Finance · scenarios · decision
A project financial model:
from assumptions to a decision
A good model does not predict the future with a single number. It links the operating logic of a project to cash flow, shows the impact of key assumptions and helps structure the conditions under which a decision remains justified.
Anton Konnov · 20 August 2026 · 10 min read
Purpose
Define first which decision the model supports
A single model may serve a product launch, fundraising, an M&A evaluation or the control of a running project. The audience, horizon, precision and set of metrics will differ. So before building anything, fix the decision options and the questions the model must answer.
A model becomes a management tool when the user understands the origin of assumptions, sees the drivers of results and can test an alternative scenario.
Six steps
How to build a project financial model
Define the decision and boundaries
Record the modelling object, audience, horizon, currency, frequency and evaluation criteria. Separately define which decisions the model supports and which remain outside its scope.
Describe operational drivers
Link revenue and cost to real logic: volume, price, utilisation, productivity, timelines, resources and constraints. Financial line items must derive from understandable causes, not grow by an arbitrary percentage.
Link the financial statements
Profit is not cash flow. Payment terms, working capital, investments, funding and taxes are needed. Linked statements help reveal when an economically profitable project may experience a liquidity shortage.
Check assumptions and sources
Separate historical facts, contractual terms, external data, calculated dependencies and expert estimates. For material assumptions, specify the owner, date, source and a range of possible values.
Build scenarios and sensitivity
A base case is supplemented with realistic alternatives. Sensitivity shows the impact of individual variables, and a value-switching threshold is the level at which the model output or the preferred option changes.
Structure decision conditions
The model's output is not only a performance metric. Key risks, a margin of safety, control points, conditions for pause or review and a set of actual metrics for post-launch monitoring are also needed.
Architecture
Which blocks must be transparent
Assumptions
A separate input block with no hidden constants buried inside formulae.
Operating model
Volumes, prices, capacity, resources, timelines and constraints.
Financial statements
Revenue and cost, balance sheet, cash flow and funding.
KPIs and investment metrics
Revenue, margin, cash flow, break-even, NPV, IRR, payback period, DSCR and margin of safety — in the composition that matches the specific decision.
Scenarios and output
Sensitivity, risks, switching thresholds and decision conditions.
Validation
What to check before use
The balance sheet balances, cash flows tie to changes in cash and cash equivalents.
Units, signs, periods and currencies are used consistently.
Material assumptions have a source and an owner.
Scenarios change drivers, not manually override output metrics.
The result can be explained without reading every formula in the file.
AI in modelling
AI accelerates model preparation and review, but does not make an investment decision
AI can extract assumptions from documents, reconcile versions, find inconsistencies, explain formulae, prepare scenarios and check metric sensitivity. An agentic loop links the model to a knowledge base, primary sources and a change log.
Critical formulae, input data and outputs must remain verifiable. Responsibility for methodology, material assumptions and the decision stays with the specialist and the project owner.
Boundaries
A model does not eliminate uncertainty
Scenario analysis is not about choosing the "right future", but about understanding a range of outcomes and the resilience of a decision. The higher the uncertainty and the irreversibility of the step, the more important it is to show assumptions, alternatives and review points explicitly.
This material is educational and does not constitute an individual investment recommendation. A specific model depends on the objective, contractual terms, tax regime and the quality of input data.
Primary sources
Methodological reference points
First step
Describe the decision the model is for
We will define the audience, available data, key assumptions and a reasonable depth of analysis.