What the numbers are supposed to mean
Earned Value Management — EVM — compares three numbers. Planned Value (PV) is what you expected to have spent by now, according to the baseline. Earned Value (EV) is what you have actually achieved, valued at baseline rates. Actual Cost (ACWP, or AC) is what you have spent. From these three, you derive the Cost Performance Index (CPI = EV/AC) and the Schedule Performance Index (SPI = EV/PV). A CPI above 1.0 means you are spending less than you have earned; an SPI above 1.0 means you are ahead of schedule.
The technique is sixty years old and deeply embedded in how major projects report upward. Government programmes, defence contracts (notably MoD CADMID Demonstration and Manufacture-phase contracts where EVMS is mandated in the contract data), and most large infrastructure jobs above a certain value have EVM requirements written into them. The appeal is that a healthy CPI and SPI — say, 0.98 and 1.02 — can be presented to a steering group in two lines and suggests the project is under control. The problem is that a project can be quietly failing and still produce those two lines, because each of the three inputs has structural weaknesses that a careful project can exploit and an incompetent one can stumble into.
This guide is about the specific failure modes of earned value in UK construction and infrastructure practice. It is not an argument for abandoning EVM. Well-run EVM on a well-structured contract is one of the most powerful governance tools in project controls. But EVM run against a poorly structured work breakdown, with progress measured by the contractor, and actual cost lagging invoices by three months, can produce numbers that give senior management a completely false impression of project health. Practitioners need to know when the numbers are reliable and when they are being moved around.
Materials on site and the front-loaded curve
One of the most common sources of EV distortion is the treatment of materials on site. On many construction programmes, materials delivered and stored — steel, mechanical plant, cabling, large components — are booked as earned value at the point of delivery rather than at the point of installation. The logic is commercial: the contractor has incurred the cost, the client is paying for the material, so the value has been earned. The consequence is that EV runs ahead of actual physical progress, sometimes significantly.
On a major mechanical and electrical package, it is routine to see 25-35% of the package value delivered as materials on site months before meaningful installation begins. If those deliveries are booked as earned, the SPI looks excellent for the first half of the package. Then installation begins, and the contractor earns very little additional value for a period because the materials have already been credited. The SPI deteriorates sharply even though physical progress is tracking the original plan. The shape of the EV curve becomes front-loaded — a large early spike followed by a plateau — which systematically misrepresents project status at both ends.
The fix is straightforward in principle and unpopular in practice. Materials on site should either be excluded from earned value entirely, with a separate line tracking material deliveries as a procurement metric, or they should be earned at a discounted rate — typically 50-70% of their value — with the remainder earned on installation. AACE Recommended Practice 10S-90 and the APM's earned value guidance both support the split. The reason this is unpopular is that it produces a less flattering EV curve in the early months, and contractors are reluctant to accept a measurement approach that makes them look behind when the commercial position is that they are on track. Client-side controls teams have to insist on it at contract award, not retrofit it after the contractor has been reporting rolled-up EV for six months.
The diagnostic question is: what percentage of earned value to date represents physical installation versus procurement? If more than 30% of total earned value on a live construction package is materials on site, the SPI is almost certainly overstating progress. Ask for the split and, if it is not available, treat the SPI with caution.
Progress measurement gamed at the package level
Earned value depends on an honest measurement of physical progress. On any programme of material size, that measurement is done by the contractor — they complete the progress claim, they estimate percent complete, they submit for payment. The client-side controls team typically reviews the claim, but review is not independent measurement. Unless the client has quantity surveyors walking every package every month, the progress figures are the contractor's figures with a rubber stamp.
The two most common gaming patterns are gradual overstatement and endpoint acceleration. Gradual overstatement is where every package runs at a steady 5-10% ahead of genuine progress — not enough to trigger alarm, but enough to keep the SPI comfortably above 0.95 throughout. The overstatement is only discovered near completion, when the contractor cannot close out the final activities because they were never actually as far advanced as reported. By then the cash has been paid, the programme has lost its float, and the client has no commercial leverage to recover. Endpoint acceleration is the opposite pattern: progress is reported genuinely for most of the package, and then jumps sharply in the final month to avoid triggering a late completion. The final jump represents work the contractor claims but has not done, and the closeout stretches on for months after the headline completion date.
Neither pattern can be detected from the EV data alone. Both require physical verification — walking the site, comparing the claim against the work in place, spot-checking specific activities. On a well-run project, the client's quantity surveyor or construction manager will independently verify a sample of claimed progress each month, and the controls team will compare the verified sample against the claim to flag discrepancies. On a less well-run project, the controls team receives the claim, processes it, and produces the CPI and SPI without anyone physically checking.
The third pattern is quieter and harder to detect: progress booked against tasks that are out of sequence. A contractor can achieve the headline earned value by front-loading the easy, high-value tasks and deferring the hard ones. The SPI looks healthy because the earned value is accruing at the planned rate, but the unearned balance is increasingly concentrated in the difficult work. When the easy work runs out, productivity collapses because the remaining scope is harder than the average assumed in the baseline. SCL Protocol 2nd edition addresses this as "opportunistic progress claiming" and notes that the only reliable defence is tracking progress by physical sequence, not just by cumulative percentage.
ACWP lagging behind actual commitment
The Actual Cost of Work Performed — ACWP, or AC — is meant to be the real money spent to achieve the earned value. In practice, it is usually the money paid, which is not the same thing. The lag between work done, invoice received, invoice approved, and payment issued can be three to four months on a typical UK construction job — sometimes longer under certain JCT payment regimes. This lag produces a structural understatement of actual cost in the reported CPI.
The consequence is that the CPI looks better than reality for the opening months of a package, because earned value is accruing but the matching cost has not yet appeared in the accounts. Then, when the invoices catch up, the CPI drops sharply. Controls teams who interpret this as a sudden deterioration in productivity are misreading the signal; the productivity was always this level, but the cost system was behind the work. A CPI that starts at 1.15 and drifts down to 0.95 over six months has not deteriorated by 20% — it was never at 1.15 in the first place.
The fix is to report accrued cost rather than paid cost. Accrued cost includes work in place that has been done but not yet invoiced, based on the contractor's own reporting of incurred cost. Most financial accounting systems support accrual-based reporting, but many controls functions do not receive the accrual data in a timely way because the finance team runs accruals as a month-end exercise and the controls report is produced mid-month. The controls team has to reconcile the gap manually, or the CPI will be systematically optimistic until month-end catches up.
The less tractable version of the problem is on lump sum or target cost contracts where the actual cost to the client is the contract value regardless of the contractor's actual expenditure. The ACWP line becomes an artefact of contract valuation rather than a measurement of cost performance, and the CPI loses meaning as a productivity indicator. On such contracts, EVM is better run on internal cost rather than contract value — measuring whether the contractor is burning through their own cost forecast — but this requires access to cost data the contractor may not share.
Schedule relationships that the SPI ignores
The SPI is a cost-based schedule metric. It tells you whether the cumulative value earned is tracking the planned cumulative value over time. What it does not tell you is whether the earned value is on the critical path. A project that is earning plenty of value but doing it on non-critical work — because the critical activities are stuck waiting for design approval, or for a permit, or for a specific subcontractor — can produce an SPI of 1.0 while steadily losing float on the actual completion date.
This is the structural weakness of the SPI. It rewards activity, not progress toward the milestone. On a complex programme with dozens of parallel workstreams, the SPI can be comfortable for months while the critical path silently slips. The contractor looks productive, the steering group sees a healthy schedule metric, and then two months before the planned completion date the critical path activities finally begin and the whole project starts reporting red. The contractor has not suddenly lost control; they lost control six months ago, and the SPI failed to show it because it was measuring the wrong thing.
The diagnostic is to read the SPI alongside the critical path float trend. If total float on the critical path is falling month on month, the project is losing schedule resilience regardless of what the SPI shows. Some organisations use a schedule metric called SPI(t) — SPI time-based, calculated using the earned schedule method — which addresses part of the problem by expressing progress in time units rather than value units. SPI(t) behaves better than cost-based SPI near the end of a project (cost-based SPI mathematically converges to 1.0 at completion regardless of delay, which is why late projects often finish with an apparently acceptable SPI). But even SPI(t) does not fix the underlying issue that cumulative value is not the same as critical path progress.
The rule of thumb is that the SPI is a useful early-warning metric when read alongside the programme, not in place of it. A project with a deteriorating SPI and a shrinking critical path float is in trouble. A project with a healthy SPI and a shrinking critical path float is also in trouble, but the trouble is hidden. Controls leads who report only the SPI without the float trend are not doing the job the steering group needs done.
Common failure modes on real programmes
The most common failure is the one above — reporting headline CPI and SPI in isolation, without the diagnostics that reveal whether the numbers are meaningful. A steering group that receives "CPI 0.98, SPI 1.02" in a monthly pack and does not see the progress verification sample, the material-on-site percentage, the ACWP accrual adjustment, and the critical path float trend is being given assurance that is not justified by the underlying data.
The second failure is allowing the work breakdown structure to drift. EVM requires a stable cost-loaded baseline that maps to the schedule and the WBS. On a long-running programme, scope changes, package re-lets, and contract variations all disturb the baseline. If the baseline is re-set every time a variation occurs — "rebaselining" — the CPI resets to 1.0 and the history of cost performance is lost. Some programmes re-baseline quarterly as a matter of course, which makes trend analysis impossible and means the CPI is always close to 1.0 by construction. A well-run EVM system maintains the original baseline and records variations separately, so that the genuine cost performance remains visible.
The third failure is running EVM on a contract structure that cannot support it. Some contracts are simply unsuitable for EVM — short-duration works, highly variable scope, research and design contracts where the work content is defined as it proceeds. Forcing EVM onto such contracts produces numbers that look like EVM outputs but do not represent anything meaningful. The controls team ends up defending the methodology rather than using it. Better to accept that EVM is not the right tool for that contract and use a simpler cost-forecast-versus-actual approach supplemented by progress narrative.
The fourth failure is sensitive to UK practice specifically: treating the EVM data as contractually privileged rather than a governance tool. Some contractors treat their cost data as commercial-in-confidence and will not share the granular information needed to validate CPI. The client's controls function then has to operate on rolled-up figures and cannot audit the detail. This is a contract structure problem, not an EVM problem, and it has to be fixed at contract award — with explicit obligations on the contractor to provide cost-loaded schedules, monthly cost detail, and access to progress verification. Retrofitting transparency after contract award is almost impossible.
When to abandon EV and what to use instead
Earned value is the wrong tool in several specific situations. Short projects — under twelve months of delivery — rarely generate enough data points for CPI and SPI trends to be meaningful. High-uncertainty scope — research, prototype engineering, contaminated land remediation — produces a baseline that cannot support EVM because the scope keeps evolving. Heavily phased programmes — a framework of sequential small packages — are better reported as package-level performance than aggregated EVM. For these, use simpler methods: cost forecast versus actual, package gate reviews, milestone reporting, or a rolling-wave approach that only baselines detail for the next two to three months.
Where EVM is the right tool but the current implementation is unreliable, the path to recovery is not to abandon the method but to restore the disciplines it requires. Rebuild the cost-loaded schedule. Separate materials on site from earned value. Set up accrual-based ACWP reporting. Introduce independent progress verification on a sample basis. Report SPI and CPI alongside the critical path float trend and the material-on-site percentage. Each of these fixes is unglamorous and takes months to embed, but the end state is a reporting system that tells the steering group the truth.
The replacement question — "what do I report if I am not reporting CPI and SPI?" — is not as hard as it sounds. A monthly report with three numbers does most of what the steering group needs: cost-to-complete forecast against the approved budget, planned completion date against the contractual completion date, and total float on the critical path. These three numbers, tracked month on month, tell the same story the CPI and SPI are meant to tell, without the structural weaknesses. Add a narrative on the top three risks and the top three issues, and the steering group has a report they can act on.
The practical point is that EVM is a technique, not a goal. If the technique is producing misleading numbers, the right response is to fix the technique or to use a different one — not to keep reporting CPI and SPI because that is what the template expects. Controls leads who recognise when their EVM is lying, and have the confidence to say so, earn credibility that no amount of well-produced S-curve reporting can buy.