Reflections from the NHS Covid Contact Tracing App Programme (Part I)
Prof. Sir Anthony Finkelstein
President, City St George’s, University of London
Published on April 10th, 2026
This is a story that has not been told, at least not in full. I will confine myself to the scientific and technical matters and will not seek to draw broader policy conclusions. What follows is a view of science advice in challenging settings from a specific example from my time as a government servant, a Chief Scientific Adviser in HM Government.
This episode is, in a modest way, an illustration of science advice operating under institutional stress. The pandemic created conditions in which established procedures were suspended, authority structures shifted, and decisions had to be made under extreme uncertainty. The questions raised in such circumstances about the role of scientific advisers, their proximity to power, and the boundaries between advice and decision, are not unique to emergencies.
The NHS Covid Contact Tracing App
This story concerns the development of the NHS[1] Covid Contact Tracing App. In early 2020, I served on the early Covid SAGEs[2], offering digital and modelling advice. It was difficult to make a meaningful contribution at this stage, when the questions were principally clinical and epidemiological. I did my best.
In March 2020, the Government Chief Scientific Adviser (GCSA), Patrick Vallance, asked me to assess contact‑tracing apps, following interest from a team at the University of Oxford and reports from China and Korea. We convened an ad‑hoc group with Oxford, NHSX and colleagues from government and academia.
Modelling & Design
There were two strands to the early work. First, modelling showed that app‑based tracing could be effective, but only with rapid deployment and high uptake. Second, we explored technical implementation, ultimately selecting Bluetooth Low Energy (BLE) for its short‑range, privacy‑preserving proximity detection.
The epidemiological case for an app was strong, and both the GCSA and Chief Medical Officer agreed the work should continue. NHSX[3] became the programme’s home, although building a consumer application at national scale, under emergency conditions, with no established team and no precedent, is not what digital transformation units are typically designed to do.
The approach was simple. Most people carry a smartphone. Smartphones can detect other smartphones in close proximity using Bluetooth. If the phones log those encounters anonymously, and a person subsequently tests positive, everyone whose phone recorded sustained proximity can be alerted before they develop symptoms and infect others. The intention was to interrupt chains of transmission before they formed. The epidemiological case was strong. The execution would prove considerably more complicated than the concept.
In terms of the technical implementation Bluetooth Low Energy was not the only option. We considered alternatives (GPS, telecoms data, manual tracing) but none matched BLE’s suitability. The modelling was the foundation of the case for proceeding. Without a credible quantitative argument that app-based contact tracing could materially affect the course of the epidemic, the programme had no justification for the resources and political capital it would consume. The Oxford team's work was rigorous. The parameters, however, were uncertain. We were modelling a novel virus, with imperfect data, against a population whose behaviour was itself changing in response to the epidemic.
The team that was assembled under pressure was not conventional, but capable and committed. Early progress was rapid but improvised. There were constraints in emergency procurement, institutional friction. There were also additional challenges. By way of example, Faculty, the AI company, was engaged on data analytics work of considerable importance elsewhere in the government's pandemic response. The two programmes were distinct but occupied overlapping institutional territory. The existence of Faculty's work complicated the public and political environment around data and technology in ways that occasionally created noise the app programme could have done without.
Institutions & Politics
Government procurement is an elaborate system designed to ensure fairness, value for money, and accountability. It is not designed for speed. The emergency procurement routes that existed were used, and used legitimately, but they created their own problems: compressed due diligence, relationships formed under pressure that were difficult to unpick later, and a perception among some observers that corners had been cut even when they had not. Procurement under these conditions is a genuine dilemma. Slowing down to do it conventionally had a cost measured in transmission events. Moving fast had costs of a different kind.
Industry wanted to help. Some of it genuinely did. Technology companies, consultancies, and a constellation of smaller firms with relevant capabilities made themselves available, and some of that engagement was valuable. There was also pressure of a less straightforward kind – companies positioning for contracts, advocates promoting particular technical approaches for reasons that were not purely epidemiological, and a general tendency for the programme to attract people who wanted proximity to something consequential. Distinguishing between these was a constant background task. It consumed attention that could otherwise have been directed elsewhere.
Ministers were engaged with the app in a way that reflected its visibility. A contact tracing app was something the public could understand, download, and see. That made it politically legible in a way that, say, improvements to the testing supply chain were not. The attention was not uniformly helpful. Programmes of this kind benefit from political cover and political patience. They are less well served by political impatience presented as interest. We experienced elements of both.
The privacy debate arrived early and intensified as the programme became more public. The centralised architecture became a proxy for a broader argument about state power that was only loosely connected to what was actually being built. The noise generated by that argument had real consequences for the programme, for public trust, and ultimately for the decisions that were made about architecture. It is possible to take privacy seriously while still concluding that the debate, as conducted, was often about something else.
For those providing scientific advice, this is a familiar environment. Technical arguments become proxies for broader political anxieties. Advisers must navigate this carefully, remaining analytically rigorous while recognising that the debate is no longer purely technical.
Science advisers inevitably operate close to political authority. That proximity is both the source of influence and the source of risk. Influence without proximity is rare. Yet proximity creates the possibility that advisers become identified with decisions that are ultimately political. Managing that boundary is part of the craft of science advice.
Technology & Constraints
When the programme moved from concept and design to testing, the Isle of Wight trial in spring 2020 confirmed BLE’s functionality on Android but revealed iOS restrictions: Apple did not permit background Bluetooth use, crippling the centralised architecture. In April 2020, Apple and Google announced a joint Exposure Notification API. This reframed the programme. Governments that wanted a functioning app on both platforms would need to build to it. Governments that wished to retain centralised architectures would have to accept that their apps would not work properly on iPhones. The choice was therefore less technical than political, expressed through technology.
NHS Test and Trace was announced in May 2020. Baroness Harding was appointed to lead it. The creation of a new organisational structure at that point in the programme had consequences. The team that had built the first version of the app, and that carried the institutional memory of the decisions made and the reasons for them, was disestablished. In complex technical programmes institutional memory is itself a form of infrastructure and losing it has costs. At this point I gave my final report to the GCSA and CMO and returned to other government work.
The app worked. It was downloaded by a substantial proportion of the eligible population. It delivered notifications that interrupted transmission chains that would otherwise have gone undetected. It also failed in ways that mattered. The governance disruption of summer 2020 was avoidable. The time lost between the Isle of Wight trial and the pivot to decentralisation was avoidable. The public confusion generated by the centralised versus decentralised debate was probably manageable with better communication.
The summer of 2021 produced the crisis the programme had always carried as a latent risk. The Delta variant was highly transmissible. The app was functioning as designed, generating isolation notifications in large numbers. The country, by that point, could not absorb the economic and operational consequences of the scale of isolation being recommended. The sensitivity settings were adjusted. The decision was understandable. It illustrated the tension between epidemiological optimisation and economic reality. The app was eventually decommissioned and its functions absorbed into the broader NHS app.
Proximity detection at population scale is a genuine capability. The pandemic demonstrated both what it can do and the conditions under which it fails. The technology itself is not the limiting factor. The limiting factors are governance, public trust, infrastructure ownership, and preparedness.
Copyright: © 2026 [author(s)]. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in Frontiers Policy Labs is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
