I'm Zahoor Bhat. I have spent fifteen years in financial modelling and equity research, working with portfolio managers and research teams around the world. It is the part of finance I actually enjoy: taking a business apart in a model and working out what moves it.
Now I am building small tools that take on parts of that same work. Reading filings, tracking what management said against what the company reported, making a repeated piece of analysis repeatable. I test them on real filings, and most of what I have learned so far is where they fall short.
Three of these six steps have something built against them. The last one is the step I am not trying to automate.
Each of these came out of a step I was doing by hand every quarter. All three are still in development.
It watches SEC EDGAR for a list of companies. When a new 10-Q or 10-K is accepted, it reads the notes and disclosures against the previous filing and reports what changed, ranked by how much it matters. The headline numbers are already in the press release. The explanation is usually in a note that most people skip.
It records what management guided to each quarter and sets it against what the company actually reported over the following four quarters. Most analysts carry a private view on whether a management team's forward statements can be trusted. This turns that view into a record.
A place to define a research question once and then run it across companies and quarters. The point is that two analysts asking the same question get the same shape of answer, instead of rebuilding the approach each time.
This is the work that happens before the analysis starts, every quarter, for every name under coverage:
Extraction, comparison and tracking are mechanical, and I think an agent can take most of them. Deciding what a change means for the thesis is not mechanical, and I am not trying to automate it. The value of the tools is that they hand an analyst better inputs, sooner.
I trained as an engineer, which is where the habit of pulling systems apart started. It is also why building these tools comes naturally to me.
Over 500 financial models built from scratch across fifteen years — LBOs, DCFs, operating models and comparables — for portfolio managers and research teams around the world, covering banks, telecom, consumer, industrials and semiconductors. That work is where I learned how analysts think about a company: revenue drivers, margins, operating leverage, cash flow, and the assumptions holding a valuation together. It is still the work I find most interesting.
Modelling taught me the shape of the work. I am now finding out how much of that shape is mechanical. These tools get built on nights and weekends and tested against real filings, which is the only way I have found to tell whether an idea holds.
If you are an analyst or a portfolio manager and any of this looks useful, or looks naive, I would like to hear it.