Methodology
Where the evidence comes from, how the numbers are computed and how the verdict is reached. The same method for every study.
What goes in
A study starts from what you know. The more you give, the more of the study rests on your own facts.
Project details
Name, country, location and sector. The sector sets the chapter outline and the financial model.
Project information
Your notes on the project: what it is, who is behind it, what has been done.
Documents
Concept notes, earlier studies, spreadsheets and reports. Each one is read and summarised.
The run
Once you start the study, it runs without stopping. Most steps run in parallel, so the whole study is written in ten minutes or less.
- 01
Set up
The sector's outline of about 33 chapters is fitted to the project. Chapters that do not apply are dropped.
- 02
Research
Agents search the web, your documents and public data for each chapter, and extract facts with their sources.
- 03
Evidence pack
The facts are merged, duplicates removed and the key figures fixed before any text is written.
- 04
Model and verdict
The financial model is run in code from the key figures, and the verdict is set by rules.
- 05
Writing
Every chapter, the conclusions and the executive summary are written from the same evidence pack.
- 06
Checks
Numbers are checked against the key figures, weak sections are rewritten and citations are resolved.
Evidence
Every fact in a study keeps its value, unit, year, source and a confidence level. When two facts disagree, the better source wins, in this order.
- 01
Your documents
Facts from the documents and information you added. These come first.
- 02
Research
Facts from web sources and public data found for this study, with a link to each.
- 03
Benchmark
Published sector or country figures, used only where research found nothing. Always labelled as a benchmark.
The data check
Each study rates its own evidence by the share of chapters backed by documents or research rather than benchmarks alone.
- High
- At least three quarters of the chapters.
- Medium
- Between 45% and 75% of the chapters.
- Low
- Under 45%. Read the study as an early sketch.
A financial model for the sector
Language models explain results. They do not produce them. The figures come from a financial model run in code, delivered as an Excel workbook.
Models by sector
- Energy
- Transport
- Water
- Agriculture
- Mining
- Digital
- A general model for other sectors
- Inputs come from the key figures: the project's own figure where there is one, else a published benchmark.
- No number is invented. An input with no evidence uses the model's labelled sector benchmark.
- Results such as IRR and NPV are computed, never written by the language model.
The verdict
The verdict is set by explicit rules applied to the financial model, the risks and the evidence. The writers explain it. They cannot change it. It concerns the next preparation step, never a permission to build or invest.
Viable
The project passes every rule applied to its model, risks and evidence.
Conditionally viable
Take the project further only once the named conditions are met.
Not viable
The project fails at least one deal-breaker test as it is structured today.
Checks
Before the study is published, code checks what the writers produced.
- Every number with a unit is compared with the key figures and across chapters. A wrong value is corrected.
- Sections that are too short, too long or cite too few facts are rewritten.
- Fact references become real citations with the publisher, title and link.
- The verdict stated in the text must match the verdict set by the rules.
What comes out
- Executive summary
- The project, the findings and the verdict in a few pages.
- Full PFS
- The complete study, chapter by chapter, with citations.
- Finance model
- The Excel workbook with its inputs and results.
- Data and sources
- Every fact used, its source and the data check.
Limits
A desktop study uses existing data and desk research only. It does not replace fieldwork, engineering or legal advice. Every study carries this notice:
Desktop pre-feasibility study. Not a bankable feasibility study, engineering design, legal opinion, land-title confirmation, permit approval, or investment recommendation. All primary-data gaps and modelled inputs require validation.

Bring your next project
Enter the project details and what you know. In ten minutes or less you have the study, the financial model and a clear verdict.