For a mid-sized company, a single investment can be worth up to a third of its invested capital. Yet it is almost always decided on a single number, an IRR or a payback, calculated in a spreadsheet and defended in a meeting. And that number is born old: in 2026 the six external assumptions most common in a corporate plan (orders, energy, rates, inflation, exchange rates, investment) left the tolerance band in a median of 3.7 months, according to Vedrai Observatory research.

For a business owner the consequence is concrete. A plan built at year-end describes the real company for a fraction of its horizon, then stops doing so while the capital stays committed. This article offers three questions to bring to the table before you sign.

The question that matters is not how much an investment returns, but with what probability it returns enough, and which variable can make it collapse.

Why a corporate plan expires sooner than you think

Capital decisions are among the most studied in management, and the evidence is harsh: according to the database of more than 16,000 projects built by Bent Flyvbjerg at Oxford, 91.5% of large projects overrun budget, schedule, or both, and only 0.5% meet cost, schedule, and promised benefits together. It is not only a problem of execution or optimism: it is the tool. A typical business case produces a single number from a single scenario, built by projecting the past.

The point is that the past, today, does not last long. The 2026 shock did not hit everyone the same way: according to Banca d'Italia, for the year firms with 20 to 49 employees expect to cut investment by 11.8%, against 1.2% for firms with more than 500 employees, with manufacturing at 9.2%. For someone running a mid-sized firm, an expired plan is not an estimation error: it is a risk to continuity.

What it means to govern the plan as a portfolio of bets

Governing the plan means to stop treating it as a document to approve once a year and to start reading it as a set of bets whose probability of success you know. This is the heart of Decision Intelligence, the category recognized by Gartner and Forrester that connects data to decisions: no longer "what does it return", but with what probability each move clears the threshold, which variable puts it at risk, and when it is worth reopening the choice.

The opposite is inertia. The McKinsey study of multi-business firms between 1990 and 2005 found, for a third of business units, a correlation of 0.99 between one year's allocation and the previous year's: the plan changes, the budget returns to where it was. Firms that instead reallocate capital dynamically generated shareholder returns about 30% higher per year. Those who govern the bets are worth more over time.

The 3 questions to ask before you commit capital

Three questions turn a plan from an estimate to defend into a decision to govern. They apply to every major investment: an acquisition, a new line, entry into a market, the sale of a division.

1. With what probability does it clear the threshold?

Two projects with the same expected IRR of 15% are not the same decision. Letting the assumptions vary according to realistic distributions, one may clear the approval threshold in nine scenarios out of ten, another in six, and when it fails it often does not recover the capital. The expected value is identical, the risk is not. The metric to ask for is not the point return, but the probability of clearing the threshold: "it passes in eight scenarios out of ten" says more than "it returns 15%".

2. Which external variable can make it collapse, and when?

Every plan rests on a few critical variables: the price of energy, the cost of debt, demand. Before approving, it is worth setting for each the band beyond which the decision reopens, and watching it. In 2026 the drop in capital goods orders was readable as early as January 30, months before official data certified the collapse in investment intentions. Review should happen by event, not by calendar.

3. Do the bets hold together, or share the same risk?

A business owner does not choose a project: they choose a combination, within a financial constraint. This is where the point estimate becomes dangerous, because risks do not add up, they correlate. In 2026 energy, inflation, and rates moved together: one cause, three plan variables in the same direction. Funding everything at once can look sustainable on paper and not be so in practice.

An example: the mid-sized firm that funds three moves at once

Imagine a company with leverage just above 2 times EBITDA and a bank covenant at 3.25 times, which decides to finance an acquisition, a new line, and a new market all at once. Under the year-end assumptions the "all-in" plan looks sustainable, with a risk of breaching the covenant close to zero.

But the swings in energy, rates, and demand observed in a single quarter are enough to move that risk into a band of 3 to 4 chances in 10, without a single cell of the business case having changed. A more prudent combination, which funds growth with a divestiture and defers the more uncertain bet, keeps a comparable expected value with a covenant risk that stays below 5%. The difference lies not in the projects, but in seeing them as a portfolio.

How to move from estimate to probability, without redoing everything

Scenario simulation is not new: it has existed for decades. What has changed is the ability to make it the ordinary way of deciding, even beyond large corporations. A Decision Intelligence system connects the plan to the external variables that make it expire, estimates from data how much they fluctuate and how they move together, propagates the shock along the KPI tree to margin, cash, and covenant, and watches the thresholds to reopen the decision when needed. In forecasting, McKinsey estimates that AI models reduce forecast error by 20 to 50% versus traditional methods.

The direction is already set: according to the Protiviti 2026 survey, AI adoption for financial forecasting rose in a year from 58% to 76%. Control stays human: the system makes explicit what each alternative costs, the owner decides. This is exactly the problem that WhAI, Vedrai's Decision Intelligence platform, is built to solve.

Key takeaways

• In 2026 the external assumptions of a corporate plan lasted a median of 3.7 months: a 36-month plan stayed valid for about 10% of its horizon.
• 91.5% of large projects overrun budget or schedule; the first cause is the tool, one number from one scenario.
• Mid-sized firms will cut investment by 11.8% in 2026, against 1.2% for large ones: the margin for error is tighter.
• Three questions before you commit capital: probability of clearing the threshold, the variable that can make it collapse, the correlation between the bets.
• Project risks correlate: an "all-in" plan can move from near zero to 3-4 chances in 10 of breaching the covenant without changing a single figure.

FAQ

What is a corporate plan and why does it expire?

A corporate plan is the document with which a company schedules investment, resources, and results over several years. It expires when the external assumptions it rests on (demand, prices, rates) leave the tolerance band set at approval: in 2026 this happened in a median of 3.7 months.

How often should a corporate plan be updated?

Not by calendar, but by event. It is best to set a review threshold for each critical variable and reopen the decision when it is crossed, rather than waiting for the annual budget. In a volatile context, a once-a-year review leaves capital allocated against assumptions that are already out of date for most of the time.

What is the difference between a point estimate and a probability in capital decisions?

A point estimate gives a single number (e.g. IRR 15%) and treats the future as known. A probability declares the uncertainty: it says in how many scenarios out of ten the investment clears the threshold and which variables make it change. At the same expected value, two projects can have very different probabilities of success.

What is Decision Intelligence applied to the corporate plan?

It is the approach, recognized by Gartner and Forrester, that connects data to decisions: it simulates scenarios, propagates external shocks through to margin, cash, and covenant, optimizes combinations of projects under constraints, and updates the assessment when the context changes.

Can an SME simulate scenarios without a data science team?

Yes. There are now Decision Intelligence platforms that make simulation accessible to mid-sized firms too, without hand-built models: they connect the external variables, estimate the distributions from data, and present the probabilities in a form the owner and the CFO can read.

Download the full report

This article summarizes a broader piece of Vedrai Observatory research. For the full data, tables, and methodology, download the report Governing capital under uncertainty

‍