How to choose assumptions for a financial-plan sensitivity analysis

A financial plan is built on estimates: income, spending, taxes, inflation, investment returns, longevity, health costs, and the timing of goals. A single projection can look precise while depending on assumptions that are uncertain. Sensitivity analysis asks how the result changes when one or more assumptions change. The point is not to guess the future exactly; it is to see which uncertainties could alter the advice.

Start with the decision the analysis should inform

Before changing an assumption, name the planning question. Is the client deciding whether to retire now, how much to save, whether a goal is affordable, or how much insurance is needed? An assumption is high priority when changing it could affect that decision. If the modeled result barely moves across a reasonable range, spending hours refining that input may add little value.

Prioritize uncertainty and impact

Two tests help choose what to vary. First, how uncertain is the assumption? Second, how strongly does it affect the plan's outcome or recommendation? An uncertain assumption with little effect may be lower priority. A highly consequential assumption that is relatively certain may still deserve a check, but the strongest sensitivity candidates are often both uncertain and decision-relevant.

Use one-variable tests before combined scenarios

A one-variable-at-a-time test changes one assumption while holding the others constant. For example, model retirement spending at several plausible levels while keeping return and retirement age fixed. This reveals the direction and approximate size of the effect. It is easier to interpret than changing five assumptions at once, because the cause of the difference remains visible.

After isolating key drivers, build combined scenarios that reflect coherent circumstances. A downside scenario might combine lower income, higher health costs, and a later retirement date if those events could reasonably occur together. A scenario should not stack unrelated worst cases simply to make a plan appear to fail. State why the assumptions belong together and how plausible the scenario is.

What to show the client

• The base-case inputs and where they came from.
• Which assumptions were changed and the range or scenario used.
• The effect on the client's stated goal or recommendation, not just an unexplained percentage change.
• Limitations: models simplify taxes, market paths, behavior, and future law.
• Available adjustments and trade-offs if the plan is sensitive to a particular input.

Avoid false precision

A spreadsheet can calculate a result to many decimal places without making the forecast accurate to that precision. Use ranges that have a defensible basis, explain that a model is conditional on its assumptions, and distinguish a sensitivity test from a probability estimate. A one-variable table does not tell the client the odds of a result unless the method explicitly models probabilities and its limitations.

CFP Board's financial-planning practice materials call for addressing how changes in key assumptions may affect whether the client can achieve financial goals. On an exam, identify the client's goal first, then focus testing on the uncertain assumptions that could materially change the recommendation. Present both what the model shows and the action the client can take.

Choose assumptions that can change the decision

A useful sensitivity test begins with the decision at stake. For a retirement date, key inputs may be savings rate, retirement spending, investment return, inflation, longevity, pension start date, and Social Security claiming age. For a home purchase, the important inputs may be interest rate, down payment, property tax, maintenance, and time in the home. Changing every input at once makes it difficult to tell what drives the result; begin with the few variables that are uncertain and consequential.

Separate inputs the client controls from those they do not. Savings rate, retirement date, housing choice, and some spending can change through action. Market returns, health events, tax law, and lifespan are uncertain. This distinction helps turn an analysis into a plan: the client may not control returns, but can choose a flexible spending rule or keep working part time if the downside occurs.

Use one-variable sensitivity and combined scenarios

One-variable sensitivity changes a single assumption while holding others constant. A table might show how projected assets change at 3%, 4%, and 5% real returns, or how a mortgage payment changes when rates move by one percentage point. This isolates the effect and helps identify break-even points. It is not a forecast that each value is equally likely. Explain the base case and the range clearly.

Combined scenarios test variables that may move together. A recession could reduce portfolio value while also causing job loss; inflation could rise with interest rates; a health event could increase spending and reduce work capacity. Construct plausible “adverse but not impossible” conditions, not a worst-case stack of unrelated disasters. Compare a baseline, a downside, and a favorable case, and show what action would respond to each.

Communicate limitations and an action plan

A projection is conditional on assumptions, not a promise. Monte Carlo success rates and deterministic charts depend on input distributions, time horizons, fees, taxes, and asset allocation. Small changes can produce very different probabilities. Tell the client what was modeled, what was not, and which conclusion is robust across scenarios. If a recommendation changes only under an extreme assumption, say so; if it fails under modest downside, address that fragility now.

Sensitivity analysis should lead to monitoring triggers. Examples include revisiting spending if the portfolio falls below a threshold, increasing savings when income rises, or reassessing insurance after a health or family change. Name who will monitor the trigger and when. For exam questions, look for a planner who identifies the decision, tests relevant assumptions, avoids false precision, and explains how the client can adapt rather than presenting one deterministic output as certain.