What-if modeling is a way of testing decisions before you make them. You build a simplified version of a real system, whether that’s a business budget, a hospital network, or a supply chain, then change one or more inputs to see how the outcome shifts. The core idea is straightforward: instead of guessing what might happen if you raise prices, hire more staff, or lose a supplier, you model it and get a number.
How What-If Modeling Works
Every what-if model starts with a baseline: a set of assumptions that reflect your current reality. You then adjust one or more variables and observe how the output changes. A retailer might ask, “What happens to profit margins if supplier costs increase 15%?” A startup might ask, “How much more runway would we have if we delay hiring to next quarter?” The model takes your inputs, runs them through the relationships you’ve defined, and gives you a projected result.
The simplest version of this is a spreadsheet where you change a single cell and watch the totals update. Excel’s built-in tools, including Scenario Manager, Goal Seek, and Data Tables, let you do exactly this. They’re deterministic, meaning they give you one answer per set of inputs with no built-in probability. For many business questions, that’s enough.
More advanced approaches layer in probability. Monte Carlo simulation, for example, runs a model thousands of times with randomly varying inputs drawn from a defined range. Instead of a single answer, you get a distribution of possible outcomes. This tells you not just what could happen, but how likely each outcome is. It’s especially useful when your inputs are uncertain, like future demand, exchange rates, or customer behavior.
Sensitivity Analysis vs. Scenario Analysis
Two closely related techniques sit under the what-if umbrella, and they serve different purposes. Sensitivity analysis adjusts one variable at a time while keeping everything else constant. This isolates the effect of that single variable, helping you identify which inputs matter most. If changing your shipping cost assumption by 5% barely moves the needle but changing your conversion rate by 5% swings revenue dramatically, you know where to focus.
Scenario analysis changes multiple variables at once to simulate a plausible future state. A “recession scenario” might combine lower sales volume, slower payment cycles, and reduced marketing spend all at the same time. This gives you a broader picture of how different conditions interact, which is closer to how the real world works. Sensitivity analysis tells you which lever matters most. Scenario analysis tells you what a particular future looks like.
Financial Planning and Pricing Decisions
Finance teams are the heaviest users of what-if modeling. The most common applications include pricing and margin sensitivity, where you test how different price points (say, a 3%, 5%, or 7% increase) affect revenue, profit margins, and customer churn. These models help you find the price point that maximizes revenue without driving away too many buyers.
Hiring decisions are another frequent use case. Expanding a team increases both costs and productivity, but those increases hit at different times. A new hire costs money immediately through salary, equipment, and onboarding, but productivity gains often lag by weeks or months. What-if models let you forecast that gap and decide whether to hire now or wait until cash flow is stronger.
Market expansion modeling works similarly. Before opening a new location or entering a new market, you can project the increase in both revenue and costs. The model helps you estimate how long it takes to break even and whether the opportunity justifies the investment. Without it, you’re relying on intuition and rough comparisons to past expansions that may not be relevant.
Healthcare and Capacity Planning
Hospitals and long-term care networks use what-if models to manage patient flow and bed capacity. A simulation model developed for a regional long-term care network, published in the journal Health Systems, captured patient movement between hospitals, skilled nursing facilities, and assisted living facilities. The model estimated wait times, bed utilization, and the number of patients waiting beyond acceptable thresholds.
Researchers ran scenarios with varying rates of bed capacity increases and changes in average length of stay (0%, 1%, 2%, and even a 1% decrease per year for long-term care patients). They tested what minimum bed capacity was needed to ensure that 80% of patients could be placed in care within three days. The model also compared proactive capacity adjustments, adding beds before demand spikes, against reactive ones, adding beds after wait times climb. This kind of modeling gives policymakers concrete numbers to work with instead of reacting to overcrowding after it happens.
Supply Chain Risk Assessment
Supply chain teams use what-if models to stress-test their networks against disruptions. You can simulate what happens when a key supplier goes offline, when shipping routes are blocked, or when demand suddenly spikes. The models measure how different inventory policies, sourcing strategies, and production configurations affect profit, output, and customer service levels.
A common scenario involves assembly and packaging line constraints. If one manufacturing facility can only produce a certain volume due to equipment limitations, the model reveals the bottleneck’s downstream effects on delivery timelines and order fulfillment rates. Running these scenarios before a crisis means you already have a playbook when disruption hits.
Common Mistakes That Undermine Models
What-if models are only as good as the assumptions baked into them, and several recurring errors can quietly make results unreliable.
The most damaging mistake is failing to involve the right people. Executives and analysts often build models without consulting the end users who rely on reports daily. Different teams may define the same metric differently, leading to inconsistent results that no one trusts. A “customer” in the sales database might mean something different than a “customer” in the billing system, and if the model doesn’t reconcile that, the output is misleading.
Granularity is another trap. Capturing too much detail makes the model slow and hard to maintain. Capturing too little means you miss important patterns. A model that tracks every individual transaction when you only need weekly totals wastes resources. A model that lumps all product categories together when margins vary widely across them hides the insight you need.
Overcomplicating the model is a related problem. Unnecessary tables, redundant fields, and overly complex relationships between variables make the model fragile and difficult for anyone other than its creator to use. The best practice is to model only what you actually need to make decisions about.
Perhaps the most dangerous error is skipping validation. A model that runs without throwing errors can still produce wrong answers. If you never check the model’s output against your actual source data, sync issues, missing records, or incorrect formulas can introduce silent errors. Business users lose confidence in the data, and decisions get made on faulty assumptions without anyone realizing it.
Choosing the Right Approach
For straightforward questions with a small number of variables, a spreadsheet-based what-if analysis is often sufficient. Excel’s Scenario Manager lets you save and compare multiple sets of assumptions side by side, and Goal Seek works backward from a target outcome to find the input needed to reach it. These tools are widely understood and require no specialized software.
When your inputs are uncertain or you need to understand the range of possible outcomes rather than a single point estimate, probabilistic methods like Monte Carlo simulation add significant value. These typically require dedicated software or add-ins, since native Excel scenarios don’t incorporate probability.
For complex systems with many interacting components, like hospital networks or global supply chains, simulation platforms that model individual entities (patients, shipments, machines) moving through a system over time give you the most realistic picture. The tradeoff is setup time: these models require more data, more expertise, and more validation to get right.
The best starting point for most teams is a simple model that answers one specific question well. You can always add complexity later. A model that’s 80% accurate and actually gets used beats a perfect model that sits in someone’s laptop because nobody else can understand it.