Forecast accuracy tells treasury how wrong a number was. Forecast uncertainty tells it how much liquidity may need to be held, borrowed or kept available before the number is known.
For years, the standard treasury question was straightforward: how close was the cash forecast to the eventual outcome? That remains important. But a point forecast can look respectable while still forcing expensive decisions. A company may expect to end a month with $100 million of available cash, yet the operational reality may be that the outcome could plausibly range from $70 million to $130 million. The midpoint is useful. The width of the range determines how much protection treasury may need.
That is the emerging logic behind measuring the cost of liquidity forecast uncertainty. The phrase is used here as an analytical framework rather than a formal accounting or regulatory metric. It refers to the financial and strategic costs created when treasury cannot confidently narrow the range of possible near-term cash outcomes: excess cash held as insurance, committed borrowing capacity kept unused, last-minute funding, conservative investment decisions, delayed capital deployment and management time spent responding to avoidable surprises.
The challenge is not marginal. The AFP — 2025 Treasury Benchmarking Survey Report reports that nearly three-quarters of treasury practitioners cite cash management and forecasting as leading departmental priorities, while more than 60% identify cash or liquidity forecasting as their most challenging task. The EY — Global DNA of the Treasurer Survey adds that 65% of surveyed treasurers did not report highly accurate 12-month forecasts. These findings help explain why the debate is moving from “How accurate is the forecast?” to “What does the remaining uncertainty cost us?”
Key analytical distinction
A forecast error is observed after the cash outcome is known. Forecast uncertainty exists before the outcome is known—and therefore changes how much liquidity, contingent funding and operational flexibility treasury must carry.
Accuracy is not the same as confidence
A forecast can be unbiased and still be difficult to use. If actual cash tends to land near the forecast with only small deviations, treasury can operate with a relatively narrow liquidity buffer. If outcomes are widely dispersed around the same average, treasury needs more room for error even when the forecast is correct on average.
The distinction is familiar in formal forecasting. The Bank of England — Forecast Evaluation Report, January 2026 evaluates forecasts along separate dimensions including accuracy, unbiasedness and efficiency, and notes that forecast errors generally increase at longer horizons because uncertainty rises. The Bank also now places greater emphasis on risks, uncertainty and alternative paths rather than treating a single forecast as the sole representation of the outlook. Corporate treasury faces the same conceptual problem on a different scale: a point estimate is only one input into a liquidity decision.
This is why confidence bands and scenario ranges can become operational rather than statistical concepts. A 13-week cash forecast that is expected to be within a narrow range may justify investing surplus cash for longer. A forecast with the same midpoint but a wide downside range may justify holding more overnight liquidity or preserving borrowing headroom. The economic decision is shaped by the distribution of possible outcomes, not just the expected value.
The first cost is the cash buffer itself
The simplest response to uncertainty is to hold more cash. That can be rational. Cash prevents missed payroll, supplier delays, covenant stress and forced financing at an inconvenient moment. But it is not free. Holding a larger buffer can reduce interest income relative to longer-duration alternatives, create tax and currency inefficiencies across jurisdictions, and compete with debt repayment, capital expenditure, acquisitions or shareholder distributions.
The AFP — 2025 Liquidity Survey illustrates the preference for protection. In the 2025 survey, 61% of organizations named safety as their top short-term investment objective, compared with 35% choosing liquidity and only 5% yield. The result does not prove that forecast uncertainty caused those allocations, but it shows how strongly treasury functions value capital preservation when managing operational cash.
The macroeconomic evidence points in the same direction. The OECD — Reigniting investment for more resilient growth argues that elevated uncertainty has weighed on investment incentives and encouraged precautionary cash accumulation. Federal Reserve research on Federal Reserve — Corporate Cash Accumulation During the COVID-19 Pandemic found that the median cash-to-assets ratio of U.S. publicly traded non-financial corporations rose from roughly 12% in the years after the global financial crisis to more than 20% in mid-2021. That pandemic surge was exceptional, but it demonstrates the balance-sheet scale that precautionary liquidity can reach when uncertainty rises sharply.
The second cost is liquidity insurance
Companies do not have to hold every dollar of protection as cash. They can purchase contingent liquidity through revolving credit facilities, overdraft lines and other committed borrowing arrangements. That flexibility is valuable precisely because it can be drawn when internal cash generation disappoints. It also has an explicit price.
A useful public example appears in Block, Inc. — 2025 Annual Report / Form 10-K. At the end of 2025, Block reported a $775 million senior unsecured revolving credit facility with no funds drawn, while the facility carried a commitment fee of 0.10% to 0.20% per year on the undrawn portion. Its covenant also measured minimum liquidity using unrestricted cash, marketable securities and undrawn revolver availability. The figures are company-specific, not a market benchmark. They nevertheless show the economics of liquidity optionality clearly: even unused capacity can carry a recurring cost.
When forecast uncertainty increases, treasury may need more of this insurance or may be less willing to reduce it. The cost is therefore not just the interest paid after a line is drawn. It includes commitment fees, documentation, covenant management, collateral requirements where relevant and the opportunity cost of reserving borrowing capacity that could support other transactions.
Banks see the other side of the same option
For banks, an undrawn corporate credit line is not idle. It is a contingent liquidity obligation. Recent Federal Reserve research makes the connection unusually clear. The July 2026 paper Federal Reserve — The Last Taxi: LCR Buffers and Bank Liquidity Provision finds that during the acute phase of the COVID-19 shock, banks with larger Liquidity Coverage Ratio buffers above the regulatory minimum provided more credit to firms with large undrawn credit lines. The result supports the idea that bank liquidity buffers help convert committed facilities into usable corporate liquidity when stress arrives.
The pricing and availability of that insurance are not fixed through time. The Federal Reserve — July 2026 Senior Loan Officer Opinion Survey reported in July 2026 that C&I lending standards were broadly unchanged in the second quarter, while a moderate net share of banks eased the maximum size and costs of credit lines to large and middle-market firms. Conditions can move in the opposite direction when lenders become more cautious. A treasury team that relies on refinancing or adding capacity only after uncertainty materialises may therefore discover that the option is most expensive when it is most needed.
A separate January 2026 Federal Reserve analysis of Federal Reserve — Supply Chain Risk and Bank Lending Amid Trade Policy Uncertainty found that banks more exposed to supply-chain risk reduced C&I loan commitments during part of the 2024–25 period of elevated trade-policy uncertainty. The episode reinforces the interaction between corporate forecast risk and bank credit supply: uncertainty can increase a company’s desire for liquidity at the same time that lenders reassess how much contingent capacity they want to provide.
The third cost is bad timing
Forecast uncertainty can also create transaction costs without producing a formal liquidity shortfall. A company that discovers too late that it has more cash than expected may invest it for only a few days instead of several weeks. A company that discovers a shortfall late may borrow overnight, break deposits early, move funds across entities, unwind hedges or accelerate intercompany transfers. Each action can be individually small. Repeated across currencies, subsidiaries and business units, the friction becomes material.
The cost of bad timing is especially visible around debt maturities, tax dates, payroll, dividends and large supplier payments. The same forecast miss can have a very different economic impact depending on when it occurs. Being $20 million short on an ordinary Tuesday is not equivalent to being $20 million short on a day when debt service, payroll and a major acquisition payment coincide.
That is why a single monthly accuracy percentage can be misleading. Treasury needs to know where errors occur in the calendar, how they correlate with committed outflows and whether errors cluster in the same business units or cash-flow categories. Uncertainty becomes expensive when it is concentrated around obligations that cannot easily be delayed.
Manual data collection keeps the uncertainty range wide
Many organizations still face a basic information problem. The PwC — 2025 Global Treasury Survey found that 38% of companies with more than $10 billion in revenue and 52% of companies between $1 billion and $10 billion still manually collect and consolidate forecasting data. Respondents identified poor data quality, ineffective tools and limited incentives for business units to contribute as common obstacles.
This matters because forecast uncertainty is often organizational before it is statistical. Treasury may have a sophisticated model but still receive late procurement data, optimistic sales assumptions, incomplete tax estimates or inconsistent intercompany information. A machine-learning model cannot fully solve uncertainty that originates in missing ownership, weak data lineage or incentives that reward business units for forecasting what they hope will happen rather than what is most likely to happen.
The EY survey provides a useful caution on technology. Treasurers that reported very frequent use of AI or machine learning for forecasting also reported higher accuracy than the full sample. But the same research highlights data quality and systems integration as persistent barriers. The inference is not that AI automatically narrows liquidity risk; it is that better modelling has more value when the underlying cash-flow inputs are timely, comparable and governed.
From a point forecast to a cost curve
A mature treasury process can translate uncertainty into a decision problem. Instead of asking only whether the forecast is $100 million, it can ask how much downside is plausible, what probability is attached to that downside and what each additional dollar of protection costs. The optimal buffer is where the marginal cost of more protection begins to exceed the expected cost of being short.
That does not require a false level of mathematical precision. Treasury can use simple scenario bands: base, downside and severe downside; or statistical ranges based on historical forecast dispersion. It can then map each range to operational actions such as cash held overnight, money market allocations, revolver headroom, commercial-paper capacity, intercompany funding and minimum cash by entity.
The result is a form of liquidity-at-risk thinking. Again, there is no universal accounting or regulatory standard called “cost of liquidity forecast uncertainty.” The usefulness lies in making an implicit trade-off explicit. A treasury team already pays for uncertainty when it holds more cash, buys more committed capacity or avoids deploying capital. Measuring the trade-off makes those decisions easier to challenge and explain.
Uncertainty can be more important than average error
Suppose two business units each miss their weekly cash forecast by an average of 5%. One misses by roughly 5% almost every week. The other is usually close but occasionally misses by 25% or 30%. The average error can look similar while the liquidity requirement is completely different. The second unit creates tail risk and may force the group to carry a larger buffer even if its mean accuracy appears acceptable.
This is one reason the distribution matters. The BIS — Annual Economic Report 2025, Chapter I noted in its 2025 Annual Economic Report that uncertain firms had expanded undrawn credit and that heightened uncertainty could weigh on investment. The broader lesson for treasury is that firms often respond to uncertainty by preserving options rather than committing capital. Better forecasts can reduce the need for some of those options, but only if the improvement is visible in the downside distribution rather than just the average.
The counterargument: confidence bands can create false precision
There is a serious limitation to the entire framework. A probability range can look scientific while resting on fragile assumptions. Historical forecast errors may not capture a new acquisition, a product launch, a tax change, a supply-chain disruption or a sudden customer default. Correlations can also break down exactly when liquidity pressure rises.
Treasury should therefore resist the temptation to replace one misleading point estimate with an equally misleading 95% confidence interval. The purpose of uncertainty measurement is not to claim that extreme events have been tamed. It is to make known uncertainty visible, preserve room for unknown uncertainty and distinguish recurring forecasting noise from genuinely discontinuous shocks.
This is also why liquidity management cannot be delegated entirely to a forecasting model. The IFRS Foundation — IFRS 7 Financial Instruments: Disclosures requires entities to describe how they manage liquidity risk and, where relevant, provide maturity information that helps users evaluate that risk. Formal disclosure requirements do not prescribe a cash-forecast uncertainty metric, but they reinforce the underlying governance principle: liquidity risk is about the ability to meet obligations as they fall due, not simply about forecast accuracy as an internal performance statistic.
Evidence versus inference
The evidence is strong that cash and liquidity forecasting remains difficult for many treasury teams; that uncertainty can encourage precautionary cash accumulation and the preservation of credit lines; that undrawn credit facilities have an explicit or implicit cost; and that banks themselves must hold liquidity capacity to honour contingent commitments during stress. It is also well established that forecast dispersion generally grows with the forecast horizon and that scenario analysis can add information that a point estimate does not contain.
The inference is that treasury teams will increasingly combine these pieces into an explicit “cost of uncertainty” framework. As of 1 September 2026, there is no universal treasury standard defining the metric, no common confidence level that organizations should use and no evidence that one formula is optimal across sectors. A utility, retailer, technology company and highly leveraged industrial business face different cash-flow structures and different costs of being wrong.
Implications for banks, fintechs, regulators and investors
For corporate treasury teams, the immediate implication is that forecast performance should be measured in economic terms as well as statistical terms. A smaller forecast error matters most when it allows the organization to reduce idle cash, shorten the period for which expensive credit capacity must be reserved or deploy surplus liquidity more confidently.
For banks, better corporate forecasting can change the pattern of deposits, revolving-credit usage and intraday liquidity demands. Banks that provide treasury services can therefore create value by helping clients connect transaction data, cash positions, payment calendars and credit capacity rather than treating lending and cash management as separate products.
For fintech and treasury-technology providers, the opportunity is broader than predicting a cash number. Useful systems need to show confidence ranges, explain the drivers of widening uncertainty, identify which business-unit inputs create the largest tail risk and connect forecast changes to the cost of funding or investing the resulting cash position.
For regulators and accounting standard-setters, there is no obvious case for prescribing a new corporate treasury metric. The more relevant questions concern governance, liquidity-risk disclosure, model controls and the reliability of data feeding material financing decisions. For investors and lenders, the quality of liquidity planning can be inferred indirectly from cash buffers, revolver usage, working-capital volatility, refinancing behaviour and management commentary, but those indicators should not be mistaken for a direct measure of forecast quality.
Conclusion: uncertainty is already on the balance sheet
Treasury teams do not need to invent a new accounting line to recognise the cost of uncertainty. It is already visible in the cash they keep idle, the credit capacity they pay to preserve, the investments they delay and the emergency transactions they execute when expected inflows fail to arrive on time.
The next step in cash forecasting is therefore not simply a more accurate point estimate. It is a clearer understanding of the range around that estimate and the financial consequences of carrying that range. The strongest treasury functions will not pretend to eliminate uncertainty. They will price it, decide how much protection is worth buying and improve the forecast where the reduction in uncertainty produces the greatest economic benefit.
References
1. AFP — 2025 Treasury Benchmarking Survey Report
2. EY — Global DNA of the Treasurer Survey
3. PwC — 2025 Global Treasury Survey
4. Bank of England — Forecast Evaluation Report, January 2026
5. AFP — 2025 Liquidity Survey
6. OECD — Reigniting investment for more resilient growth
7. Federal Reserve — Corporate Cash Accumulation During the COVID-19 Pandemic
8. Federal Reserve — July 2026 Senior Loan Officer Opinion Survey
9. Federal Reserve — The Last Taxi: LCR Buffers and Bank Liquidity Provision
10. Federal Reserve — Supply Chain Risk and Bank Lending Amid Trade Policy Uncertainty
11. BIS — Annual Economic Report 2025, Chapter I
12. IFRS Foundation — IFRS 7 Financial Instruments: Disclosures
