Twelve years on, what the Horizon replacement tender tells us
Computer Weekly confirmed this week that Accenture has joined IBM and DXC on the shortlist to replace Fujitsu’s Horizon system. This is the Post Office’s third attempt in twelve years. The same firms keep appearing. The interesting question is not who wins, it is why the shortlist is always so short, and what that tells buyers about the procurement and delivery model itself.
On 14 May 2026, Computer Weekly confirmed that Accenture has joined the bidding for Lot 2 of the Post Office Horizon replacement tender, a £160 million contract for an off-the-shelf EPOS system. Two months earlier, the same publication confirmed IBM was bidding for Lot 1, the £323 million “walk in and take over” contract, with DXC as partner. A handful of further bidders are in competition for both lots. The decision is expected in July 2026.
Most coverage of the Horizon replacement has framed it as a single procurement story, complicated by the inquiry into the scandal it is intended to close. A more useful reading, for buyers in any sector running large integrator-led IT programmes, is structural. This is the Post Office’s third attempt to replace Horizon in twelve years. The first two failed in different ways with different sets of suppliers. The shortlist now in front of the Post Office board is drawn from substantially the same group of firms that featured in both. The pattern is repeating because the procurement and delivery model has not changed.
Three attempts, one shape
The first attempt began in 2013 as a £100 million multi-supplier programme, structured as a service integration and management (SIAM) arrangement with IBM, Computacenter, Atos, Verizon and Accenture covering the service towers, and Atos providing the integration layer. It was cancelled in 2015. Public reporting noted that the Post Office paid IBM several million pounds in compensation. None of the planned replacement of Horizon was delivered.
The second attempt, the Strategic Platform Modernisation Programme (SPMP), began in 2021 with a different structure. The Post Office would build a new platform, New Branch IT (NBIT), in-house on AWS, with external partners augmenting internal capability. Accenture’s involvement began with a £27 million contract in May 2022, expanding subsequently to a reported £50 million. Coforge held a complementary £18 million contract under the same Digital Capability Framework. By April 2024, a further £75 million three-year framework was being competed, with Accenture and Coforge named as the main development partners.
By May 2024 the numbers were difficult to defend. The Infrastructure and Projects Authority rated NBIT “currently unachievable.” The £180 million budget had become a £1.1 billion projection. Implementation had slipped by up to five years. Around 80% of the staff working on NBIT were contractors, a detail the then-Chairman drew specific attention to when he told the public inquiry he was “too reliant on Accenture.” When the Post Office board responded by cutting short-term spending, roughly 70% of the project team was let go, mostly contractors, with the predictable consequences for retained knowledge and morale. By late 2024, the in-house build was being wound down in favour of an off-the-shelf alternative.
The third attempt, the procurement now under way, is that off-the-shelf alternative. The shortlist is drawn from substantially the same group of firms that featured in the 2013 SIAM attempt, with IBM bidding for Lot 1 (DXC as partner) and Accenture bidding for Lot 2. The procurement structure differs in detail. The structural shape is the same: large scope contract, scale-capable bidder pool, integrator-led delivery, balance sheet sized to absorb the risk. This is not a criticism of any of the bidders. The structural question is why all three attempts have produced shortlists drawn from the same small pool, and whether the resulting market is functioning the way the buyer thinks it is.
Why the shortlist is always the same
Three features of how large-scope IT outsourcing is procured consistently produce the same shortlist.
Scale capability bias. Procurement scoring frameworks weight evidence of having previously delivered at comparable scale, which by construction favours the firms that have already done so. Few firms outside the top tier have the references that score well, regardless of whether their proposed solution is genuinely better. The effect is cumulative: the firms that won the last large contract are the firms best positioned to win the next one, because they now have a fresher reference.
Balance sheet as procurement gate. Contracts of this scale carry termination liabilities, performance guarantees and indemnity expectations that only a relatively small number of firms can credibly underwrite. This is reasonable from a risk perspective, but it has the effect of pre-selecting the bidder pool before any evaluation of technical fit. A firm with a better operating model and a smaller balance sheet rarely makes it to the shortlist.
Single accountable supplier preference. Buyers consistently prefer a single supplier they can hold accountable, rather than a more competitive multi-supplier arrangement with an independent integrator. The 2013 Horizon replacement attempted the latter; it failed for reasons specific to its execution, but the broader lesson the market took from it was that SIAM is hard, which has pushed buyers back toward single-supplier structures. Single-supplier structures heavily favour the firms with the breadth to bid for everything.
Each of these features is individually defensible. In combination, they produce a procurement model whose outcome is largely determined before the technical evaluation begins, and which selects consistently for a small group of large integrators.
What the second attempt actually proved
The NBIT story is worth dwelling on because it is the most instructive of the three attempts and the most likely to be forgotten in the next round.
The second attempt was structured to address the perceived failure of the first. Single accountable buyer rather than SIAM. In-house build to retain platform control. Cloud-native architecture. External partners augmenting internal capability rather than replacing it. Each decision was a reasonable response to a specific lesson from 2013.
What NBIT demonstrated is that the choice between SIAM, in-house build, and single-supplier outsourcing is not the most important variable. All three can fail. What separated NBIT from a successful programme was not the delivery model but the foundations underneath: the maturity of the in-house engineering function, the balance of permanent against contractor resource, the realism of the budget against the scope, the governance structure when the programme came under pressure, and the buyer’s ability to challenge its delivery partners when the evidence started pointing the wrong way. None of these are choices about which supplier to use. All are choices about how the buyer organises itself to be capable of running the programme. We have written before about why independent technical delivery management matters in complex programmes; the NBIT story is one of the cleanest illustrations of the argument. The IPA review did not find that Accenture or Coforge were poor partners. It found that the programme was unachievable on the terms it had been set, and that governance had not surfaced that fact early enough.
Why this matters beyond the public sector
It would be easy to read this as a public sector observation. It isn’t. The same procurement model, with minor variations, is how most large private sector IT outsourcing is bought. Regulated financial services buyers, large manufacturers, multinational retailers, all use weighted scoring frameworks, balance sheet thresholds and single-supplier preferences to manage the risk of large transactions. The same firms appear on the shortlist. The same in-house-versus-outsource pendulum swings every five to seven years, shaped by the failure of the previous swing. The same contractor-heavy delivery models accumulate the same dependencies. We have written before about how unexamined dependencies on a small number of strategic suppliers limit organisations’ ability to act (Are your IT suppliers holding you back?); the procurement and delivery model is the upstream cause of that downstream dependency.
The foundations questions
Before running another competitive tender or another in-house build, three foundational questions are worth asking honestly. The NBIT experience sharpens each one.
Is the in-house capability honest about its limits? The NBIT decision to build in-house was reasonable in principle but depended on a level of permanent in-house engineering capability that did not exist on day one. Substituting contractors for that capability bought delivery capacity at the cost of retained knowledge and long-term ownership. Either the buyer builds the capability properly before relying on it, or accepts the structural cost of the dependency.
Is the delivery model deliberate? SIAM, single-supplier outsourcing, in-house build with augmentation, off-the-shelf with implementation partner: each has known strengths and known failure patterns. The Post Office is trying its third combination across twelve years. The lesson is not that any of them is structurally wrong. It is that none of them works without the foundations underneath.
Does the governance challenge the programme honestly? The NBIT IPA review surfaced the unachievability rating only after the budget had grown sixfold. Earlier challenge would have produced different decisions. The question every large buyer should ask of their own programme governance is whether it is structurally capable of telling the board something the board does not want to hear, and whether the people most able to do so are independent of the parties whose work is being evaluated.
Where smaller independents change the picture
The structural pattern is not a counsel of despair. It is a description of where the model leaves room for a different kind of contribution, and where the buyer can reach for help that does not come with the conflicts the large incumbents carry by definition.
Specialist Independent advisory firms have a different commercial geometry from the integrators that win the contracts. They do not bid for the implementation work. They do not need the buyer to choose a particular delivery model to protect a future revenue line. They can be honest about whether the in-house engineering function is ready to lead the build, because they are not the firm that would augment it. They can challenge the scoring framework without being scored against it. They can advise the governance forum without being one of the parties whose work is being evaluated. These are not differences in personnel or capability. They are differences in incentive.
The “perception point” matters as much as the substance. Even where a large consulting firm acts with complete integrity, and most of the time they do, the structural conflict means the advice can reasonably be perceived as conflicted. It does not matter that the supplier chose the option on its merits if half the organisation thinks the choice was made for biased reasons. Decisions made under that cloud get re-litigated for the life of the programme, drain the buyer’s political capital to defend, and undermine the governance forum’s ability to make the next decision. Independent advice does not just avoid the conflict. It avoids the perception of conflict, which is often the more expensive of the two.
What that produces, in our experience, is a different quality of conversation early in the procurement and delivery cycle. Sharper questions at the shortlisting stage. An integration model chosen for the buyer’s actual constraints rather than by default. An in-house capability assessed honestly before being relied upon. A contract shape sized to the market the buyer wants to compete, not the one the incumbents prefer. Governance designed to surface uncomfortable findings rather than to manage them.
None of this is a substitute for the large integrators that ultimately deliver. The contracts being competed in May 2026 will be won by firms with the scale to absorb the risk, and rightly so. The role of the specialist independent advisor is to make sure the buyer enters that procurement with a clear-eyed view of what they are actually buying and how it will need to be governed.
Final thought
Three attempts in twelve years is the empirical evidence that the procurement and delivery model, not the supplier choice, is the variable that needs the most careful thought. The integrators are good at what they do. The in-house build was a credible response to the failure of the SIAM attempt. The off-the-shelf approach is a credible response to the failure of the in-house build. None of them has succeeded yet, because the foundations underneath the supplier choice have not changed across the three attempts. Smaller, independent voices in the buyer’s ear at the right moments, before the shortlist is fixed, before the delivery model is locked in, before the governance is designed around the parties whose work it is meant to evaluate, are one of the most efficient ways to change that.