While during their studies a student may be able to create a perfect DCF model, when their CFO asks them to rebuild the model in similar circumstances as a result of the changes in one of the revenue assumptions, they might be at a loss. The reason for this is that there is a big difference between simply knowing the theory and being able to use it in practice, especially when it comes to financial modeling. This is where most classroom teaching falls short.
Financial modeling classroom teachings focus on syntax, not judgment
Most financial modeling classes are built as a math problem. You get a case, some assumptions, and an answer. In fact, that is the correct way to learn the general syntax of a financial model. It is good while learning how to make a Three Statement Model balance or to understand how the LBO waterfall is structured. However, it does not prepare you for the actual job because, in the real world, you are not given the correct set of assumptions. Usually, you either find a model and have to reverse-engineer the sometimes illogical assumptions or come up with your own that can be justified and have flexibility to perform different scenarios. This prepares you for the actual job, where you have to have judgment about which assumptions to bend and which to break. If someone is looking to develop a certain skill set, then the FME courses focus on building the real application of financial modeling, not theoretical syntax.
How you build the formulas is actually a small part of the job, think 25-30%. The bigger part is knowing why you are building them, such as knowing that the Customer Churn % is more sensitive to change than the Employee Churn % and by how much, or realizing that the revenue seasonality if not incorporated correctly will wreak havoc on the P&L. It is all about using the model setup and the outputs it provides to justify your conclusions and convince your stakeholders, which sometimes requires running scenarios on the fly by changing some assumptions based on information from another context. Financial Modeling Education courses offer structured practice in three-statement models and 13-week cash flow forecasts, keeping the emphasis on building and working with models rather than only learning the syntax.
Real models are dialogues, not statements
A student model is a statement, while a professional model is a dialogue. Your model may have been graded by a TA, but at work it goes through 5 people over 6 weeks. Someone added the revenue build based on the last call with Sales, someone noted that the cost assumptions were wrong, and the CFO asked to add a downside case last night for the Board meeting. Version control and collaboration are now integral parts of modeling that are not taught in school.
Sensitivity analysis is more than just a one-variable data table. In reality, the executive wants to see what will happen if three variables change at once, and they want to know yesterday. Developing a model that can withstand the scrutiny of multiple changing variables at once without errors or without having to completely rebuild it is a skill that is best developed in the workplace, where time is of the essence, as opposed to school projects with due dates a month away.
AI will take over most of the mechanics of modeling
While AI will be able to take the vast majority of the calculation-based, rote part of building a model when the right triggers are implemented, it will not replace the judgment skills required for financial modeling. As professor Andrew Lo put it, “If you’re doing analysis with Excel, you’re already competing with the machine and losing.” Copilots can already create formulas, identify errors, recognize patterns, and generate a first-cut model based on a data file. Abstraction is easier when you have the ability to quickly test out ideas based on different data points, and then actually create an output based on your hypothesis. Columns of numbers are actually boring to look at, however, a machine can generate charts based on your company’s performance that you can then use to base your own judgment on. The mechanics of the task are less important than the reasoning behind them.
Unstructured data is a big part of the job
The clean data sets of school projects are very far removed from the messy data of the real world. It comes in chunks, it is formatted incorrectly, has random entries, structural issues, and often contains errors or missing values. In other words, the data needs to be preprocessed before you can plug it into your model. And the amount of time you spend on preprocessing determines how much value you will extract from it, since garbage in equals garbage out. Learning how to preprocess messy data sets and then building a model on top of that which can withstand an audit is a different skill altogether, one that is not taught in school. Error checking, audit trails, and overall data integrity are important parts of the job that are separate from building the actual model but are often ignored nonetheless.
Bridging the gaps requires more practice than just watching other people do it
Bridging the gaps requires more practice than just watching other people do it, such as through financial modeling courses. To develop these skills, one has to practice with unstructured problems similar to an actual M&A or FP&A deal, where there are moving parts, changing assumptions, and incomplete data. AI can be an excellent tool for this type of practice, as it can act as a copilot that allows the user to become familiar with the mechanics of a financial model without getting bogged down by the details.
The bar is raised every time you think you are close
Knowing the underlying principles enables you to solve the problem at hand, but the moment the ball is in your court, you will be asked to justify your assumptions under time constraints. This is where those who have the ability to practice under these conditions have an advantage over those who have only learned the theory.
