer the bulk or export interface over per-item lookups wherever one exists: it is kinder to the publisher, faster for you, and gives a reproducible snapshot rather than a series of point-in-time answers you cannot reconstruct later.

Method	Format	Cadence	Notes
Annual bulk export by supplier or recipient	CSV	once per annual release, conventionally in March	The correct primary method. Pull broad slices rather than narrow queries so that you hold a local copy you can re-filter without going back to the interface.
Query interface for targeted questions	HTML	ad hoc	Right for a one-off analytical question, wrong as a collection mechanism. Record the exact filter parameters alongside any figure you take from it.
Fact sheet ingestion	HTML	annual	The accompanying trends fact sheet states what changed in the edition and what SIPRI's own analysts consider significant, which is the fastest route to knowing whether your baseline has shifted.
Companion database pulls	HTML	annual	The arms industry and arms embargo databases are separate resources on the same site and answer adjacent questions – who produces, and who is prohibited from receiving – that transfer data alone cannot.
Edition archival	bulk	every release	Keep each year's export permanently. Since the historical series is revised in place upstream, your own archive is the only way to reconstruct what the data said when you published a finding.

Ingesting it into the platform

Every step below is idempotent and cursor-based: interrupt one and it resumes from where it stopped rather than duplicating rows or losing progress. Collection is recorded per source, so a feed that quietly stops publishing shows up as a stale timestamp instead of silently thinning your coverage.

Register as an annual-cadence source — Add the database in sources.php with an annual collection rhythm and a note that the entire historical series is mutable between editions, so that downstream consumers understand why last year's figures may not match.
Schedule the release-window pull — Configure collect.php to attempt collection in the release window and let cron.php own the retry logic, since a single annual job that fails silently costs you a year.
Normalise entities on import — Use import.php to map supplier and recipient names to canonical country entities and to resolve historical states and name changes across a seventy-year series, which is a real and underestimated data-engineering problem in this dataset.
Preserve the comment field verbatim — Carry the free-text comments through ingest untouched and index them for search. They contain the second-hand, licensed-production and intermediary qualifications that make individual rows interpretable, and stripping them produces a clean dataset that quietly lies.
Resolve the industrial layer — Push weapon designations through enrich.php to attach producing companies and countries of origin where they are known from other registered sources, producing the company-level view that supports supply chain and sanctions work.
Correlate with embargo and sanctions state — Use correlate.php to overlay transfer records with sanctions and embargo timelines so that transfers occurring inside a restricted period surface for review rather than sitting undetected in a table.
Build the country baselines — Populate country.php and country-risk.php with supplier-recipient relationship histories so that any new reporting about a transfer can be evaluated against decades of pattern rather than against an analyst's memory.
Version every derived figure — Stamp the edition and access date onto any metric exposed through analytics.php or exported via export.php, because a SIPRI-derived number without an edition is not a reproducible finding.

Registered sources and their last-collected state are listed in sources.php, and the scheduled chain that keeps them current is in automation.php.

How it is wrong, and how to tell

Every dataset is wrong in characteristic ways. Knowing which ways is the difference between using a source and being used by one, and it is the part of source evaluation most often skipped because it is the part that takes work.

This is among the highest-quality open datasets in the security field, and the basis for that judgement is transparency of method rather than any claim of completeness. SIPRI publishes what it counts, how it values it, what it excludes and why, and it flags estimates and uncertainty in the data rather than presenting everything with equal confidence. The research is manual and adjudicated, which means it is slower than automated collection and considerably more accurate about what it does cover. The known weaknesses are structural rather than sloppy: scope excludes whole categories of trade, opaque suppliers produce thinner records than transparent ones, and the retroactive revision policy that keeps the series accurate also makes it unstable for citation. The trend-indicator value is a defensible construct that is widely misused by people who did not read the methodology, which is a quality problem in the ecosystem rather than in the data. Confidence should be high for the existence and rough scale of transfers between well-documented parties, moderate for quantities and timing, and explicitly low for anything involving covert supply, non-state recipients or the most recent two or three years, where reporting has not yet matured.

Characteristic false positives

Reading the trend-indicator value as money. It is not dollars, not contract value and not comparable to customs or trade statistics, and every published claim that a country exported a given sum of arms based on TIV figures is simply an error.
Counting orders as deliveries. Outstanding orders include deals that will be reduced, delayed for a decade or cancelled outright, and the status filter is the difference between a description of reality and a description of intentions.
Reading a single year as a trend. Delivery schedules are lumpy by nature and one large platform batch dominates a recipient's annual total, which is precisely why SIPRI's own analysis uses five-year aggregates.
Treating absence as evidence of no transfer. The scope exclusions mean a supplier can be central to a recipient's military capability through small arms, components, ammunition or sustainment and appear nowhere in the database.
Confusing recipient with end user. The receiving state is recorded; onward transfer, use by allied forces on its territory, and re-export are only visible when someone documented them and SIPRI recorded it in comments.
Comparing figures across editions. Retroactive revision means the same historical year can carry different values in successive releases, and an apparent change over time may be a change in SIPRI's information rather than in the world.
Double counting licensed production. A licensed-production arrangement and the deliveries of components or complete units associated with it can be read as separate transfers by an analyst skimming designations without reading comments.
Attributing supplier identity through an intermediary. Deals routed through brokers, transhipment states or third-party governments are attributed according to SIPRI's stated rules, which may not match the attribution your analysis needs, and the comments are where the difference is disclosed.

None of these make the source unusable. They make it a source that requires corroboration before an assertion built on it goes into a product, which is true of every source and admitted by few.

Ageing

The dataset ages on an annual clock and in two directions at once. Forwards, the most recent covered year is always the weakest, because reporting on transfers matures over several years as national reports appear, deliveries become visible and deals are confirmed, so recent-year figures are systematically revised upward and should be treated as provisional. Backwards, older entries become more complete and more stable but their analytical currency decays for capability questions, since equipment delivered thirty years ago may be retired, cannibalised, destroyed or re-exported without any of that appearing here. Between annual releases the whole dataset is frozen, which makes it useless for anything time-sensitive: if a transfer was reported in the press last month it will not be in the database for up to a year and possibly longer. A stale record in practice looks like an outstanding order from a decade ago that was quietly abandoned, a recipient state that no longer exists under that name, or a figure you quoted from an earlier edition that no longer matches the current one.

What this source feeds

A source is only worth what it lets you conclude. These are the disciplines that collect through it, the mission domains it serves and the data points it yields — every one is a tag, so you can follow any thread from here into the rest of the library.

Collected by these intelligence disciplines

Serves these mission domains

Yields these data points

How each sector uses SIPRI Arms Transfers Database

The same dataset is worked very differently depending on who you are, what authority you hold, and what you are ultimately producing. A military analyst is supporting a commander’s decision; a journalist is meeting a publication standard; an NGO caseworker is protecting a person. The records are shared — the constraints, thresholds and outputs are not.

🎖 Military and defence

For capability planning and threat characterisation this is the standard reference for establishing what an opposing or partner force has actually received and from whom, over a horizon long enough to include everything still in service. The outstanding-order backlog is the closest thing available to a forward-looking indicator of capability change, and delivery-year distributions tell you when new systems will reach units rather than when they were announced. Use it to build the materiel baseline that loss and attrition reporting is measured against, and to identify sustainment dependencies, since a force flying a supplier's aircraft is dependent on that supplier for parts in ways the equipment count alone does not show.

🕵 National intelligence

This is a primary collection route for WEAPINT and ECONINT work on proliferation, embargo circumvention and supply relationships. Its analytical function is baseline: to say that a reported transfer is anomalous you must first be able to characterise what normal looks like for that pair over decades, and this is where that characterisation comes from. Combine it with the embargo timeline to surface transfers inside restricted periods, use the comment field to identify brokers and transhipment routes that appear repeatedly, and treat the absence of small arms, components and services from scope as a standing collection gap to be filled from elsewhere rather than as a finding.

👮 Law enforcement

For customs, export control and arms trafficking enforcement the database is context rather than evidence. It establishes the legitimate trade pattern against which an intercepted shipment or a suspicious licence application can be assessed, identifies the intermediaries and transhipment states that recur in documented deals, and supports the country-risk assessment that underpins licensing decisions. It will not tell you about the illicit flows that constitute most enforcement work, because those are out of scope by construction and because small arms are excluded. Route intelligence on suspected illicit transfers to the mandated national authority and to the relevant international mechanism rather than treating a database gap as a lead.

🔍 Private investigation and corporate security

Corporate investigators and defence sector due diligence teams use this to test claims. Whether a company's stated export history is consistent with the recorded transfers of the systems it makes, whether a claimed end user matches the recorded recipient, and whether an agent or intermediary appears in the documented record of deals in that market are all answerable questions with real commercial consequences. The comment field is where brokering and offset arrangements surface. Remember that the dataset is annual and lagging, so it is a poor tool for anything about a transaction in progress and a good one for establishing a counterparty's historical footprint.

📰 Journalism and OSINT media

For arms trade reporting this is the standard citation and the standard trap. The trend-indicator value is not a monetary figure and reporting it as one is the most frequent error in published arms trade coverage; the second most frequent is comparing single years. Cite the database by name with the access date, use five-year aggregates for trend claims, and read the comment field on any row you intend to build a story around, because it usually contains the qualification that changes the story. SIPRI's annual fact sheet is written for exactly this audience and is a better starting point than the raw query interface.

🌍 NGO, humanitarian and human rights

For arms control and human rights advocacy this is the evidentiary foundation for Arms Trade Treaty compliance work, embargo monitoring and risk-of-diversion arguments. It lets an organisation demonstrate that a supplier maintained a transfer relationship with a recipient during a period of documented violations, which is a factual claim that survives contestation in a way that assertions do not. Pair it with the arms embargo database for the legal overlay and with field-level weapons documentation for the evidence that specific items reached specific users, and be explicit that the exclusion of small arms means the database understates the flows most relevant to civilian harm.

🎓 University and research

This is one of the most heavily used datasets in quantitative international relations and security studies, and the norms of good practice around it are well established: cite the edition, archive the extract, use TIV as an ordinal capability measure rather than a monetary one, and address the scope exclusions explicitly in the limitations section. The retroactive revision policy is a genuine reproducibility hazard and is the main reason published results diverge between studies nominally using the same data. Researchers building new indices on top of it should read the sources and methods documentation closely enough to know what the valuation rules do to second-hand and licensed-production deals, because those rules drive a large share of the variance in aggregate figures.

Playbook: working SIPRI Arms Transfers Database end to end

A repeatable sequence from first pull to finished product. Each phase states what you are trying to establish, not merely what to click — the objective is a defensible chain of reasoning, not a completed checklist.

Phase 1 — Read the methodology before the data

The sources and methods documentation defines what a transfer is, what the trend-indicator value measures, and which categories are excluded. Nearly every apparent anomaly an analyst finds in this dataset is explained there. Reading it first converts two days of confused investigation into ten minutes of orientation, and it is the difference between using the data and misusing it confidently.

Phase 2 — Fix the question to the right unit

Decide whether you need capability delivered, money spent, or items present. Only the first is what this database measures. If the question is financial, you need trade statistics or national export reports and you should accept their distortions; if it is about what is in service now, you need inventory sources. Choosing the wrong unit at this stage produces an answer that is internally consistent and irrelevant.

Phase 3 — Pull broad and filter locally

Export the full supplier or recipient history rather than a narrow slice matching your current hypothesis. Local re-filtering costs nothing and lets you test alternatives without going back to the interface; more importantly, it forces you to see the deals your hypothesis would have excluded.

Phase 4 — Establish the baseline relationship

Before evaluating any specific transfer, characterise the supplier-recipient pair over the longest available period: volume, categories, continuity, and the political events that interrupted it. This baseline is the analytical product; the individual transfer is only interpretable against it. Write the baseline down, because you will reuse it and because it is the part a reviewer will challenge.

Phase 5 — Separate orders from deliveries explicitly

Build two views of every relationship, one of what was agreed and one of what arrived, and look at the gap. Persistent large gaps indicate financing problems, political friction, industrial capacity limits or quiet cancellation, and each of those is a finding in itself that the combined figure conceals.

Phase 6 — Read every comment on rows that matter

For any deal your conclusion rests on, read the free text. It carries second-hand status, refurbishment, aid characteristics, intermediaries, uncertainty about quantities and the identity of parties other than the recorded recipient. Analysts who work only from the numeric columns produce clean, wrong answers.

Phase 7 — Overlay the legal restrictions

Bring in the embargo and sanctions timeline for both parties and mark transfers that occur inside restricted periods. Some will be legitimate under exemptions and grandfathering, which is exactly why the overlay must be examined by an analyst rather than turned into an alert that fires on coincidence.

Phase 8 — Fill the scope gaps deliberately

List the categories your question needs that this database excludes – small arms, ammunition, components, services, technology – and assign each to another source before you draw conclusions. The gap list should appear in your product, because a reader who does not know about the exclusions will read absence as evidence.

Phase 9 — Cross-check against independent reporting

Test your emerging picture against national export reports, declarations to international registers, customs trade data and field weapons documentation. Disagreement between SIPRI and a national report is common, usually explicable by definitional differences, and always worth understanding before you publish either figure.

Phase 10 — Aggregate over five years, not one

Construct trend claims from multi-year aggregates. Delivery lumpiness makes annual comparisons unstable for all but the largest flows, and a rise or fall that reverses next year is a scheduling artefact rather than a policy change. Say which window you used and why.

Phase 11 — Version and archive the extract

Store the exact export with the edition and access date, and treat that archive as part of your finished product. Retroactive revision guarantees that a future reader querying the live database will get different numbers, and only your archive can explain why.

Phase 12 — State the unit in the product

Whenever a trend-indicator figure appears in a finished assessment, say in the same sentence that it is a measure of military capability rather than value. This single sentence prevents the most common downstream misreading and costs you nothing.

The platform ships this as a step-checked workflow in playbooks.php, so progress is recorded against a case rather than held in someone’s head.

What to pair it with

No single source carries a finding. These are the datasets that corroborate, extend or contradict this one — and a source that contradicts is worth more than one that agrees, because it is the only thing that will tell you when you are wrong.

Source	Relationship	What it adds
SIPRI Military Expenditure Database	extends	The demand side. Transfers tell you what a country received; expenditure tells you what it could afford and how its priorities shifted, and the two series read together are far more informative than either alone.
SIPRI Arms Embargoes Database	prerequisite	The legal overlay. Without the embargo timeline, transfers cannot be assessed for compliance, and the same transfer can be routine or a violation depending on dates the transfer record does not carry.
SIPRI Arms Industry Database	extends	Links the systems being transferred to the companies producing them, which is the bridge from state-level transfer analysis to corporate due diligence and supply chain work.
UN Register of Conventional Arms	corroborates	States' own voluntary declarations of imports and exports in seven major categories. Divergence between a state's declaration and the SIPRI record is itself analytically interesting.
UN Comtrade	contradicts	Monetary trade statistics including arms categories. Structurally disagrees with SIPRI because it measures a different thing in different units, and the disagreement is informative rather than a problem to be reconciled.
Arms Trade Treaty Secretariat	extends	The treaty framework, state party reporting and the obligations that make a transfer lawful or not. Essential context for any compliance argument built on transfer records.
Small Arms Survey	extends	Covers the entire category SIPRI excludes. If your question involves the weapons most used in contemporary conflict and crime, this is where the answer lives.
Conflict Armament Research iTrace	corroborates	Physical documentation of weapons recovered in the field, which can confirm that a recorded transfer actually reached a theatre and can reveal diversion the transfer record cannot show.
Wassenaar Arrangement	extends	The export control regime and control lists that govern much of the trade being recorded, and the vocabulary national licensing systems actually use.

Legal, ethical and operational constraints

The database itself is lawful to consult and to analyse; the constraints are on redistribution and on what you do with the conclusions. SIPRI's terms permit free non-commercial research and educational use with attribution and require permission for commercial exploitation and substantial reproduction, and those terms have been restated over the years, so verify the current wording before embedding the data in a product. The more consequential legal dimension is downstream: analysis of arms transfers touches export control law, sanctions and embargo regimes, and in some jurisdictions brokering and technical assistance offences that can be triggered by activity well short of moving a weapon. Analysts working on live transfer questions should understand that identifying a potential circumvention route is a matter for a competent authority rather than a private actor, and that in most jurisdictions there are affirmative reporting obligations attached to knowledge of certain sanctions breaches. Nothing in this database supports operational assistance to any party seeking to acquire controlled goods, and the library does not provide guidance in that direction. Where analysis feeds accountability or litigation work, the retroactive revision issue becomes an evidentiary one, and the archived extract with its access date is what makes a figure defensible.

Operational security

Querying is low-exposure but not anonymous. SIPRI operates ordinary web infrastructure and sees the address, user agent and query parameters of every request, which means the pattern of your queries discloses which supplier-recipient relationships you are examining and when your interest began. For most users this is unremarkable. For an analyst working a live sanctions circumvention question involving a specific pair of states, a burst of narrowly targeted queries from an attributable institutional range is a meaningful disclosure to anyone with visibility of those logs, including through lawful process in the hosting jurisdiction. The mitigation is straightforward and also better collection practice: pull broad extracts once per release and run your narrow queries against your own copy, so that your analytical questions never leave your infrastructure. Avoid registering for any optional account with an identifying institutional address if the interest itself is sensitive.

Two rules that hold regardless of jurisdiction. Collection that is lawful is not automatically proportionate, and a dataset assembled for one purpose does not carry consent for another. Where the records concern identifiable people, the question is not only whether you may hold the data but whether holding it serves the purpose you are accountable for.

Is it earning its place?

Sources accumulate. Feeds get added during an incident and are never reviewed again, and a decade later the pipeline is carrying dead weight that nobody dares remove. These are the measures that show whether SIPRI Arms Transfers Database is contributing anything, and they are worth baselining now so the answer is available later.

Annual collection success: whether the release-window pull completed and produced a plausible row count, since a missed annual job is a year-long blind spot rather than a gap.
Revision magnitude: how much the previous five years of figures changed between editions, which tells you empirically how provisional recent-year data is for your particular suppliers of interest.
Entity resolution completeness: the proportion of supplier and recipient names successfully mapped to canonical country entities, including historical states, which gates every join you will attempt.
Comment coverage: the share of rows in your area of interest carrying free-text qualifications, which is a proxy for how much of your dataset needs analyst reading rather than aggregation.
Order-to-delivery conversion: the realised fraction of ordered quantities by supplier and recipient, tracked over time, which is the practical measure of how much weight to put on outstanding orders.
Embargo overlap hits: how many transfers your correlation surfaces inside restricted periods and how many survive analyst review, which measures whether the overlay is producing signal or noise.
Citation hygiene: the proportion of internal products carrying the edition and access date, which is the difference between a reproducible finding and an anecdote.

Beware of volume. Indicator counts rise easily and say almost nothing. Unique contribution — findings this source produced that no other source in your stack would have — is the measure that matters, and it is usually far lower than anyone expects.

Tradecraft notes

The distinctions that separate a competent analyst from a fast one:

The trend-indicator value is a capability index, not a currency. Say this to yourself before every figure you publish, because the entire ecosystem around this dataset is contaminated with people who forgot.
The comment field is not metadata, it is the data. Second-hand status, aid characteristics, brokered arrangements and quantity uncertainty live there, and a row read without its comment is a row read wrong.
Orders are intentions and deliveries are facts, and the gap between them is one of the more revealing indicators in the whole dataset because it exposes financing, politics and industrial capacity at once.
Scope exclusions are the single largest source of wrong conclusions. Before writing that a state does not supply another, check whether the supply you are looking for would even be in scope.
Historical state names, dissolutions and successions across a seventy-year series are a genuine data problem, not a formatting nuisance, and a careless join across them will silently drop or duplicate entire relationships.
Five-year aggregates exist because single-year figures lie. If your finding disappears when you widen the window, it was a delivery schedule and not a trend.
Retroactive revision means the database is a living document and your citation is a snapshot. Archive every extract, and expect to defend the difference between your figure and the current one.
Licensed production changes the meaning of both quantity and timing, and treating a production licence as equivalent to a delivery of the same number of units misstates both capability timing and industrial dependency.
A gap in a long-standing supply relationship is often more informative than a large delivery, because relationships in this domain are sticky and their interruption usually has a political cause worth identifying.

Questions analysts actually ask

Is the trend-indicator value in dollars?

No. It is SIPRI's own unit representing the military resources embodied in a delivered weapon, benchmarked on known production costs of comparable systems. It is designed for comparison across time and countries and it deliberately does not track what anyone paid. Do not convert it, sum it with financial data, or describe it as value.

Why does the database disagree with my government's arms export report?

Because they measure different things. National reports typically count licence approvals or financial values under national category definitions; SIPRI counts deliveries of major conventional systems under its own definitions. The divergence is expected and usually explicable from the two methodologies rather than indicating an error in either.

Can I use this to detect sanctions evasion?

You can use it to establish the baseline that makes evasion recognisable, and to identify transfers occurring inside embargo periods. You cannot use it as a detection feed, because it updates annually, excludes the covert and illicit trade almost entirely, and lags reporting by years. Treat it as context for other collection.

Why did last year's figures change?

SIPRI revises the historical series as new information emerges, which is a feature. It also means figures are only reproducible against the edition you used, so archive your extracts and always cite an access date.

Does it cover small arms?

No, and this is the most consequential exclusion in the dataset. Small arms, light weapons and most ammunition are outside scope, so the database is silent on the weapons that dominate contemporary conflict casualties and illicit trafficking. Use dedicated small arms research for those questions.

Are non-state recipients included?

In a limited set of documented cases, yes, but coverage of transfers to non-state armed groups is necessarily much weaker than for states because such transfers are rarely reported through the channels SIPRI draws on. Absence of a group from the database says nothing about whether it was armed.

Is there an API?

There is no public programmatic API. The practical approach is broad exports at the annual release, stored locally and queried internally. Given the annual update cadence, this is not a limitation worth engineering around.

How current is the most recent year?

Provisional. Reporting on transfers matures over several years, so the latest covered year is systematically the least complete and is revised upward in subsequent editions. Treat recent-year figures as a floor and say so.

What does the platform add?

Scheduled annual collection with edition archival, entity resolution across seventy years of changing state names, correlation against embargo timelines and other registered sources, and versioned figures in exports. The analytical judgements remain SIPRI's; nothing in the pipeline generates a transfer record.

Standards, formats and interoperability

What this source speaks natively, and what it has to be translated into before a partner can consume it. Work that arrives in a recognised format is easier to defend, easier to hand over and easier to automate against:

SIPRI's weapon categories are its own controlled vocabulary and do not map one-to-one onto the seven categories of the UN Register of Conventional Arms, the Wassenaar munitions list, or national export control classifications, so any cross-walk needs an explicit mapping table.
The trend-indicator value is a SIPRI-internal unit with no external standard equivalent, and no valid conversion to any currency or to customs valuation exists.
Country identification should be normalised to ISO 3166 codes on ingest, with an explicit historical mapping for states that have dissolved, merged or been renamed during the covered period.
The UN Register of Conventional Arms provides the nearest thing to an internationally agreed reporting taxonomy for major conventional arms and is the sensible interchange vocabulary when reconciling sources.
Export control classification systems – national munitions lists and the Wassenaar control lists – are the vocabulary licensing authorities use, and translation into them is required for any work that touches enforcement.
In the platform, transfer records normalise to relationship-style structures between country and organisation entities and export as STIX 2.1, MISP, CSV, JSON and JSONL.
Arms Trade Treaty reporting templates define the categories state parties report against, which is the framework any compliance argument will ultimately be assessed in.

References

Primary documentation and authoritative references for this source. Publishers revise and retire material, so treat the retrieval date as part of the citation and re-check before relying on any of it in a formal product.

SIPRI Arms Transfers Database — Stockholm International Peace Research Institute. The database landing page, with the current statement of scope, access terms and release timing. The correct starting point and the correct citation target.
Arms Transfers Database sources and methods — SIPRI. The definitive explanation of what counts as a transfer, how the trend-indicator value is constructed and what is excluded. Read this before your first query, not after your first anomaly.
SIPRI Arms Transfers query interface — SIPRI. The working application where queries are built and results exported. Note the exact filter parameters with any figure you keep.
SIPRI databases — SIPRI. The index of all SIPRI databases. Worth knowing in full, because several questions analysts bring to the transfers database are better answered by one of the others.
SIPRI Arms Embargoes Database — SIPRI. Multilateral and unilateral embargoes with their scope and dates. The legal overlay without which transfer records cannot be assessed for compliance.
SIPRI Arms Industry Database — SIPRI. Arms-producing and military services companies by revenue. The bridge from state-level transfer analysis to corporate entities and supply chains.
SIPRI arms and military expenditure research — SIPRI. The programme page tying the transfer, expenditure and industry work together, and the route to the annual fact sheets that explain what changed in each edition.
UN Register of Conventional Arms — United Nations. States' own declarations of arms imports and exports. The primary independent cross-check, and a source of informative disagreement.
UN Comtrade — United Nations Statistics Division. Monetary trade statistics including arms-related categories. Measures a different thing in different units, which is exactly why it is useful alongside SIPRI.
Arms Trade Treaty — ATT Secretariat. The treaty text, state party reports and reporting templates. The legal framework any transfer compliance argument is ultimately assessed against.
Wassenaar Arrangement — Wassenaar Arrangement Secretariat. Export control lists and best practice guidelines for conventional arms and dual-use goods, and the vocabulary national licensing regimes are built on.
UN Office for Disarmament Affairs — United Nations. The institutional home of the conventional arms transparency instruments and the disarmament machinery that shapes what states report and to whom.

Link integrity: every reference above was verified with a live request when this page was generated. Where a publisher had moved or withdrawn a document, the link was repointed at a preserved copy in the Internet Archive and marked as archived. Anything with no reachable copy anywhere had its link removed rather than left to rot — the source is still credited, it simply cannot be linked.

Put it into practice

The Quantus Intel threat intelligence platform operationalises this source: it collects each annual edition on schedule, archives every extract with its access date so revisions are visible rather than silent, resolves seventy years of changing state names to stable country entities, and overlays embargo timelines so that transfers inside restricted periods surface for analyst review.. Browse the full source catalogue, or follow any tag above into the rest of the library.