Convert a PDF to a Structured Forecast
This guide explains how to take a financial PDF and turn it into a structured Model Reef forecast, using the PDF import pipeline, automatic classification and the Data Library.
Model Reef does not try to replicate spreadsheet style modelling. Instead it uses AI to infer structure from the PDF, create variables and drivers, and then run the accounting engine to generate a complete three statement model and valuation. Your job is to review, refine and extend the forecast, not to rebuild everything by hand.
Before you start
You should have:
A source PDF containing financial information, such as:
Annual report or investor deck.
Bank or lender pack.
Management accounts exported to PDF.
A clear understanding of what the PDF represents:
Single company or group.
Historical only or includes a management forecast.
A Model Reef workspace with permission to create models.
If you are new to the platform, read:
What Is Model Reef
PDF Import Overview
What you will build
By the end of this guide you will have:
A new Model Reef model generated from your PDF.
Mapped variables for revenue, COGS, Opex, staff, assets, liabilities and equity.
Historical periods populated using imported series.
Forward looking periods prefilling with default assumptions that you can refine.
A working set of outputs:
P&L.
Balance Sheet.
Cashflow Statement.
Cash Waterfall.
Valuation and dashboards.
Prepare and review the source PDF
Before import, quickly review the PDF and note:
Which pages contain:
P&L style information.
Balance Sheet style information.
Cashflow details.
Driver tables, such as volumes or prices.
The time span covered, for example FY2021 to FY2024.
The currency and units used:
Dollars, thousands, millions etc.
If the PDF is extremely noisy, you may want to generate a cleaner export from the original system if possible.
Start a new model and upload the PDF
Create a new model in your workspace. Give it a clear name such as:
Client - PDF Import - Initial Model.
Use the PDF import feature to upload your file.
Provide a short description of the PDF if prompted, for example:
"FY22 and FY23 management accounts" or
"Bank forecast pack including three year projection".
Model Reef uses this context to guide table detection and classification.
Review detected tables, dates and units
The PDF pipeline will:
Detect tabular regions.
Infer header rows and date columns.
Identify units and scaling (for example dollars vs thousands).
On the import review screen:
Check that date columns are interpreted correctly:
Period labels map to proper calendar or financial year periods.
Confirm units:
Adjust scaling if the PDF uses thousands or millions.
Remove any tables that are clearly not financial or useful.
This step ensures that the Data Library entries created later will be consistent.
Map lines to variable types, categories and branches
Model Reef will propose a mapping for each detected line item, for example:
Revenue.
COGS.
Opex.
Staff.
Asset or liability balances.
Equity or dividend flows.
You should:
Review the proposed variable type for each line and correct obvious errors.
Assign categories and sub categories aligned with your reporting structure.
Decide which branch each line belongs to if you want a multi branch model:
For now you can keep everything in a single branch if structure is not critical.
Do not worry about perfect categorisation on the first pass. You can refine mapping once the model is compiled.
Let Model Reef compile variables and statements
Once mapping is approved, Model Reef will:
Create Data Library entries for each imported series.
Create variables of the appropriate type, category and branch.
Apply default logic to:
Historical periods (use imported values exactly).
Forecast periods (pre fill based on simple continuation or growth rules).
Run the accounting engine to generate:
P&L.
Balance Sheet.
Cashflow Statement.
Cash Waterfall.
Valuation outputs.
You now have a fully working model derived from the PDF, even before you adjust any logic.
Inspect historical statements for accuracy
Open the P&L, Balance Sheet and Cashflow Statement for the historical period covered by the PDF.
Check that:
Total revenue and key cost lines match the PDF.
Assets and liabilities line up with the source Balance Sheet.
Cash movements and balances are credible.
If you find misclassifications, adjust:
Variable types and categories in the model.
Mapping settings if you repeat the import.
The goal is to match the PDF view for historical periods as closely as is practical.
Review and refine forecast assumptions
The automatically built model will include default assumptions for forward periods, such as:
Extending trends in revenue and cost.
Continuing capex patterns.
Carrying current working capital behaviour into the future.
You should replace or refine these defaults by:
Editing driver based assumptions for growth, margins and costs.
Adjusting timing and delays to reflect realistic payment terms.
Revising capex and funding schedules.
At this stage you are turning the imported skeleton into an actual planning model.
Sanity check the Cash Waterfall and valuation
Use the Cash Waterfall to check whether:
EBITDA and operating cashflow lines look sensible.
Change in Net Working Capital behaves convincingly.
Capex, debt and equity flows reflect your understanding of the business.
Open the valuation view to see:
FCFF and FCFE patterns.
NPV, IRR and Money Multiple.
If any of these look unreasonable, trace back to the underlying variables and Data Library entries and adjust assumptions.
Save this as a baseline model for future work
Once the model feels credible:
Rename it to remove "Initial" and treat it as your baseline forecast.
Optionally create copies for:
Upside and downside scenarios.
Funding or transaction analysis.
Document key assumptions and any deviations from the original PDF in notes.
You now have a structured, machine readable forecast model built from the original PDF.
Check your work
Historical statements in the model broadly match the PDF.
Key variables are correctly classified as Revenue, COGS, Opex, Staff, Assets, Liabilities, Equity or Dividends.
Forward assumptions are based on deliberate drivers rather than unreviewed defaults.
Cash and valuation outputs are internally consistent.
Troubleshooting
Related guides
Last updated