The Kitchen Is Leaving the Building: Why Central Production Is Reshaping European Food Service

Michael Ingemann
8 hours ago
10 min read

Something structural is happening in European foodservice and it is not a trend that will reverse when conditions improve. Across the UK, Scandinavia and Southern Europe, operators of every size and format are arriving at the same conclusion: the model in which skilled chefs produce food from scratch in every kitchen, at every site, every day, is no longer commercially viable at scale. Labour is too expensive, too scarce and too inconsistent. Food costs are too volatile. The margin arithmetic simply does not work.

The response is to move production off-site into centralised facilities and redesign front-of-house kitchens around assembly, finishing and service. This shift is gathering serious momentum. Central production units or CPUs, are no longer a niche solution deployed by a handful of large contract caterers. They are becoming a structural feature of how premium foodservice operates across multiple formats and multiple markets.

But the transition is not straightforward and the gap between operators who benefit from it and those who do not is almost entirely execution. Our research across European foodservice markets, including conversations with operators, caterers, supply chain specialists and food entrepreneurs, reveals a consistent picture: the model works brilliantly when committed to fully and fails expensively when implemented halfway. Understanding why is the starting point for any operator considering this shift.

The Forces Making the Old Model Unworkable

The pressures driving this shift are not new but they have reached a tipping point. Labour cost inflation in hospitality has outpaced general wage growth for several consecutive years across most European markets. The skilled chef pipeline, already under strain before 2020, did not recover after it. The hospitality sector lost experienced kitchen staff to other industries during the pandemic and has been unable to fully replace them. The result is a structural skills shortage that cannot be solved by paying more alone.

Food cost volatility has compounded the challenge. Input prices across proteins, dairy, produce and packaging have remained elevated and unpredictable, making margin management at the site level an increasingly difficult discipline. When both major cost lines, labour and food, are under simultaneous pressure, the traditional kitchen model, which optimises for neither, starts to look unsustainable.

The operators who have moved earliest and most decisively toward centralised production are those who recognised that these pressures are structural rather than cyclical. They are not waiting for conditions to normalise. They are redesigning their operating model for the conditions that exist and are likely to persist.

The operators who benefit most from central production are not those with the best suppliers. They are those who redesign everything, staffing, forecasting, logistics and menu planning, around the new model simultaneously.

What Central Production Actually Does and Does Not Do

A central production unit, at its most effective, is not simply a larger kitchen producing food in bulk and distributing it. It is a fundamentally different operating philosophy. The CPU separates two activities that the traditional kitchen conflates: creative development and technical production. These require different skills, different mindsets and different management approaches. Conflating them in a single kitchen at every site is inefficient, expensive and increasingly unrealistic given the available labour pool.

In a well-designed CPU model, trained production specialists who excel at consistency, volume management and technical execution handle the manufacturing of components, sauces, proteins, bases and finishing elements. Creative chefs and culinary leads work at the development level: designing dishes, testing recipes, responding to seasonal and market changes and ensuring the output of the CPU retains genuine culinary ambition. The two functions are distinct but connected and keeping them connected is one of the most important and most underestimated challenges of the model.

What the CPU does not do is eliminate the need for kitchen skill at the point of service. The finishing, plating, temperature management and quality control that happen at the front end still require competent people. What changes is the nature of the skill required: less raw production expertise, more assembly intelligence, quality awareness and speed. This is a genuinely different labour profile and operators who do not update their recruitment and training approach to match it undermine the economics of the whole model.

The CPU also does not guarantee consistency. It creates the conditions for consistency, which is different. Consistency is achieved when the production process is tightly documented, the logistics are reliable, the cold chain is unbroken and the front-end team understands exactly what they are receiving and how to handle it. Every one of these conditions requires active management. A CPU that produces excellent components and distributes them to sites with poor logistics, undertrained staff or inconsistent handling will produce inconsistent results regardless of what happens at the production end.

The Commercial Case: Why the Numbers Work at Scale

The economic argument for central production is compelling when modelled correctly. Most operators who reject it do so because they have modelled it incorrectly. The comparison is not between the cost of a component bought from a CPU and the cost of the same ingredient produced on-site. That comparison will almost always favour on-site production, because it ignores most of the real costs of the traditional model.

The correct comparison accounts for the full labour cost of on-site production, including management time, training costs, wastage driven by skill variability and the cost of inconsistency in the finished product. It accounts for the capital cost of maintaining full kitchen infrastructure at every site. It accounts for the regulatory and compliance cost of running a production kitchen, extraction, gas, fire suppression, environmental health, which is concentrated and amortised across volume in a CPU rather than replicated at every location.

When the full cost picture is drawn correctly, the economics of central production become significantly more favourable. Operators who have made this transition with genuine commitment report site-level EBITDA improvements that reflect not just food cost changes but the compounding effect of labour rationalisation, reduced capex requirements and improved consistency driving higher guest satisfaction and repeat spend. In some cases, the space freed by the removal of full kitchen infrastructure at the site level has been converted into revenue-generating use, additional covers, event space or other front-of-house capacity, delivering a double commercial benefit.

The cost of the CPU itself, expressed as a percentage of revenue, also reduces materially with scale. An operator running a CPU to serve three or four sites faces a very different unit economics picture than one serving fifteen or twenty. This is one of the strongest arguments for operators to think about central production not as a facility they build for their current estate but as infrastructure they design for where the estate will be in three to five years.

The comparison is not between the cost of a component from a CPU and the same ingredient produced on-site. That comparison ignores most of the real costs of the traditional model.

The Failure Modes: Why Half-Measures Are Worse Than Nothing

The most consistent finding from our research across European markets is also the most sobering: the majority of operators who attempt to introduce central production without fully committing to the model fail to realise meaningful benefit. Worse, many end up in a position that is commercially worse than before, carrying the cost of a CPU or a third-party component supplier on top of an unreformed on-site kitchen operation.

The most common failure mode is labour doubling. An operator introduces menu components from a central source but does not reduce on-site kitchen staffing to match. The result is a kitchen paying twice: once for the component and once for the people who used to produce it. The savings that justified the model evaporate before they are realised. This happens most often when operators lack the management conviction to make the staffing changes that the model requires or when they introduce components incrementally rather than committing to a threshold that makes the labour restructuring defensible.

The second failure mode is scope creep. A CPU that starts by producing three or four core components gradually accumulates requests: a new sauce here, a seasonal garnish there, a special for a particular event. Eventually it is producing forty SKUs for an estate that only has the volume to justify fifteen efficiently. The unit economics deteriorate, the production team loses focus, quality control becomes harder and the CPU starts to feel like a problem rather than a solution.

The third failure mode is the loss of culinary identity over time. This is the most insidious because it is slow and is often not recognised until the damage is significant. Production kitchens optimise for efficiency and replicability. Without deliberate structural investment in culinary development, dedicated chef time, NPD processes, regular review of what the CPU is producing against what the market wants, the output slowly gravitates toward the safely producible rather than the genuinely compelling. The brand loses its culinary edge and guests notice, even if they cannot articulate why the food feels less interesting than it used to.

The fourth failure mode is logistics dependency without logistics investment. A CPU model requires a cold chain, a distribution operation and a receiving process at every site that are reliable and well-managed. Operators who invest in production but not in logistics find that the quality achieved at the CPU is degraded by the time it reaches the plate. The logistics are not a supporting function of the CPU model. They are an equal part of it.

Which Formats Benefit Most and Which Are Not Ready

Not all catering and foodservice formats are equally suited to central production and one of the most useful things an operator can do before committing to the model is an honest assessment of where they sit on the spectrum.

Event and venue catering sits at the highly suited end. Volume is predictable with advance notice, menus are designed around the event rather than daily variation and the logistics of delivery are inherent to the format. Operators in this space are often already using centralised preparation, whether through in-house CPUs or third-party suppliers, for a significant proportion of their offer. The opportunity is to formalise and extend what is often already happening informally.

Multi-site restaurant groups with stable core menus are the next most naturally suited format. The conditions that make a CPU commercially attractive are present: volume, menu consistency and predictable demand and the operational complexity of managing production across multiple sites creates a genuine incentive to consolidate. The transition requires menu discipline: a willingness to reduce the number of SKUs being produced and to design menus around what a CPU can produce excellently rather than everything a creative kitchen team would ideally like to offer.

Premium workplace and contract catering presents a more complex picture. Daily menu variation, which is central to the value proposition of many B&I caterers, creates a fundamental tension with the logic of centralised production, which benefits from consistency and repetition. Operators in this format who have made the CPU model work have typically done so by separating their offer into a stable core, sauces, bases, proteins, snack components, that is centralised and a variable daily element that retains on-site preparation. This hybrid approach captures some of the economic benefit of centralisation without requiring the full menu standardisation that daily-changing formats resist.

Daily-changing scratch catering at the premium end of the B&I market is the format least naturally suited to full CPU integration. The culinary identity of this format is built around freshness, daily change and the visible presence of skilled chefs. Introducing central production without compromising what guests value about the experience requires extremely careful design and a level of culinary sophistication at the CPU that most production facilities do not yet have. This does not mean the model is irrelevant here, but it means the starting point is component-level support rather than menu-level outsourcing.

What Operators Who Get It Right Have in Common

Across our research conversations, a consistent set of characteristics emerged among the operators who have implemented central production successfully and sustained the benefit over time. They are worth articulating clearly because they are as much about organisational culture and leadership as they are about technical capability.

The first is wholehearted commitment from leadership. The CPU model requires decisions that are difficult and sometimes unpopular: about staffing, about menu scope and about logistics investment. Operators who have implemented it successfully have done so because senior leadership understood the full model, committed to it publicly and held the organisation accountable for executing it properly. Operators who positioned it as a trial or who allowed the front-of-house teams to opt out of the staffing changes, consistently failed to realise the benefit.

The second is a clear and maintained distinction between creative development and production execution. The best operators treat these as separate disciplines with separate teams, separate success metrics and a structured process for moving ideas from the development kitchen into the production environment. They invest in culinary talent at the development end not despite running a CPU but because of it, understanding that the creative credibility of what the CPU produces is the thing that prevents the slow commoditisation of their food offer.

The third is menu discipline. Successful CPU operators are ruthless about what goes into the production scope and what does not. They design menus with the CPU in mind, building dishes around components that can be produced at volume and finished with skill at the site. They resist the natural tendency to add complexity over time and they review the production scope regularly against both volume and culinary performance.

The fourth is investment in logistics and front-end capability. The operators who sustain the quality of their CPU output understand that the cold chain, the delivery schedule, the receiving process and the front-end assembly training are not secondary to the production investment. They are equal parts of the same system and they require the same management attention.

The Horizon: Where This Is Going

The direction of travel in European foodservice is clear. Central production will become the dominant model for multi-site foodservice operations across most formats over the next five to ten years. The economic pressures driving the shift, labour cost, skills scarcity, food inflation, capex intensity, are not going to resolve. They may ease at the margins but the fundamental arithmetic of scratch-cooking at scale in high-cost European labour markets has become untenable.

What will change is the sophistication of the CPU model itself. The current generation of central production in foodservice is largely focused on the basics: sauces, proteins, bases and semi-finished components. The next generation will extend further up the culinary stack, producing finished or near-finished dishes of genuine quality that challenge the assumption that centralised production and premium output are incompatible. Technology, in refrigeration, in packaging, in reheating and finishing equipment, is advancing to support this.

The specialist third-party CPU operator is also emerging as a significant category in its own right. Rather than every operator building and running their own central facility, which requires a scale of operation and a management capability that many groups do not have, the market is developing suppliers who offer production-as-a-service: high-quality menu components produced to a client's specification, delivered on a schedule that works for their operation, at a price that reflects genuine volume efficiency. This model lowers the barrier to entry for operators who want the benefit of central production without the capital and operational complexity of running the infrastructure themselves.

The operators who will be best positioned in five years are those who are making these decisions now: committing to the model, investing in the logistics, restructuring the labour approach and building the culinary development capability that keeps their CPU producing food that guests genuinely want to eat. The kitchen is leaving the building. The question is not whether to follow it. It is how quickly and how well.