86% Lower Modelled Overflow Volume—Yet 62.5% of Samples Crossed a Health-Risk Threshold: Is the £4.6bn Super Sewer Measuring the Wrong Benefit?

An 86% success and a 62.5% warning can both be true

A project controls leader should be able to hold two apparently conflicting numbers in the same board pack without forcing either one to disappear. Tideway’s first annual Benefits Realisation Report, published on 12 October 2026, says modelled annual storm-overflow volume into the tidal Thames fell by approximately 86%: from a pre-tunnel average above 21 million cubic metres to just under three million cubic metres in 2025. The report also estimates that the tunnel prevented 84 tonnes of sewage-derived litter from entering the river and reduced days associated with elevated health risk from an average 75.7% to 10.4%. These are substantial first-year results. The asset is 25 kilometres long; Thames Water describes the delivered project cost as £4.6 billion. That is a project cost, not annual expenditure, and it should not be confused with the additional £2.5 billion that Thames Water says it invested across all three phases of the wider London Tideway improvements scheme.

Four days later, the Financial Times reported that Fulham Reach Boat Club’s weekly testing found 62.5% of samples taken between February 2025 and October 2026 exceeded 1,000 coliform unit equivalents of E. coli per 100ml, the health-risk threshold used in that reporting. The FT also reported that Thames Water had not conducted bacterial or pathogen testing after the tunnel became fully operational and offered no guidance on recreational safety at that location. This is not proof that the tunnel failed, nor is the club series a river-wide random sample. The tunnel targets combined sewer overflows in a complex tidal system; bacteria at one reach may reflect upstream discharges, tributaries, rainfall, tides and other inputs. Conversely, an 86% reduction in modelled overflow volume is not direct proof that every river user now encounters safe water. The figures describe different populations, periods, methods and positions in a causal chain. Their tension is precisely why benefits realisation belongs in project controls after construction ends.

The asset is an output; a healthier river is a system outcome

The official report is more careful than the headline debate suggests. Its 86% figure combines calculated pumped volumes with modelled gravity-overflow volumes, and the report explicitly notes that 2025 was unusually dry. Its water-quality assessment uses dissolved oxygen data from two Environment Agency automated stations; those instruments are serviced on a best-endeavours basis and are not MCERTS-certified. It recorded one sewage-linked dissolved-oxygen depletion event lasting five days, while another depletion event was attributed to heat and low flow rather than sewage. The litter estimate applies a proxy rate measured at Abbey Mills to sewage intercepted elsewhere. These are defensible methods when direct measurement is impracticable, but they are models and proxies, not an all-purpose certificate of river health. The report commits Thames Water to annual publication for ten years, an important recognition that a first operational year—still within commissioning—cannot close the benefits case.

This distinction matters far beyond water. A railway can open while passenger benefits remain uncertain; a digital system can go live while adoption stalls; a hospital can be built while pathways and staffing limit its effect. Output control asks whether the specified asset was delivered safely, legally, to the authorised scope, cost and schedule. Benefits control asks whether the asset changes the intended condition for the intended beneficiaries, by how much, over what period and compared with what would otherwise have happened. In the Thames case, overflow capture is an intermediate outcome strongly related to the tunnel; ecological resilience and recreational safety are end outcomes influenced by the wider river system. A good board should celebrate movement in the first without claiming sole control over the second. It should also refuse to use system complexity as an excuse not to measure the public result.

New UK evidence: ownership thins out after handover

A new open-access empirical paper provides a sharp explanation for this governance gap. Sarah Coleman and Richard Kirkham’s The role of leadership in ex-post project benefits realisation, published in Project Leadership and Society in 2026, examined documents and 47 semi-structured interviews across eight UK government major projects. Participants included project directors, sponsors, benefits leads, operational managers, commercial and finance managers, analysts and asset users. The authors found that leadership for benefits after delivery was distributed across project, commercial, operational, sponsorship, benefits and change functions; formal processes were applied inconsistently, and reviewing and exploiting benefits were less mature than planning and delivering them. Most importantly for infrastructure owners, benefits did not simply emerge from the asset. Data quality, organisational systems, time horizons, operational capacity and senior attention shaped whether anyone could establish what happened later.

The study is qualitative and purposively sampled; it reveals experienced practitioners’ interpretations rather than a statistically representative failure rate, and it does not evaluate Tideway. Its transfer is nevertheless unusually direct because it examines British government major projects and the boundary between temporary delivery organisations and permanent operations. A second 2026 study by Morten Welde synthesised standardised ex-post evaluations of 14 Norwegian road projects. It found that operational delivery could be acceptable while wider goals remained weakly formulated or lacked clear ownership, and that ex-post benefit-cost results changed substantially when assumptions and demand differed. The sample is small, sector-specific and not statistically representative, but the lesson converges: finishing the asset is not the final act of control; it is the point at which forecasts begin meeting observable reality.

Build the benefits control account before the project team leaves

For Tideway, Thames Water, regulators and river authorities, I would create one public benefits control account with four linked levels. Level one records asset availability, operating constraints, maintenance and accepted scope. Level two records causal operational measures—intercepted volume, residual spills, rainfall, tunnel closures and treatment capacity—stating what is monitored, calculated or modelled. Level three records environmental response through dissolved oxygen, bacteria, litter, ecology and tributary conditions, with locations, sampling frequency, laboratory method and uncertainty. Level four records beneficiary outcomes: safe-use guidance, reported incidents, river-user confidence and amenity. Each measure needs a baseline, denominator, data owner, publication date, target or decision threshold and explanation of what the tunnel can and cannot cause. Independent community data should be reconciled, not dismissed; official data should be questioned, not presumed wrong. Where evidence conflicts, publish the competing explanations and the test that would distinguish them.

Accountability must survive handover. Tideway should retain responsibility for evidence about design intent and asset performance; Thames Water should own operation and the ten-year benefits series; environmental and public-health bodies should own the measures within their mandates; and one named senior benefits owner should integrate the story for decision-makers and the public. Budget the monitoring, data stewardship and independent review as part of the investment, rather than treating them as discretionary aftercare. NISTA’s Gate Review 5 guidance, updated on 20 August 2026, asks whether long-term evaluation, benefits ownership, operating arrangements and evidence align after delivery. Project directors, heads of controls and council leaders can apply the same discipline now: select the three outcomes that justified the investment, preserve the counterfactual and assumptions, name the post-handover owner, and schedule the first two evaluation decisions before demobilisation.

The optimistic conclusion is not that £4.6 billion has bought the wrong thing. The first-year evidence strongly indicates that the tunnel is intercepting a vast volume of pollution, while the independent samples warn that river health and human exposure remain larger system challenges. That combination is valuable intelligence. Britain can treat it as an argument about whether the project succeeded, or as the beginning of a mature learning cycle that protects benefits across the asset’s life. The better course is clear: keep the impressive 86%, publish the uncomfortable 62.5%, explain their limits, close the missing-data gaps and make someone accountable for the next decision. When boards measure the whole chain from concrete to lived outcome, great engineering becomes durable public value—and successful projects become even more credible.