When Board Commentary Hides Unvalidated Scenario Assumptions

Start With the Assumption Log, Not the Commentary

When the deck opens with a confident summary, the board rarely asks which rows of the forecast were actually tested. The narrative is smooth. The numbers look round. The time pressure is real. Beneath the commentary, a scenario model can still carry assumptions no one has validated.

This is a familiar position in FP&A. A pricing scenario, a working-capital bridge, or a revenue ramp gets built in Excel, then summarized into a one-page board pack. By the time the CFO's commentary is added, the assumptions have become part of the story. The story sounds authoritative. The underlying parameters may be stale, single-sourced, or drawn from a range without any supporting logic.

The failure is not the board's lack of curiosity. It is that the commentary layer hides the validation status of the inputs. The fix is not to make every assumption visible to the board. The fix is to separate validation from narrative and to make validation a named step before commentary is drafted.

Validate Before You Narrate

Assumption validation has a repeatable order: extract, classify, validate, re-run, then narrate.

1. Extract every scenario input into a log. List the assumption name, the cell or table where it lives, the owner, the last validation date, and the source of the value. A simple Excel table works. VBA or Power Query can pull named ranges from the scenario model into the log so the log cannot drift from the model.

2. Classify each assumption by materiality. Not every input needs full documentation. A materiality threshold turns noise into an owned list. For a revenue scenario, the elasticity parameter is material. The rounding convention in a formatting cell is not.

3. Validate material assumptions against an external source, a prior study, or a defined range. An assumption sourced from a single email is not validated. An assumption that has not been compared against the prior quarter is not refreshed. The log should show a red flag when validation is blank or older than the approval window.

4. Re-run the scenario output after each material assumption changes. This is the validation of the model itself, not just the inputs. CFA Institute Standard V(A) is direct on this point: practitioners must understand the assumptions and limitations inherent in any model and must test the output before incorporating it into decision-making.

5. Draft board commentary only after the log shows no red flags, or after the red flags are explicitly called out as limitations. Commentary that smooths over a red flag is not communication. It is concealment by omission.

A Worked Example: The 3.8 Percent Pricing Uplift

Consider a quarterly board pack for a manufacturing company. The revenue bridge shows a 3.8 percent uplift from pricing. The CFO's commentary reads: 'Pricing remains robust across key accounts.' The scenario model behind that line has three pricing assumptions: input-cost pass-through at 60 percent, a customer mix shift to higher-margin SKUs, and an elasticity parameter of negative 0.15.

An analyst reviewing the model before the board meeting finds two problems. The elasticity parameter was carried forward from the prior quarter without checking against the updated elasticity study, which shows negative 0.35. The 60 percent pass-through came from a single procurement conversation and was never tested against a range of 45 to 70 percent. Both assumptions are material. Both appear in the board commentary as fact.

The control layer here is simple. A scenario log flags the elasticity parameter because its validation date is blank. A sensitivity check reruns the pricing uplift with the elasticity at negative 0.35 and the pass-through at 45 percent. The uplift drops from 3.8 percent to 1.2 percent. Before the commentary is drafted, the pack shows the unvalidated assumptions in red and the range of outcomes. The board still sees a concise summary, but it is a summary built on tested inputs, not on narrative smoothing.

This is not a one-off. The same dynamic appears in working-capital forecasts, acquisition scenarios, and FX exposure slides. The specific numbers change. The control layer does not.

What the Standards Require

Standards and frameworks support this discipline. The CFA Institute Standard V(A) requires that analysis incorporate a broad range of assumptions sufficient to capture potentially negative outcomes, not only the comfortable case. The FASB Conceptual Framework states that forward-looking information that cannot be verified until a future period should disclose its underlying assumptions and compilation methods so users can decide whether to rely on it. The SEC's order approving PCAOB AS 2501 notes that significant assumptions in estimates can be susceptible to manipulation or bias and that independent review must evaluate whether there is a reasonable basis for those assumptions. These are not optional guidelines for FP&A. They are the same obligations that apply when a model output moves from a spreadsheet to a boardroom.

One specific tool helps here. A well-specified scenario has three components: a proposed event, differential impacts across risk factors, and a path over time. A single fixed-point assertion, such as a flat 10 percent downside with no timing or factor-level effects, is a weaker standard. Board commentary built on a fixed point invites the smoothing problem because there is no dynamic to validate. When the scenario includes a path and factor-level impacts, the analyst can check each component against a source or a prior study. That is the difference between a scenario and a slogan.

Where the Control Layer Breaks

A limitation of this approach is straightforward. Automated validation checks cannot tell whether a 2 percent share-shift assumption is strategically plausible. They can only prove the arithmetic is clean, the source is documented, and the scenario definition is complete. Judgment is still required. If the business has a new product launch with no historical analogue, no validation log will produce certainty. The best the analyst can do is mark the assumption as undetermined, show the sensitivity band, and let the board see the honest range. That is still better than commentary that pretends the assumption was settled.

Monday Morning Takeaway

Reconciliation before presentation: validate the scenario inputs as named assumptions before any commentary is drafted. If the log is not clean, the narrative is not ready.

Extract first, then validate, then narrate. A story that runs ahead of the control layer eventually runs into the audit committee.