01 / Set a goal
Start with a question worth answering.
Define the question, the decision, and what a useful result would look like.
Name the decision
I start by writing down what I want to find out and why it matters. I want a clear question before I spend time collecting data or building a model.
Set the scope
I decide which company or process I am looking at, the time period, and the information I can realistically obtain. I keep the goal small enough to work through properly.
Choose a way to judge the work
I decide what would make the result useful: a clearer funding requirement, a comparison of scenarios, or a repeatable task that is easier to check. I separate the result I want from any claim about improvement.
Leave room to refine the question
The first goal gives me a direction. As I learn about the industry and inspect the data, I may need to narrow it or change how I answer it.
Why I put this stage here
A clear goal helps me decide what information matters and when the work is enough to support a decision.
A related project: an answerable cash question
For the 13-week model, the current bank balance was not public. The useful question became how much opening liquidity an illustrative scenario would require.
Explore the project ↗02 / Industry
Understand the business before the numbers.
Map the industry, its leading companies, buyers, and suppliers.
Understand the industry structure
With the goal in mind, I work out how the industry operates: its main activities, how businesses create value, and how money moves through the system.
Identify the leading companies
I identify the largest companies and distinguish their roles and business models. This helps me understand the competitive landscape and place the company I am studying in context.
Define the buyers and suppliers
I map the main customers and suppliers, what they exchange, and how they depend on one another. I want to understand what drives demand and what shapes supply before interpreting the numbers.
Turn that understanding into questions
I use this industry map to identify the questions that matter for the project. Those questions give the financial analysis a starting point grounded in how the business actually works.
Why I put this stage here
The same figure can mean different things in different businesses. Understanding how the industry works helps me decide what a number can actually tell me.
A related project: liquidity and payment timing
The 13-week model distinguishes profit from the timing of receipts and payments. Its case study shows why the flow of money matters to the question being asked.
Explore the project ↗03 / Data
Make the evidence fit to work with.
Collect, cross-check, clean, and visualize the inputs before modeling.
What led me here
Extracting data, organizing it in Excel, and cleaning it made me notice how much time could go into repetitive preparation. I started looking for a more efficient way to do that work. That experience is part of why I care about both usable data and the process used to prepare it.
Collect information and extract the data
With the goal and industry context in place, I gather the information needed for the analysis. I keep track of each source, period, unit, and definition so that the inputs remain traceable.
Verify, cross-check, and clean
I double-check the figures and cross-check them against supporting sources where available. I then organize the data, resolve inconsistent formats and units, and examine missing values before using it in the model.
Visualize what the data is saying
I use tables and charts to examine patterns, compare scenarios, and make the main drivers easier to see. Visualization helps me interpret the analysis and explain the reasoning behind it.
Decide what the evidence can support
I check which parts of the question the data can answer and where the gaps remain. I keep missing values and uncertain estimates visible, and carry those limits into the financial analysis.
Why I put this stage here
An analysis can look convincing even when its inputs are wrong. Data verification matters especially when AI makes it easy to produce a fluent answer quickly.
A related project: assumptions that need inspection
The DCF case presents historical model inputs, scenario outputs, and limitations. It is an example of making the basis of an output inspectable, rather than treating the output alone as a conclusion.
Explore the project ↗04 / Finance
Turn the evidence into a financial answer.
Turn checked data into a model, compare scenarios, and explain the result.
Choose the method for the goal
I return to the question I set at the start and compare ways to answer it. I choose a method that fits the industry and the data I can support, rather than adding complexity for its own sake.
Build the model in clear steps
I separate source data, assumptions, calculations, and outputs. I want to see how a change in an input moves through the model and affects the answer.
Compare scenarios and explain the drivers
I change the assumptions that matter and compare the results. I explain which variables drive the difference, what the model suggests, and what it cannot tell me.
Compare with actual results
Where comparable actual figures are available, I compare them with the model using matching periods and definitions. If they are not available, I record the indicators I need to follow later.
Why I put this stage here
The financial model comes after the goal, industry context, and data checks. That gives the calculations a clear purpose and a basis I can explain.
A related project: an answerable cash question
In the liquidity project, the current bank balance was unknown. The model therefore asks what opening liquidity a scenario would require, rather than claiming to predict the company’s actual funding shortfall.
Explore the project ↗05 / Automation
Keep learning after the model is built.
Track actual results, automate follow-up, and capture what I learn.
What led me here
I did not want to keep spending so much time repeating data extraction, Excel organization, and cleaning. Looking for a more efficient method led me to become interested in AI, technology, and coding. For me, that interest started with a practical question: how could I make this work more productive?
Follow the actual results
After reaching a conclusion, I return to the relevant indicators as new information becomes available. I compare observed results with the original estimates, keeping the time period and definitions consistent.
Investigate the differences
When the results differ, I look at which variables changed and which assumptions did not hold. I use relevant news and other supporting information to understand possible explanations, while distinguishing evidence from my own interpretation.
Automate recurring follow-up
I look for repeatable steps in this review process: collecting updated indicators, extracting data, and crawling or gathering relevant articles. I build automation around those steps to reduce repetitive work and make follow-up easier to maintain.
Keep responsibility for the interpretation
Automation helps collect and organize information, but I still verify the results, assess what the differences mean, and make the final calls. The workflow supports my judgment; it does not replace it.
Close the loop: what I learned
Finally, I write down what I learned, what I would change, and which methods were useful. I carry those lessons into the next project, connecting new knowledge to work I can actually use it for.
Why I put this stage here
A project does not end with a model output. Following the results helps me improve the analysis, and automating repetitive checks leaves more time to understand the differences and learn from them.
A related project: the local n8n workflow
The research-to-content prototype connects collection, structured information, and AI-assisted output. Its case study includes the existing workflow artifact and the limits of the current demonstration.
Explore the project ↗