Winning the Streets, Losing the War: The Haitian State’s Tactical Trap

An operation is judged by what it retakes. A campaign is judged by what is still held, prepared, and covered the next day. Six months of incursions into the capital’s strongholds, followed by a massacre in the center of Kenscoff, can be read through the same lens: holding without freezing the offensive, going in knowing where you are going, and never giving the adversary back the time or the space it needs to rebuild. It is on these three capabilities, and not on the number of operations, that the difference between winning streets and winning the war is decided.

There are two ways of counting a war against gangs. The first counts operations: strongholds penetrated, weapons seized, leaders neutralized, and, on the other side, leaders who escape, armored vehicles immobilized, neighborhoods where the shooting resumes. That is how the past six months have been told, in official statements and in criticism alike. The second counts something else: what each operation leaves behind, what it costs the next one, and what it teaches the adversary. This piece adopts the second approach, because the first cannot explain why, between December and August, the state was able to go into Bel-Air, Delmas 6, downtown, Martissant, and Village de Dieu, and still found itself facing 47 dead and more than fifty abducted in a municipality it said it was consolidating.

The facts are known and need not be told a second time. The police, its Task Force, the Armed Forces of Haiti (FAd’H), and, since April, the Gang Suppression Force (GSF) conducted a continuous offensive sequence from late December to June. Positions were retaken, a few of them held. Kenscoff, under attack since January 2025, then swept, patrolled, and declared consolidated, was hit in July and then struck in its center on 23 August. What is not known, or not discussed enough, is the mechanism that links these two sets of facts.

This piece therefore rereads the timeline with three questions in mind: who holds the ground once the breakthrough has been made; with what knowledge of the terrain and of the target the men set out; and how much time and space the adversary was given while the effort was concentrated elsewhere. These three questions point to three bottlenecks. Three foreign campaigns, in Colombia, Rio, and Mosul, show what happens when they are ignored. Five decisions, none of which costs an armored vehicle, can loosen them. And the conclusion, a few months before an election, no longer belongs to the forces but to the political leadership that employs them.

The day after matters more than the assault

An assault is judged by what it destroys, seizes, or retakes. A campaign is judged by what remains possible the next day. Is the ground still held? Have the men who led the assault been relieved? Can we strike elsewhere without weakening what has just been won? Do those setting out know where they are going, whom they are looking for, and where the neighborhood can close in on them?

The problem in the first half of 2026 is not a lack of action. It is that successful actions, taken one by one, do not turn into pressure that accumulates. That transformation depends less on equipment than on a system: personnel, relief, intelligence, preparation, command, the capacity to act in several places at once. Tempo, in this sense, is not the frequency of sorties. It is the speed at which a force can prepare, strike, hold, learn, and start again elsewhere before the adversary has rebuilt its means. A very active campaign can lose the initiative as soon as a single one of these steps becomes the bottleneck. The timeline of the past six months shows which ones.

Six months, read at the right scale

The timeline first shows a buildup. From late December, the Haitian National Police (PNH) and the Task Force, then the FAd’H and, from April, the GSF, intervened in several sectors of the Ouest department. Strongholds were penetrated, roads reopened, weapons seized, a few points consolidated. But the same timeline reveals something else: the nature of the effort demanded of the forces was gradually changing, and it is that change that matters.

Table 1 · Operational milestones in the Ouest department, December 2025 to June 2026. The sources for each row are listed in the note on method.

Two official figures frame the intensity. On 25 March, the PNH reported 32 large-scale operations in the first quarter, with 46 rifles, 19 pistols, and 6 drones seized. On 26 May, its spokesperson spoke of nearly 118 operations and patrols in the metropolitan region. The second figure does not mean four times as many offensives as the first; it adds up actions of unequal value, the presence patrol as well as the penetration of a stronghold. It measures activity, not what activity produces.

That is why the table should not be read like a scoreboard. As gains accumulate, three constraints tighten. The first is human: every point retained consumes a presence. The second is informational: you do not enter a contested space without a picture of the terrain, the target, and the risks, and that picture cannot be improvised. The third is geographic: concentrating your means here opens up time and space over there. Let us take them in order.

The tax on success

In December and January, the force posture was almost entirely geared toward maneuver. Units moved in quick succession through Bel-Air, Bercy, Delmas 4, Delmas 6, Delmas 2, and Marché Salomon, without yet having to bear the weight of positions that would need guarding.

From March onward, the equation changed. The gains had become numerous enough to create a second, permanent mission: preventing the armed groups from coming back in behind the advancing units. The authorities openly acknowledged this. On 25 March, Le Nouvelliste reported the planned deployment of 400 FAd’H soldiers, out of a total strength of about 2,000, to consolidate the positions recovered by the PNH; on 29 April, 200 of them were on the ground. In early June, the defense minister inspected the downtown sites placed under military guard, including Unitech, the Faculty of Human Sciences, IERAH, and INAGHEI, which the police had retaken in mid-April.

This consolidation is not fictitious, and that is precisely where the problem appears. Every position held has a daily cost: guard duty, patrols, resupply, relief, liaison, protection of the approaches. The reconquest draws on its own personnel to protect what it has just won. It is a tax on success. The wider the secured perimeter grows, the more the question becomes: how many forces remain available to keep going?

Village de Dieu brings all of this together. After two weeks of pressure in Bicentenaire, the security forces entered, on 1 June, a neighborhood where no unit had set foot in five years; six assault rifles were seized. The tactical feat is real. But a police source described the intervention as “essentially a reconnaissance mission,” and the PNH itself set fire to a heavy vehicle immobilized on site so that it would not fall into the hands of the gangs. From the very next day, the yardstick was no longer the same: who holds the approaches, who stays in contact, who gathers intelligence, and which units can still act elsewhere?

If the assault units stay behind to guard, they lose their mobility. If they leave without being relieved, the armed group takes back the space, street by street. The point is not to choose between offensive and consolidation, but to organize them as two distinct functions. The 21 June operation in Tabarre, conducted by the GSF, offers a glimpse of this: cordon, search, and clearance, then day and night patrols and the removal of barricades by bulldozer.

Yet Haiti does have rare skills. Nearly two hundred soldiers have been trained in Martinique in urban combat, marksmanship, and combat first aid; the PNH has its specialized units. The question, then, is not only how many men there are, but what they are used for. A unit that took months to train should not end up as a guard unit by default if another force can hold the ground behind it.

That is the first trap: achieving enough breakthroughs to have to tie down a growing share of one’s means, with no relief system to free up those who maneuver. The interviews conducted for this article reveal a second bottleneck, less visible and probably deeper. It lies before the operation.

Going in blind

Several police officers, including members of specialized units, describe training in which survival on contact logically takes center stage: advancing together, covering one another, reacting to fire. Their accounts converge on one point: other skills, just as necessary, remain far less systematic. Reading a map, visualizing the terrain and the routes, making use of an aerial image, preparing the objective on a plan, visually identifying the target: not all the operators deployed master these skills to the same degree.

The same accounts describe briefings that are insufficiently structured. What is sometimes missing is a shared projection of the terrain, a clear view of the routes, the likely obstacles, the passage points, and the exits. Officers recount situations in which the men sent to find a target knew the target’s face, or its immediate surroundings, only sketchily. These interviews are not an audit of the PNH. But their convergence raises a question that no count of armored vehicles can measure: with what shared picture does a unit go into an operation?

This question weighs all the more because the adversary, for its part, is watching. Several officers report enemy drones over parts of Bicentenaire; in Kenscoff, the attackers of 23 August had flown over the town for several days before attacking. United Nations reports document more broadly the use of commercial drones by armed groups, 5 Segonn in particular, to reconnoiter, conduct surveillance, track targets, and coordinate attacks. The PNH itself seized six drones in the first quarter.

That is where the asymmetry lies. A column can be observed before it has identified who is observing it; a movement reveals an intention; a poorly protected communication exposes an entire deployment. Knowing how to take cover remains indispensable, but it is no longer enough. Forces must be able to read and update a map, cross-check what comes from the ground against what comes from sensors, protect their communications, detect the other side’s drones and, within the appropriate legal framework, jam them. This requires technicians and engineers as much as fighters: telecommunications, geomatics, image analysis, target databases. The H-TAC project of the Organization of American States (OAS) is already strengthening the PNH’s tactical preparation; the next step is to make map reading, planning, and secure communications ordinary skills, present in every unit, rather than know-how that exists only when a specialist is on hand.

The same reasoning applies to the state’s drones. A drone is first and foremost a sensor. It is useful only if the image is read, cross-checked, turned into targeting, and passed on to a unit that understands the same situation. With a weak intelligence chain, a precision tool becomes an amplifier of error. And error costs twice. The United Nations Integrated Office in Haiti (BINUH) and the Office of the High Commissioner for Human Rights have documented civilian casualties during operations against gangs, including in strikes by explosive drones. Legal considerations aside, every targeting error damages human intelligence, feeds the adversary’s recruitment, and alienates the residents whose cooperation is needed to identify movements, faces, and networks. Precision is part of information superiority.

The second bottleneck can be summed up in one sentence: a force does not lose the initiative only when it lacks men. It also loses it when it sees later, understands less well, or prepares less thoroughly than the adversary. That leaves the third bottleneck, the one that catches up with even a well-prepared, well-relieved force.

The time given back, the space left open

A gang is not a stock of armed men. It is an organization that recruits, collects levies, arms itself, controls roads, and moves its revenue points. A pause gives it time to move its leaders, replenish its stocks, repair its positions, and observe the new habits of the security forces. Human intelligence follows the same logic in reverse: a lasting presence learns the streets, the flows, the faces; a withdrawal means rebuilding that knowledge for the next operation.

Time is only half the problem. On 8 May, BINUH counted at least 1,642 people killed in the first quarter and noted that, despite the advances downtown, violence was continuing to spread toward the Artibonite and Centre departments. Kenscoff, attacked in July and then struck in its center in August while the effort remained concentrated on downtown and Village de Dieu, is the costliest example. This does not mean that the same groups move mechanically from one area to another. It means that a numerically constrained force is constantly trading off concentration against coverage, and that the adversary picks the point where its coverage is thinnest. Two curves must therefore be tracked at once: the pressure in the area under attack and the vulnerability of the areas covered less. If several logistically linked strongholds are hit within the same window, the groups have fewer options for moving men, weapons, or vehicles. If operations become sequential and predictable, the adversary absorbs the blow here and prepares over there.

The three bottlenecks converge. Striking several areas without spreading oneself thin requires holding units that free up the assault units. Using them at the right moment requires shared preparation and intelligence. Avoiding gaps requires a command that sees the whole theater and makes the trade-offs. The problem stops being that of a single unit. It becomes one of campaign architecture. Others ran into it before Haiti, in very different conditions.

Bogotá, Rio, Mosul: three mechanisms, not three models

Colombia in the 2000s, the favelas of Rio, and Mosul bear no comparison with Port-au-Prince in scale, adversary, or means. They are invoked here not to be copied, but because state forces there ran up against the same difficulty: taking a space is an operation; preventing the adversary from returning while continuing to advance is a campaign.

Colombia is the most telling case, because this distinction became doctrine there. The security policy of June 2003 was organized around a sequence of “clear, hold, consolidate”: drive out the armed groups; maintain order and prevent their return; reinstall enough institutions that control no longer rests on soldiers. This was not merely doctrinal. According to the US Government Accountability Office, military and police personnel rose from 279,000 in 2000 to 415,000 in 2007; a dedicated joint force, Joint Task Force Omega, was created in 2004 to act against the FARC in their stronghold in the center of the country; in La Macarena, a civil-military fusion center coordinated the army, the police, development, and justice from 2007 onward. According to the Colombian Ministry of Defense, the share of territory under full or partial state control rose from about 70% in 2003 to 90% in 2007. Plan Colombia’s counternarcotics goals were not met, and the FARC remained a threat. But the organization was designed precisely to keep a local victory from condemning the state to start over in the same place: the mobile units press on because other structures hold.

Rio tells almost the opposite story. For years, the security forces carried out incursion after incursion into the favelas: armed entry, firefights, withdrawal. From 2008, the Pacifying Police Units (UPP) replaced incursions with a permanent presence; at their peak, more than 9,000 police officers were assigned to 38 UPPs, and Human Rights Watch found a far steeper drop in police killings in communities with a UPP than in the rest of the state. Then the program expanded faster than its personnel, the promised public services did not follow, abuses undermined trust, and the budget crisis ate away at the whole arrangement. For Haiti, the warning is direct: a position can be occupied without the surrounding space being controlled. When personnel numbers no longer allow for patrols, intelligence, and surveillance of the approaches, the post being held becomes a mere point of state presence in an environment the adversary crosses and observes at will.

Mosul offers a third lesson, with no question of comparing Haitian gangs to the Islamic State. The analysis published in Military Review after the battle shows something counterintuitive: the further an attacker advances into a dense city, the more it slows down, because attrition, logistics, and the burden of consolidation all increase. Some of the best Iraqi units were used up early on; the coalition had to pause to regenerate its combat power. The authors’ recommendation: keep fresh forces for the decisive phases, organize rotations, maintain a reserve. Pausing is not a mistake. The problem begins when one unit’s pause becomes the pause of the entire campaign.

Three paths, one destination. Separate the functions so they can follow one another; do not hold without depth; plan relief and reserves before engagement. That is exactly what the Haitian timeline says. The question is no longer whether the PNH, the FAd’H, or their partners can penetrate a stronghold. They have done it. The question is how to build a system in which each territory retaken increases state control without reducing, in the same proportion, the state’s capacity to retake the next one. Five decisions answer that question, and none of them begins with an equipment order.

Five decisions that do not cost a single armored vehicle

The first: turn coordination into the conduct of a campaign. A joint operations room for the PNH, the FAd’H, and the GSF, placed under the authority of the Superior Council of the National Police, must produce a shared picture of the situation, set priorities, and organize the sequences. Whoever runs it answers not only for the success of an operation, but for what it makes possible or impossible afterward.

The second: impose a minimum standard of preparation before any major operation. A briefing cannot be reduced to reminding the men to advance together and cover one another. Units share a representation of the terrain, the identification of the target, the constraints, the responsibilities, the communications, and the options if the situation changes. Map reading and the use of geographic data become basic skills, maintained through training.

The third: plan the relief before the breakthrough. Specialists break through; other units hold; intelligence stays; investigators and magistrates turn arrests and seizures into case files; the administration returns. The holding plan exists before the assault unit goes in, with the personnel and logistics needed so that a victory does not turn into one more force tied down.

The fourth: build an information-superiority chain. Human intelligence, mapping, image analysis, secure communications, drone detection and jamming within the authorized framework. This chain brings engineers, computer scientists, and analysts to work alongside the forces, including through structured partnerships with universities and technical schools.

The fifth: change the indicators. The number of operations, of people neutralized, or of weapons seized is not enough. Command tracks the interval between major operations, the length of effective control over retaken points, the number of areas kept under pressure at the same time, the time it takes to reoccupy when a group returns, and the speed at which intelligence becomes a decision. These indicators serve first and foremost to conduct the campaign; some of them can be published in aggregate form.

The prerogative to be taken back

The electoral calendar adds a deadline: on 27 July, the Provisional Electoral Council (CEP) set the first round for 13 December and the second for 21 February 2027, provided the security climate is acceptable. As an election approaches that will put the state’s authority directly on the line, security will no longer be measured by the intensity of operations, but by the capacity to make lasting what has been retaken.

This analysis therefore does not end with a request for armored vehicles, drones, or soldiers. Kenscoff was attacked despite a rising police budget, a private company under contract, and an international force in the process of deploying. Haiti already has men capable of going where, only a few years ago, the state no longer went. It has units that fight, soldiers who reoccupy positions, technologies the armed groups fear, partners ready to supply more. What remains to be proven is harder: turning these scattered means into state power.

For a state does not reconquer a capital when its armored vehicle reaches the end of a street. It reconquers it when the officer going in knows what he will find; when the one who stays can be relieved; when intelligence precedes the assault instead of following it; when the state’s communications are less vulnerable than those of the people it is fighting; when the courts return behind the police; and when the next operation begins before the adversary has understood the previous one.

That is what retaking the initiative means. And that is where the question becomes political. Sustaining this continuity requires choices: where to concentrate resources; which units fight and which hold; investing in cartographers, engineers, and communication systems as seriously as in weapons; organizing the PNH, the FAd’H, and the partners around a campaign rather than around their respective institutions; measuring a victory not by the number of press releases, but by the number of territories the state can return to the next day, the following week, and six months later without having to retake them. These choices are not up to the police officer climbing into the armored vehicle. They are up to those in power.

For years, Haiti has lived with an inversion that has become almost ordinary: it is the armed groups that decide which roads are open, which neighborhoods are accessible, at what hours people can travel, which businesses can operate. They do not just control territory. They hand out time, permission, and prohibition. The reconquest begins on the day this prerogative changes sides. It will not be measured by the next neighborhood an armored vehicle manages to enter, but by the day the state decides where to go in, where to stay, where to move, and where to strike next, without the adversary being able to wait calmly for it to wear itself out.

Sovereignty is not a flag planted on a reconquered position. It is the capacity to ensure that what has been retaken today never has to be retaken tomorrow.

A note on method

The timeline was cross-checked against the following open sources. Bel-Air, 30 December 2025: PNH statement of 31 December, Le Nouvelliste. Bercy, 2 January: PNH, Le National, Juno7. Delmas 6, 14 January: PNH, Le Nouvelliste, Reuters. Delmas 2, 19 January: Le Nouvelliste. Carrefour-Aéroport: police substation reopened on 7 February (AP). Marassa, Duval, Santo 25, 1 March, and Martissant, mid-March: Le Nouvelliste of 2 and 23 March. First-quarter results (32 operations, 46 rifles, 19 pistols, 6 drones): PNH press conference of 25 March. FAd’H: Le Nouvelliste of 25 March and 29 April; ministerial inspection of the downtown sites, iciHaïti, 6 June. Village de Dieu, 1 June: PNH, Le Nouvelliste, Rezo Nòdwès. Tabarre, 21 June: briefing by the commander of the Gang Suppression Force to the Security Council, 20 July. Soldiers trained in Martinique: Outre-mer la 1ère, August 2026. Electoral calendar: CEP statement of 27 July. Kenscoff, 23–24 August: tolls of 47 killed, 22 executed in a church, more than fifty abducted, and 2,600 displaced (AP, Haitian Times, Le Nouvelliste, UN Secretary-General); earlier attacks from 4 to 9 July (UN); drone surveillance reported by Le Nouvelliste. The figure of “nearly 118 operations and patrols” is the one reported by Gazette Haïti on 27 May; it is retained as an indicator of activity, not as a count of offensives.

For Colombia, the data on personnel, doctrine, Joint Task Force Omega, La Macarena, and territorial control come from report GAO-09-71 (October 2008). For Rio, Human Rights Watch, “Good Cops Are Afraid” (2016). For Mosul, Thomas D. Arnold and Nicolas Fiore, “Five Operational Lessons from the Battle for Mosul,” Military Review, January–February 2019. On armed groups’ drones: reports S/2025/420, S/2025/597, and S/2026/32.

The passages on operational preparation are based on direct interviews with police officers, including members of specialized units, whose identities are not published. They constitute neither a representative sample nor an audit of training programs; they are used where several accounts converge, and the wording has been confined to what these testimonies make it possible to establish.