The Rodri Transfer Lesson: Why Performance Does Not Move With the Asset You Buy

A deal that is not done, and a question bigger than the deal

At the time of writing, neither Manchester City nor Barcelona has made an official announcement about the transfer of Rodri Hernández. What credible reporting establishes is that the player has agreed personal terms with the Catalan club, and that the remaining gap is the fee: an offer around €70 million against a demand close to €80 million. That followed a full run of contradictory news, including links to Real Madrid that then collapsed, and widely shared clips claiming an official announcement that never happened.

That is a first lesson worth pausing on: in a fast-moving market, the cost of building a decision on unconfirmed news is higher than the cost of waiting one day. But what matters for your business is not the fate of the deal. It is the question every analysis circles around: if Rodri does move, does what he delivered move with him?

Performance is a property of the system, not of the component

Rodri was not simply a good player in a good team. He was the hinge an entire system was built around: a side that monopolises possession, relies on a single deep midfielder to receive under pressure and distribute in minimal touches, and kills the counter-attack before it starts. The output was four Premier League titles, the Champions League in 2023, and the Ballon d'Or in 2024.

But that output was not produced by the player alone. It was produced by the interaction between the player and a specific system: a manager with a defined philosophy, team-mates drilled for years in fixed movement patterns, and a defensive line willing to push up because someone reliable covers behind it. Remove any one of those and you do not get the same ratio of output, however good the player himself is.

This is precisely the mistake most often made when buying technology: buying the component and expecting the performance of the system.

The reverse test: what happens when the component is missing

The methodological luck here is that the test actually ran. On 22 September 2024 Rodri ruptured his cruciate ligament against Arsenal and missed the rest of the season, returning in May 2025. In that season Manchester City finished without major silverware and lost the FA Cup final 1-0 to Crystal Palace on 17 May 2025.

It would be wrong to reduce an entire season to the absence of one player; other factors were in play. But the observation stands: a high-performing system saw its output drop visibly when a single component with no equivalent replacement fell out of it.

That gives us the rule in both directions: the system depends on its critical components, and the component depends on the system it operates inside. Anyone selling you one and promising you the output of the other is selling you an illusion.

Translating the rule into technology buying decisions

Every manager has met this situation in one form or another:

• Bought an ERP that works superbly at a competitor, and it did not produce the same result, because the competitor had ordered its processes before the purchase rather than after it.
• Brought in an AI model topping the benchmarks, then discovered the problem was never the model but that the company's data sits in files the model cannot reach at all.
• Hired a manager with a striking track record elsewhere, who then struggled, because that earlier success rested on authority, a team, and tooling that did not exist in the new environment.

In all three cases the error is the same: assuming performance is an attribute glued to the asset that travels with it wherever it goes. Performance is the result of an interaction, and whoever buys the component without preparing the system has bought the most expensive part of the equation and neglected the part that determines its value.

The right question before any technology purchase is not "is this the best on the market?" but "what has to be ready on my side for this best to give what it can?"

The second dimension: readiness and AI in risk assessment

The other element in valuing this deal is physical: a player of 30, back from a major injury, priced near €80 million. Clubs no longer settle such a decision on impression. They draw on training-load data collected by wearable devices and match and training tracking systems, and build models from it to estimate injury probability and manage return-to-play loads.

Here we should be precise, because this field is sold for more than it delivers: systematic reviews of dozens of studies on machine learning and injury prediction conclude that results are promising but of limited practical value so far. The reasons are well known — small samples, severe class imbalance because injuries are rare against thousands of healthy sessions, and weak external validation beyond the club whose data trained the model.

This is exactly what we repeat to clients in a completely different context: predictive maintenance does not start with a prediction model. It starts with a clean asset record and disciplined measurement over a sufficient period. Running a model on incomplete data produces false alarms, and within two months everyone stops paying attention to them — so you have paid for a tool that cost your team its trust in alerts. The rule is the same whether the asset is a pump in a plant or a player in a squad.

The gap between €70m and €80m is not haggling — it is two different valuation models

When one side offers 70 and the other asks 80, the disagreement is usually not about willingness but about method: how many productive years remain, what the probability of re-injury is, what the commercial value is, and what the replacement would cost if the deal fell through. Each side arrives at a different number because each runs a different model.

The professional parallel is direct: when you negotiate a system or a development contract, a dispute over price is usually a dispute over the value model rather than the figure. Whoever enters that negotiation with a written total-cost-of-ownership calculation across the system's full life — training, support, migration, and likely downtime included — negotiates from an entirely different position than someone comparing headline quotes.

Five questions before you buy "the best on the market"

• What specific outcome am I buying, and how will I measure it in three months? If the answer has no number, you are buying an impression.
• What must be ready on my side before go-live — data, written processes, permissions, a named owner — and what does preparing it cost?
• Was this solution's success elsewhere caused by the solution itself, or by an environment I do not have?
• What is the exit plan? If I want to switch in two years, what do I lose, what does it cost, and is my data genuinely portable?
• What is the cheaper alternative that delivers 80% of the result, and why did I rule it out?

The Origami view

The request that reaches us most often starts with a sentence like: we want that particular system, or we want AI like the one we saw at such-and-such company. Our first answer is always a question: which decision do you want to change after go-live? If the answer is clear, delivery becomes an engineering matter. If it is not, then whatever tool is bought, however advanced, will join the list of systems that were paid for and nobody uses.

That is why we start our projects by describing the process before choosing the tool, and insist that every system has a named owner inside the company and at least one indicator measured before and after go-live. Companies that do this get excellent results out of an average tool; companies that skip it get a document folder and unused licences out of the best tool on the market. The component matters, but the system decides its value.

Sources

• BBC — Rodri ruled out for the season with an ACL injury: bbc.com
• Associated Press — Crystal Palace beat Manchester City in the 2025 FA Cup final: apnews.com
• The Athletic — developments in Barcelona's offer for Rodri: nytimes.com/athletic
• Systematic review of machine learning methods in sport injury prediction and prevention: link.springer.com