Speedboat or oil tanker, and why the answer changed

Running the old emergency services radio network costs the Home Office about £1.7 million a day. A finished Emergency Services Network, the thing meant to replace it, would cost about £0.7 million a day. Both figures are the Home Office’s own, reported by the National Audit Office in its May 2019 assessment of the programme.

The difference is roughly a million pounds a day, and it is not being paid for radios. It is being paid because a decision taken in 2015 about who would build the replacement turned out to be difficult to revisit, and every month of difficulty is billed at the incumbent’s rate.

That is the useful way to think about choosing a technology partner, and it is not the way the speedboat and the oil tanker invite you to think about it.

The metaphor sorts the market on the wrong axis

The image is a supplier’s, and it flatters both ends of the market. The boutique gets to be quick and responsive. The multinational gets to be solid and safe. A buyer is invited to decide how much speed they want and to accept the corresponding trade.

Speed is close to irrelevant. What decides whether a programme survives contact with reality is what it costs to change your mind, and that price is set by contract structure, by who owns the joins between suppliers, and by what happens to your data and your people at the end. None of those correlate with how many staff a supplier has.

The Home Office did not buy small. In 2015 it awarded the three main contracts to EE for priority access to its mobile network, to Motorola Solutions for the software and systems, and to KBR as delivery partner. Later it added Vodafone and Samsung. Every one of those is an oil tanker by any definition anyone uses, and the programme still ran into trouble that had nothing to do with sluggishness.

The tanker’s failure mode is unowned seams

When the Home Office realised in 2017 that its plan was not achievable, it commissioned an independent review, and that review identified five causes of delay. Read them as a list about buying rather than a list about radios.

The delivery partner failed to provide planning and collaboration between the other contractors after its role was downgraded. Motorola and EE had solutions based on different versions of the technical standards. There was disagreement on accountability for systems integration and technical design, and the review found that the Home Office and Motorola had not agreed the true scope of Motorola’s role in integrating the systems end to end. There was no effective process for signing off software in a timely manner. And the related projects that the Home Office kept separate from the main contracts and controlled itself were late, among them the handsets, the vehicle equipment, coverage on the London Underground, and the air-to-ground service for helicopters and aeroplanes.

Not one of those is a supplier being slow. Three of the five are questions about who is responsible for the space between two suppliers, and one is about work the buyer retained and could not deliver.

This is what large-supplier procurement actually buys you. It does not buy integration. It buys several substantial firms, each competent within its own boundary, each with a contract describing its own scope, and a set of joins that belong to whoever was least specific in the negotiation. If nobody wrote down that a named party owns the end-to-end behaviour of the whole thing, the buyer owns it, whether or not the buyer has anyone capable of doing so.

The bill for that is public. The programme is now forecast to cost up to £9.3 billion to 2037, an increase of £3.1 billion or 49% on the 2015 business case. Of that increase, £1.4 billion is the cost of extending the old Airwave system while the replacement is finished. The point at which total financial benefits are expected to outweigh the costs of simply having continued with Airwave moved to July 2029, seven years later than the 2015 forecast.

Look at how that total is composed and the picture gets stranger. In the Home Office’s forecast, the previous service, Airwave from 2015 to 2022, accounts for £2,921 million. The mobile communication service from EE accounts for £1,672 million and Motorola’s user services £1,192 million. In other words the largest single line in the cost of the replacement programme is the thing being replaced. Contingency stands at £714 million, 9% of forecast costs, which the NAO points out is enough to fund an extension of Airwave of less than two years if there are no other cost increases.

Nor is the buyer a small one. When it is ready, 470 organisations are expected to use the network: all 107 police, fire and ambulance services in England, Scotland and Wales, and another 363 organisations across the public, private and third sectors expected to use it and contribute to its costs. This is about as much purchasing power as exists in British public safety, concentrated in one department, buying from firms with the deepest delivery benches in the market. Scale on both sides bought neither speed nor certainty.

And then the sentence that should be read by anyone about to sign a long contract with an incumbent still in place. The Home Office considered that the only options available to it were to reset the programme or to cancel it and continue with the more costly old system. It did not evaluate other options, because such changes would have required an even longer extension of Airwave.

The incumbent’s daily rate had become the thing that defined which futures were affordable to consider. That is what an oil tanker actually does to you. It is not slow. It is expensive to be next to when you want to move.

The speedboat files accounts, and almost nobody reads them

The other end of the market is now unusually easy to examine, because several of the firms doing this work are listed and publish.

Made Tech Group is a supplier of digital, data and technology services to the UK public sector. Its results for the year ended 31 May 2025 report revenue of £46.4 million against £38.6 million the year before, contracted backlog of £92.2 million, sales bookings of £82.1 million, profit before tax of £2.0 million after a loss of £3.0 million in the prior year, and total headcount excluding contractors and partners rising from 349 to 374. Four customers each accounted for more than 10% of revenue, and the largest represented £8.7 million, or 19% of the group total.

Everything a buyer needs to reason about the risk of hiring that firm is in that paragraph, and none of it appears in a capability deck.

Three hundred and seventy four people is the entire company. If your programme needs forty of them, you are asking for more than a tenth of the firm, and you are competing for those people with every other client. A backlog of £92.2 million against revenue of £46.4 million is roughly two years of contracted work, which is genuine security and also a constraint on how quickly they can take on something new. A swing from a £3.0 million loss to a £2.0 million profit in one year is a firm whose margin is thin enough that a delayed payment run is a real event for them.

The concentration figure is the one buyers misread most often. A supplier with a customer at 19% of revenue is usually presented as a risk to the supplier. It is at least as much a problem for the customer, because a supplier that cannot afford to lose you cannot afford to tell you that your plan is wrong. Organisations hire small specialists precisely for candour and then quietly buy enough of their capacity to make candour expensive.

The other thing the filings tell you is that the firm you hired is not the firm that will deliver. A supplier growing headcount by a quarter over two years is staffing your work with people it had not met when it wrote your proposal. Acquisition changes it faster still, and the digital services market in Britain has spent a decade consolidating.

What you are buying is optionality

Both hulls are fine. What sinks programmes is the water in between, and the specific form the water takes is a decision that turns out to be more expensive to reverse than anyone priced.

Most enterprise technology is now bought as a running service rather than as a build. That is the change that makes the old version of this question obsolete. When you bought a build, capacity and speed were the sensible things to compare, because the relationship had a defined end. When you buy a service, the relationship has no defined end, and the only property that matters over ten years is what it costs you to stop.

Which means the questions worth asking during selection are not about method or velocity. They are about what exists on the day you want out. Whether you receive the data in a documented format that somebody else can read without a professional services engagement. Whether the code written for you is yours, and whether anyone has checked that claim against what is actually deployed. Whether you may hire the people who have been embedded in your organisation for three years, or whether a non-solicitation clause makes your own institutional memory unpurchasable. Whether a transition period is defined in days and paid at a rate agreed now rather than at a rate negotiated at the worst possible moment.

None of that is adversarial. A supplier confident in its work is usually relaxed about exit terms, because exit terms only bite when the client wants to leave. It is the negotiation position of the party that expects to earn most of its money after year three that hardens on those clauses, and that is a signal worth reading.

The reset the Home Office announced in September 2018 shows what optionality looks like when you have to buy it back. The approach changed from launching the whole programme at once to introducing services in stages, so that emergency services could test discrete elements and build confidence before committing. That is a better design. It was also arrived at after the contracts were let, which meant revising the whole programme, extending timetables and renegotiating agreements, a process the NAO recorded as still under way at the time it reported. And the crucial unknown was recorded honestly: users would be offered a limited service from late 2019 or the full system from 2021, but their appetite to adopt those early products was not yet known.

That is the position a buyer ends up in when the decision to stage delivery is taken second rather than first. The staging is now a renegotiation rather than a design choice, and the question of whether anyone will use the early increments is being asked after the money is committed rather than before.

Four documents to read before signing

The filed accounts, including the customer concentration disclosure and the headcount. If the supplier is private, the Companies House filing still tells you turnover band, employee numbers and whether the auditors said anything.

The backlog, and whether it is contracted or pipeline. Those are different words and suppliers use them interchangeably in conversation and precisely in filings.

The exit schedule, read out loud in a room containing the person who would have to execute it. Most exit schedules have never been read by anybody who would be responsible for the exit, which is why they contain obligations nobody could meet.

And the integration accountability, in writing, naming a party. Where more than one supplier is involved, this is the single clause that decides whether a programme has a systems integrator or has an argument.

The sequencing question underneath all of this, which is whether the problem was understood before the product was chosen, is worked through in the solution is the last thing you should pick. The running cost of the automation that increasingly forms the substance of these contracts is set out in intelligent automation, and the wider programme coverage sits under digital transformation.

The Home Office is not slow, and neither are its suppliers. It is stuck, and stuck has a daily rate.