Detection is the easy part

The chain runs detect, intercept, identify, process, charge, evidence, court, disposal. Money enters it at the first link. Most of the delay sits in the last four.

This is not an argument against sensors. It is an argument about sequence and proportion, and it is made easier by the fact that several governments have now published the numbers.

What detection costs, and what it buys

The best-funded maritime detection system in the world sits in the central Mediterranean. Independent researchers put Frontex spending on aerial surveillance there at more than half a billion euros since 2017, including leased long-endurance drones operating from Malta. Frontex declines to publish flight-hour costs as commercially sensitive; researchers working from the contract value have calculated a figure in the region of sixteen thousand euros per flight hour.

It works. Boats are found. What happens next is where the argument sits. Human Rights Watch and independent researchers have documented that a substantial share of the resulting interceptions were carried out by the Libyan coastguard, returning people to a state with no functioning asylum system. Arrivals fell. The chain did not get faster; the far end of it was moved to somebody else’s jurisdiction.

The measure you choose determines what the capability becomes. If the metric is arrivals, detection technology will find a way to reduce arrivals, and it may not be the way you intended.

Four places the chain actually breaks

The published record is unusually clear on this, and it comes from small jurisdictions rather than from theory.

• Detection itself, where it is genuinely poor. Aruban authorities have been quoted estimating that they identify three to five boats a year while believing that for every boat identified at least ten more are not. That is a real detection gap and the stated cause is institutional resourcing rather than a missing sensor.
• Interception and response, where detection outruns the ability to act. Latvia completed a one hundred and seventy-three kilometre border fence, and by August 2026 had stopped close to ten thousand crossing attempts in eight months against roughly twelve thousand across the whole of 2025. Migrants tunnelled beneath it. The country’s border chief attributed the binding constraint to a shortfall of around two hundred guards.
• Processing, where the queue grows even as arrivals fall. Cyprus ended 2025 with about twenty-eight and a half thousand pending asylum applications, up eleven per cent year on year, and an average processing time of twenty-seven months, in a period when irregular arrivals were falling. Fewer people arriving did not shorten the queue, because the queue was never a function of arrivals.
• Disposal, where there is nowhere for the outcome to go. Trinidad and Tobago operates a single immigration detention facility with a capacity of around one hundred and fifty, historically overcrowded to the point of improvised overflow accommodation, with no statutory maximum period of detention.

What the auditors found

Two independent public audit bodies have examined two entirely separate border systems on two continents, and reached compatible conclusions.

The European Court of Auditors reported in 2021 that Frontex had not been sufficiently effective in helping member states manage the external border, citing weaknesses in information exchange, incomplete risk-analysis data, and a failure to disclose the real cost of joint operations.

The United States Government Accountability Office reported in January 2026 that the Coast Guard had missed its migrant-interdiction targets in six of the previous ten financial years and that the interagency task force had not fully shared lessons learned. Its recommendations were about performance measurement and information sharing.

Neither audit recommended more sensors. Both recommended better measurement and better handoffs. That is a striking convergence from two bodies with no connection to each other, and it is the strongest evidence available that the constraint in mature systems is organisational rather than technical.

What data analysis is actually for

The common assumption is that analytics will find more. In practice its highest value is elsewhere: it links what has already been found.

A detection is a point. A case is a network. The work that converts one into the other is matching a vessel to a previous sighting, a phone to a route, a facilitator to a payment, a person to a prior encounter under a different name, and an exhibit to the officer who seized it. Every one of those is a data problem, none of them require a new sensor, and all of them shorten the distance between arrest and charge.

This is also where evidential discipline pays. Continuity of exhibits, tasking records, retention schedules and disclosure practice designed in at the start rather than reconstructed for the first contested trial. A prosecution that collapses on disclosure has consumed the entire chain and produced nothing, and no amount of upstream detection recovers it.

The measure that matters

If an institution measures only detections and interceptions, it will optimise for detections and interceptions. Two better measures are available and neither is expensive to collect.

• Time from detection to disposal, whether disposal means charge, removal, release or transfer. Measured as a median and as a worst decile, because the tail is where the political risk lives.
• Attrition at each handoff. Of a hundred detections, how many become interceptions, how many become identifications, how many become charges, how many reach court, how many conclude. The largest single drop in that sequence is the place to spend, and it is very rarely the first step.

Both can be assembled from records an organisation already holds. Producing them for the first time is usually uncomfortable, which is a reason to do it rather than a reason not to.

The layer that is cheap and largely solved

Small jurisdictions do not need to build detection sovereignly and mostly should not try. The functioning model is subscription to shared regional services.

• In the Caribbean, CARICOM IMPACS operates a joint regional communications centre screening on the order of forty million passenger movements a year through advance passenger and cargo information systems. A single state could not build that.
• In Europe, EMSA provides remotely piloted aircraft on request to member state authorities, and Frontex shares a fused surveillance picture rather than funding twenty-seven national systems.
• Border management information systems supplied through IOM cost less than most people expect. A 2025 handover in Belize covered servers, workstations and training for hardware valued at around two hundred and eighteen thousand Belize dollars, roughly a hundred and ten thousand United States dollars.
• Interpol’s stolen and lost travel document database is available to any member state and is among the cheapest genuine capability uplifts available at a border.

Set against a single leased long-endurance drone, the entire list above is a rounding error.

What we would establish first

Before any sensor is specified, three things, none of which cost anything beyond the effort of assembling them.

• The attrition curve from detection to disposal, with the largest drop identified.
• The median and worst-decile time through the chain, and which handoff consumes it.
• Whether the authorisation and evidential practice at the front end will survive contest at the back end, because if it will not, everything upstream of the courtroom is wasted effort.

If the answer to any of those is unknown, that is the finding, and it is worth more than a procurement.

Our interest, and one honest caution

We advise on sensors, communications architecture, unmanned systems and data, and we can help specify and source them. An argument that detection is the easy part is not an argument against our own work; it is an argument about where in our own work the value sits. Read it accordingly.

The caution is this. One of the clearest reductions in irregular arrivals recorded anywhere recently was in the Bahamas, where apprehensions fell by around seventy-eight per cent between 2024 and 2025. The cause publicly credited was a change in destination-country immigration policy, not any detection investment. Flow responds to law and to opportunity at least as much as it responds to capability, and any adviser who tells you otherwise is selling something.

Read next · Procurement

The price of a gift

Donated capability is rarely free. The question is not whether to accept it, but who wrote the requirement it answers.