The Art of Forecasting, 6 Tips to Create Awesome Forecasts
There are many things I love about my job, but the nerd in me enjoys forecasting the most. In some ways, it is equivalent to design in other fields of study. The forecast is not the main thing, it is not your actual business, but it gives you a view of how your business may perform in the future. There are a myriad of advantages to having and maintaining great forecasts, from managing cash, raising capital, evaluating projects, and financial planning. Without it, the future of your business is a set of scattered puzzle pieces, not a big picture. I am a big believer that a solid strategic plan is just not complete with great forecasting.
To be fair, forecasting is detailed work. It requires understanding how your business works, from sales to manufacturing, operations to administration. As such, it’s not for everyone. But if you want to create great forecasts, there are few things that you need to keep in mind.
Focus on the Objective First
Every part of your business can benefit from some form of planning, and planning will require some form of forecasting. Marketing, human resources, operations, manufacturing, and sales, can all set internal targets and metrics based on what the business needs to achieve. If a forecast assumes or calls for 10% sales growth, or a 1% improvement in EBITDA, or a 3% improvement in cash to sales, different departments can start to plan an implement initiatives to meet targets. Similarly, if an investor wants to know the intrinsic value of a company or project, they’ll need forward looking statements that they can rely on.
However, not every forecast answers the same question, so you have to understand exactly what you’re looking for before you start. For example, if I’m evaluating a capital project, I will look at a simple “as-is” forecast and I will overlay it with a forecast for the capital project. This gives me the intrinsic value of the project, separate from the status quo. On the other hand, tax planning may require detailed forecasting of asset depreciation and financing flows. A payroll forecast may require other forecasts as inputs, such as production, distribution, and sales forecasts.
Pick Appropriate Time Intervals
The time horizon and periodicity of a forecast is also important. Capital projects often need to be evaluated on annual periods, typically 3-5 years out, depending on the nature of the project. Working capital (cash, receivables, payables, inventory) management needs monthly forecasting and targets, typically 12 months out or current fiscal year. Sales forecasts need to take into account seasonality and the user needs to decide if a month over month forecast is more or less informative than a year over year forecast. For example, I forecast monthly top line sales as year over year, but I forecast production targets and capacity based on seasonality adjusted trends. This is because an annual growth trend might give you a great high level view, but products can change significantly from one year to the next, so a more current view is needed.
Lastly, one has to remember that the further out you look, the less information you have and the more you’re likely to be off. Do not pick a time horizon that is longer than necessary, and put less weight on it the further out you go.
We will sometimes see a forecast as a set of numbers on a page. That will give us the impression that each of those numbers was calculated separately and scrutinized separately. This is not so, as it would make forecasting a somewhat intractable problem. Instead, forecasts rely on models and those models rely on assumptions.
Whenever I make a forecast, I try to make sure that the model has internal consistency. When this involves financial forecasts, I also make sure that they are integrated. That is, your income statement forecast should match your balance sheet and cash flow forecast. Assumptions on capital expenditures and depreciation should match long term assets in your balance sheet. Debt financing flows should match interest and cash, to name a few. This can be done without too many assumptions if you do the following:
Use formulas that can be carried across the time horizon. If formulas need to change, your model will become overly complex.
Understand which numbers will be driven by sales, or by the size of your balance sheet, for example.
Use a plug to close your model. In financial forecasts, your balance sheet needs to, well, balance. You can and should use one variable such as cash to be dependent on the rest of your balance sheet assumptions so that your model always balances.
Make Few but Sound Assumptions
Closely related to the choice of time horizon is the choice of assumptions. It is impossible to tell the exact future of a business, and knowing that your forecast will be wrong in one way or another is a great starting assumption. Remember that you’re trying to provide guidance based on what is likely to happen, not to tell the future. But a common temptation when building a forecast is to make many assumptions over a long time horizon. This will quickly undo a forecast because it becomes indefensible. For example, is it possible that you will know the average cost of all your product inputs 10 years out, or even that you will be selling the same products as today? Will you know your general administration costs 5 years out, or your marketing budget?
It is tempting to dive into every line in your income statement and balance sheet and try to answer all questions. Instead, focus on the few assumptions that drive the model. In your income statement, focus on sales, cost of goods sold, and EBITDA to cancel out some of the noise. Make an assumption for tactical capital expenditures instead of figuring out an exact depreciation amount far into the future. Use inventory days, receivable days, and payable days for shorter balance sheet forecasts, but consolidate non-cash working capital for longer forecasts.
As a rule of thumb, if you are making more than half a dozen assumptions, consider the scope of your model or the questions you’re trying to answer. If possible, simplify the model.
Don’t Forget about the Physical World
People tend to forget this when they are working with numbers on a computer: The business is affected by the physical world. While this sounds cynical, consider these examples:
If a retail brand is expected to grow at X% per year, how many stores are opening up? Where? Does it have a physical distribution network to support sales growth?
If a resources company is going to grow, does it have a pipeline of new production sites? How much gold/oil/potash is it able to extract from its current sites?
If a car company will make more cars, what is its current capacity to do so? Will new factories be required? Are new acquisitions taking place?
So while you have have a simple assumption for growth in your forecast and model, you need to understand both the capacity of the firm and how its physical footprint matches your model.
Beware the Fly in the Ointment
Lastly, and this is very important, don’t make a technical mistake in your forecast. Don’t, as the Simpsons scientist says, forget to carry the one. Let’s say you picked the right time horizon, and you made the necessary assumptions, you also need to make sure that you can back up everything you put into your model. Make sure that your assumptions are backed by the correct data and make sure that every formula and calculation is correct.
If you forget this and someone finds an error in your forecast, it will bring into question everything else in your forecast, much like the proverbial fly in the ointment.