What Makes a Good Financial Model?
Characteristics of a good financial model
Modeling is used in various disciplines ranging from natural sciences andand finance. With such a wide array of applications, it is safe to say that modeling itself is partly a discipline.
However, it is not a full-fledged discipline because there is no specialist body governing the use of models or acting as a source of accumulated specialist knowledge. As a result, most people practicing modeling are self-taught.
However, despite the lack of standardization in the “pseudo-discipline,”among the enormous populace developing and using models in their jobs every day that following certain principles can bring some order.
It is important to note that standardization cannot be brought overnight. It has its problems, i.e., costs in terms of time and effort in obtaining and applying the knowledge.
Nonetheless, modeling gurus agree that the benefits of having orders far exceed the associated costs. Although it is not uncommon to see some disagreement about the specific practices that bring order, there are a few general pointers on which all agree.
Models are structured quantified representations of real-world settings. Each real-world setting has context and a story.
A good model should act as a presentation of the quantified facts of the matter. In addition, it should tell the story of the real-world setting that it is representing.
In this article, we explore the objectives of modeling, the attributes of good models, and the benefits that accrue from using good models.
Good financial models should be well-structured, logical, transparent, accurate, and appropriate for their intended use.
Key modeling practices include understanding requirements, defining clear assumptions, effective formatting and presentation, avoiding high-risk functionalities, and implementing error checks.
Benefits of good models include standardization, improved modeling practices, time savings, enhanced communication, and positive user impressions.
Modeling lacks full standardization, and many practitioners are self-taught.
Effective financial models act as structured representations of real-world settings, conveying context and quantified facts.
“So, there is no specialist body, no standards, and no benchmarks for models. Then what exactly makes a good model?”
There is no rulebook. Let us know if you come across one; we’re looking too! However, for the time being, there are a few rules of thumb that are generally followed by everyone using models day in and day out.
Models are parts of processes. Therefore, the only definitive answer to the above question is that a good model makes the process, of which it is a part, easier and more efficient.
Generally, the following attributes are good to have in a model:
A model comprises various worksheets that complete it and make it functional. At the minimum, it should contain all the components that are sufficient to meet the needs of its users.
These components should be logically structured. It is highly advisable to plan the layers in the model before diving into building the model. Users should plan how the input variables, processing workspace, and the necessary output will be organized.
A logical structure ensures that models are easy to build, understand, use, audit, and fix.
Experts suggest creating layers in models in the given order:
- Cover page
- Input worksheets
- Processing worksheets
- Output worksheets
- Further analyses (such as and scenario analyses)
- Findings and conclusions
These may be further divided into multiple parts to meet the objectives.
A financial model is a quantified reflection of a real-life situation with economic, industrial, and market aspects and reproduces the perspectives of the model developer and the users. Therefore, it should be logically sound, and consider the fundamental relationships between various relevant factors.
Often, a model is created by one but used by many, some of whom may not be technically literate. Hence, it must be easy to understand for all users and have a simple interface. In addition, the entire modeling process from inputs to conclusions should be easy to follow.
There should be clarity in the flow of data. A model should seem like a neatly laid-out presentation. Keeping it simple also allows modelers to focus on its key variable or aspects.
It should be simplified so that even users with no modeling experience can follow it. Designing the model in such a way involves steps such as making logic easy to follow and simplifying formulas by breaking them into easier steps.
Each model should also have proper documentation. Other than that, it may also be helpful to have a user guide in the same file so that it is always available to all users, even when the file is transferred between them.
Minimizing model risk
SR 11-7 is a standard issued by the US Fed on managing model risk. It defines model risk as “the potential for adverse consequences from decisions based on incorrect or misused model outputs and reports.”
It also states that model risk can result in losses, poor decision-making, or reputational damage.
Model risk primarily arises in two situations. First, when a model has fundamental flaws that produce inaccurate results. Second, when it is used incorrectly or inappropriately.
Therefore, models should be designed so that the chances of these two situations materializing are minimized.
The processing of inputs in a model (the computations) must be accurate. While some calculations lie in plain sight, some may be hidden, and while some may be as simple as “2 + 2”, others can be more complex. For the model to be useful to the decision-makers, its calculations need to produce the results required.
The way the model processes the inputs and assumptions shapes the results that are used for decision-making. As a result, using flawed models can lead to losses of enormous magnitude.
A model may need to be modified as per the variety of the use cases in which it might be intended to be used or when new data becomes available. Flexibility generally refers to how much the model can adapt.
For instance, if a model needs to be used frequently under varying circumstances, it must be adaptable, i.e., flexible. However, a flexible model would produce less reliable forecasts as it cannot capture the specifics of each case.
On the other hand, a model built to assist with a single decision will not be flexible, but it might capture the case’s specifics well and hence have better predictive power.
Flexible models allow users to run various analyses and make modifications when new data becomes available. Flexibility can be achieved by keeping the model simple. Making it complex and introducing innumerable devices for every eventuality would take away from its flexibility.
Speaking of “flexible,” did you know that WSO offers some amazing flexible templates for free?
And not just for investment banking models, but plenty of other categories too! Private equity model templates.. Resume templates. Cover letter templates. Presentation templates. Phew! You can sign up below to receive them.
It can also be referred to as the granularity of a model, which is the level of details incorporated into the model. However, a model must not be excessively granular as they are simplified and structured representations of real-world situations.
They are simpler because it is impossible to capture all the details from the real world. Additionally, simplification can be intentional to maintain focus on the key aspects.
However, some cases may need models to be more granular. Such models may be more complex and time-consuming to build, but they are more precise.
Other models may be back-of-the-envelope calculations that do not require as many details but lack the level of precision necessary for high-stakes decision-making.
A good model bridges the gap between its costs and benefits. It maximizes the precision of forecasts as much as possible while keeping it simple at the same time.
Models are tools used for decision-making, so they need to be effective in how their results are communicated to the decision-makers. Building a model would be a wasteful exercise if others cannot understand it.
The assumptions used to build the forecasts need to be clearly defined. It must have its results and conclusions neatly laid out. It is ideal to have an executive summary as well. Furthermore, a good model should enable its users to present the results differently as per their needs. It should also allow further analyses of the data.
“These principles sound good. But how do I build a flexible model? Do I have my model do yoga every day?”
Principles can sometimes be vague blanket statements so that they can be applied in vastly different scenarios alike. This vagueness can make it confusing to follow.
The following practices generally ensure that a model abides by the above principles.
Understanding the requirements
To begin with, it is paramount that model developers answer a few questions to have a complete understanding of their assignments. If they do not understand their assignment well, they cannot expect a good model.
Good questions to address before starting to build a model are:
- What is the purpose of the model?
- Who are the end-users?
- What are the requirements of the model?
- What are the inputs to the model, and are they adequate?
- What is the output supposed to be like?
- How will the input be processed to have the desired output?
- Are there any specific requirements? (Is the model going to be a template? Is it a project-specific model?)
When developing a, modelers need to decide what variables they need to account for to obtain accurate forecasts. These assumptions they make will be the base upon which the model is built.
The assumptions should be realistic and well-grounded in the company’s experience and knowledge of the industry. They should also be clearly stated, easily changeable, and adjustable if needed.
For example, if you’re creating a model for a restaurant, it might make sense to assumeis directly proportional to their sales.
Formatting and presentation
Model developers and the end-users of a model are often not the same. Therefore, a model and its conclusions need to be presented to its end-users effectively to avoid any gaps in communication, which can lead to poor decision-making or, worse, reputational damage.
Like any other management report, a model should be formatted pleasantly. Good formatting improves readability and saves time.
Such a model is also easy to review, audit, and update. It will also save effort if another modeler takes over the assignment.
Good formatting involves consistency through the model, using reader-friendly fonts and appropriate colors. Inputs, assumptions, formulas, and results need to be differentiated from each other. Modelers must supplement the model with a format key to guide the users to use different formats.
Excel and other spreadsheet software offer users a wide range of functionalities, some of which are riskier to use. These are called vulnerable functions as they tend to be unreliable in many cases.
For example, they may have a higher risk of errors, reduce calculation speed, or be difficult to check.
They may slow down the spreadsheet, may not be updated dynamically, or may even corrupt the model, especially when transferred to users with older versions of Excel.
Therefore, modelers must avoid using high-risk functionalities unless necessary.
- Circular references: There is no guarantee that circular references produce the correct solution, especially in cases where there is more than one possible solution. Additional layers of circularity can exponentially complicate the model. It can get much harder to check for errors and audit the model. It also requires a lot of computational power.
- Volatile functions: These functions produce different values each time they are calculated (each time the spreadsheet is changed).
Modelers must ensure that each row and column has an accurate label. Labels should be structured, descriptive, and well-formatted. They must identify the values in each line item. Each column must have a clear purpose.
Each time series should be allotted one column in a worksheet. A mismatch of time series in a column makes models inefficient. A worksheet should cover only one continuous period. For example, the same worksheet should not have a periodicity of 2021-2031 and 2000-2005.
Error checks run throughout your model and assist in eliminating errors. You can imagine them as the safety system of your model.
For example, you would not dare think about building a gas station without fire safety systems. The safety measures are installed simultaneously and begin to monitor everything right then. Much like that, error checks are put in place as the model is being developed.
Modelers must have a column in each spreadsheet to aggregate all the error indicators.
In addition, they must have a particular section dedicated to organizing and summarizing all error checks in the entire workbook.
The points above are not exhaustive, and neither are they hard and fast rules. The only rule is that whatever steps you take to make your model better must bring ease of use or efficiency. To learn more about this, please check out WSO’s guide on the best practices in financial modeling.
Since modeling is essentially a no man’s land, following the best practices is the closest thing to abiding by its law. It brings about standardization. The race towards better models will improve modeling as a practice in the long run.
It also saves time by avoiding errors and making them easier to understand and review. Few things can be more confusing than models that are developed inconsistent with best practices.
Standardization in any field indicates the presence of a framework. Models developed with a framework make conveying a complicated message a piece of cake.
They also ensure the understandability of a model. It enables everybody to be able to answer all questions about the model.
Besides, following the best practices makes simple and neat models that leave a fantastic first impression.