The question is not whether you could have produced the numbers. It is whether you can tell that they cannot be true.
A quarterly progress review on a defence contract. The supplier's cost control specialist works through a sequence of indices and acronyms. Performance to date is running at roughly seventy per cent against both cost and schedule. The recovery plan targets ninety-two per cent by completion. Everyone nods. The presentation moves on.
One person in the room does the arithmetic. Given how much of the budget has already been consumed at seventy per cent efficiency, reaching ninety-two per cent overall would require every remaining hour of work to be performed at a rate the supplier has never come close to achieving — and, depending on how far through the contract it is, possibly at a rate above what is physically possible. The recovery plan is not ambitious. It is arithmetically unavailable.
This account comes from John P. Sahlin, writing in PM Network in 1998, and it is his recollection rather than a documented case. [FACT CHECK REQUIRED] — the underlying earned-value logic, in which a cumulative performance index can only be lifted by future work performed at an efficiency the remaining budget must actually support, is standard and can be verified independently; the specific figures are his. But whether or not that particular meeting happened as described, every executive reading this has sat in its equivalent.
The uncomfortable question is not whether the presenter was being dishonest. It is whether anyone in the room was equipped to find out.
The Strategic Context
Senior leaders are asked, continuously, to approve continuation. Another tranche of funding. Another six months. A recovery plan for a program that has already missed twice. In each case the decision rests on a claim about the future made by people with more information than the decision-maker and, frequently, an interest in the answer.
The conventional defences are structural: independent assurance, a project management office, a gated review process, external audit. These are worth having. But they share a limitation — each of them substitutes another party's judgement for the executive's, and each can itself be presented with numbers it does not test. Assurance functions get snowed too.
There is a second defence, less discussed and much cheaper, which is the decision-maker's own capacity to falsify a claim. Not to originate the analysis. Not to know the discipline. Simply to hold enough of the underlying logic to ask one question that the claim must survive.
This is a governance property, not a personal accomplishment, and most organisations have never treated it as something they could deliberately build.
What Leaders Commonly Misread
The first misreading is that this is an argument for technical executives. It is not, and the distinction matters. Sahlin's own position — he had no technical degree — was that a project manager does not need intensive technical training, and that leadership and management skills common to all application areas matter more than deep domain expertise. The claim here is narrower and compatible with his: you do not need to know how the work is done, but you need to hold enough of the structure of the claim to test whether it hangs together. That is a different and much smaller body of knowledge.
The distinction between the technical and managerial content of a leadership role is treated at length in [Related article: From Specialist to Delivery Leader: The Promotion That Is Actually a Career Change]. This article assumes it and moves on.
The second misreading is that the risk is deception. Deliberate misrepresentation exists, but it is rarer than the more ordinary failure: a team that genuinely believes its recovery plan, has never tested it against the arithmetic, and is presenting an aspiration in the grammar of a forecast. Sahlin's own instinct in the moment was that the contractor's engineer had not previously understood what the figures implied — which is a far more common condition than fraud, and a far more dangerous one, because nobody in the room has any reason to look harder.
The third misreading is that more reporting protects you. It usually does the opposite. Volume and technical presentation are the two conditions under which an unexamined claim travels furthest. A twelve-page pack with four indices and a heat map is harder to interrogate than a one-page statement of what is being asserted and what would have to be true for it to hold. This is distinct from — though it compounds with — the way reporting layers compress and discard information as it travels upward, which is examined in [Related article: Whose Knowledge Does Your Governance System Actually Hear?]. Here the information arrives intact. The problem is that nobody tests it.
Reframing the Issue
The reframing is to stop asking do I understand this? and start asking what would have to be true for this to be right?
Understanding is unbounded and intimidating; an executive can always be told they lack the background. Falsifiability is bounded and specific. It requires only that the decision-maker can identify the claim's load-bearing assumption and ask whether it is available.
Applied to the opening example, the load-bearing assumption is future productivity. The plan asserts a level of future performance; the arithmetic tells you what that level must be; the record tells you whether it has ever been achieved. Three sentences, no modelling, and the recovery plan either survives or it does not.
This generalises well beyond earned value. A demand forecast rests on a conversion assumption. A synergy case rests on an attrition or retention assumption. A clinical business case rests on a throughput assumption. In every instance, the number presented is downstream of a rate that can be compared against something the organisation has actually done. Executives who ask that question consistently are not more numerate than their peers. They have simply learned where to point.
Strategic Analysis
Four working rules, and what they are for
The practitioner literature offers a compact set of rules for leaders operating above their own technical depth. Read as personal advice they are unremarkable. Read as governance norms they are considerably more interesting.
Never tell the team how. Setting what and when, and leaving how to the specialists, is usually justified as respect for expertise. Its harder value is informational: a leader who has prescribed the method has an interest in it working, and will not test the claim that it is working. Prescribing how contaminates the assurance you will later need.
Know what you do not know, and know where the knowledge is. The useful executive skill is not domain knowledge but an accurate map of your own ignorance and a short list of people who can close it — with the availability and reliability of those sources considered in advance, not at the point of crisis.
"I don't know" is acceptable — once. As a personal position this preserves credibility. As an institutional norm it does something more valuable: it makes the admission of uncertainty low-cost, and an organisation in which uncertainty is expensive to admit is an organisation whose reporting is systematically optimistic. The corollary matters as much as the rule — never answer the same question that way twice.
Learn enough not to be snowed. This is the operative one. Sahlin's own framing is that people conceal faults with technical jargon and statistics, and that a leader needs enough command of the numbers to work through them to their real meaning. Note what is not being asked for: not the ability to produce the analysis, only the ability to detect that it does not close.
The economics of one question
There is a portfolio argument here that is worth making carefully, because the evidence is reasoning rather than data. A continuation decision on a troubled program is typically the largest single discretionary allocation an executive makes in a given month, and it is the decision most likely to be made on an untested claim. If a single falsification question changes even a small proportion of those decisions, the return dwarfs anything achieved by improving the reporting pack.
Which is why this belongs in the conversation about what an organisation is buying when it funds delivery capability — a conversation examined in [Related article: What Does the Enterprise Own After a Capability Investment?]. Numeracy of this specific, narrow kind is among the cheapest capability investments available, and among the least often made deliberately.
Why the instruments make this harder than it should be
Earned value, critical path, and the rest of the standard performance apparatus were built for a particular kind of work, and they carry that inheritance into contexts where the underlying assumptions no longer hold — a problem taken up in [Related article: What Kind of Work Were These Instruments Built For?]. One practical consequence belongs here: instruments designed for stable scope tend to express performance as a ratio against a baseline, and a ratio against a baseline that has been rebaselined three times is not a measurement of anything. Before testing whether a recovery is arithmetically possible, it is worth asking what the denominator is.
Decision Framework
A four-question test, applicable to any forward-looking claim presented for approval.
1. What is the claim, stated as a single sentence with a number and a date? If the presenter cannot produce this, the pack is describing activity rather than asserting an outcome, and there is nothing to test.
2. What rate does the claim require, and what rate have we achieved? This is the falsification question. It converts a forecast into a comparison against the record. Most unsound claims fail here, in public, in under a minute.
3. What would have to change to make the required rate available? A credible recovery names a specific change — different people, reduced scope, a removed dependency — and the change is verifiable. An incredible one names effort, focus or commitment.
4. If this claim is wrong, when will we find out, and what will it have cost by then? The answer determines how much interrogation the claim deserves. A wrong claim discovered in three weeks at low cost warrants less scrutiny than a wrong claim discovered in nine months after two further tranches.
Two supporting conventions make the test work. Require the assumption on the page — every forward claim states the rate it depends on. And separate the presenter from the assurer: the party asserting the recovery should not be the only party that has tested it.
From Strategy to Execution
Immediate. At the next investment committee or program board, apply question two to whichever claim carries the largest financial consequence. Do it once, in the room, without warning. The result tells you a great deal about the standard of evidence currently being accepted, and it will change the next pack more effectively than any policy.
Medium term. Change the template rather than the training. Add one required field to every continuation paper: this forecast assumes a rate of ___; our achieved rate over the last two periods was ___. A single line, mandatory, does more than a course, because it makes the omission visible when it is missing.
Long term. Build the capability deliberately in the small number of people who sit on continuation decisions. This is not a general numeracy program; it is a short, specific curriculum on the half-dozen instruments that actually carry consequence in your organisation, taught as how these can mislead rather than how these work. It is inexpensive, and almost nobody does it.
Signals to Monitor
- Recovery plans that name effort rather than change. Language such as increased focus, renewed commitment or management attention in place of a specific structural change is the most reliable single indicator of an untested forecast.
- Rising presentation sophistication as performance deteriorates. More indices, more charts and more acronyms appearing in the pack as the news worsens is a pattern worth naming explicitly when you see it.
- Repeated rebaselining. Each rebaseline resets the ratio and erases the record. Three in a program's life means the performance indices no longer measure performance.
- Forecasts that improve without an intervening event. A projection that recovers between two reporting periods with nothing having changed is an adjustment to the forecast, not to the work.
- Silence after the falsification question. How a room responds the first time someone asks what rate is required is diagnostic of the whole governance culture, and worth observing deliberately.
Questions for the Leadership Team
- On our three largest in-flight investments, what rate of future performance does each business case assume, and how does that compare with what we have actually achieved?
- When did this leadership team last decline a recovery plan on the ground that it was not arithmetically available?
- Who in our governance chain is expected to test a claim rather than receive it — and do they have the standing to say no?
- What is the cost, in this organisation, of a delivery leader saying they do not know?
- Which of our performance measures have been rebaselined more than once, and do we still believe what they report?
- If a supplier presented us with the numbers in the opening of this article, what would happen?
Closing Perspective
Executive judgement is usually discussed as a matter of experience, instinct or strategic breadth. Those things matter. But a substantial share of the decisions that damage an enterprise are not failures of judgement at all. They are failures of interrogation — approvals given to claims that would not have survived one properly aimed question.
The capability required to ask that question is small, specific and learnable. It does not require an executive to become a specialist, and it does not require a larger assurance function. It requires knowing that every forward-looking claim rests on a rate, that rates can be compared against a record, and that a claim which fails that comparison should be refused however confidently it is presented.
The organisations that lose most to this are rarely the ones that were deceived. They are the ones where everyone in the room assumed someone else had checked.
About the author
Kevin Jogin is Founder & Principal Advisor at EraNorth. Meet the Founder.