CFO Connect logo
Forecasting targets light pink
Meetup Recap

The Art & Science of Forecasting

Luc Hancock
Luc Hancock CFO Connect

“Forecasting isn’t a hard science. The most effective forecasting is both an art and a science.”

Financial planning and analysis has become increasingly important to businesses thanks to a turbulent few years. Leaders want to know where they’ll stand in the quarters and years to come, and they use data from the finance team to plan and budget accordingly.

FP&A enables finance teams to analyze the company’s financial data, predict future outcomes with forecasting, and then share this information with the executive team. The executive team goes on to make informed strategic business decisions based on this data and the FP&A team’s recommendations.

Forecasting is still an emerging discipline, and even seasoned finance professionals are interested in learning more on the topic. So to help, CFO Connect had the pleasure of hosting a webinar with FP&A expert Christian Wattig on “The Art and Science of Forecasting.” 

Read on for a recap of the webinar.

About the expert

Christian spent over 13 years in FP&A leadership roles in the consumer goods sector before earning his MBA at NYU Stern School of Business. He’s currently the Head of Training at FP&A software company Datarails, where he combines his FP&A expertise and love for teaching.

Christian has also developed a course for finance teams and business leaders: FP&A Bootcamp. It’s four interactive workshops with breakout room discussions and role plays to help finance professionals improve their FP&A skills.

Objectives of forecasting

Forecasting is a foundational pillar of FP&A that uses historical data to predict what will happen in the future. Finance teams use forecasting models with different inputs to estimate future outcomes, often using best and worst case scenarios to find a likely middle ground.

Christian started the webinar by breaking down forecasting into three distinct stages or objectives: 

1. Predict business performance 

First of all, why do companies need to predict business performance? What’s the purpose? 

“Mainly, your investors and other stakeholders want to know what they can expect in terms of a return on their money. They want to know what’s happening next quarter and the following year. But effective forecasting can achieve more than that. When done well, it helps you to make better decisions that drive your business forward.”

There are a few popular forecasting methods, and most companies use some combination:

Top-down forecasting

“For top-down forecasting, start with your high-level result in mind beforehand. Figure out what you need to do to reach your desired result. For example, what do you need to invest and/or change in order to hit your goal?” 

Bottom-up forecasting

“With bottom-up forecasting, there’s no end result in mind. Rather, you start with a blank slate. Look at your projects and ask how each project can help deliver your financial goals. Add up all your projects, get to a number, and see where you land.” 

Driver-based forecasting

“First, determine what your business drivers are, and then estimate how each of those business drivers impacts your growth and your bottom line. It’s an approximation, not an exact science. But iteration results in developing a better understanding of your business.”

2. Create a baseline for business performance

In order to create a baseline for business performance, you need to compare your forecast to your actual results. 

“Did the business perform as expected? Usually, you don’t hit your exact forecast, given the amount of uncertainty that’s out there. Look at the forecast and your actuals, and try to find the root cause. Dig into the variances to find out why the business performed the way it did.” 

Christian said that a typical pitfall is ‘analysis paralysis.’ This happens when people get lost in the details. 

“In variance analysis, people want to explain everything. Going line by line on the P&L and trying to compare to the forecast and then explain each variance” is a surefire way to get stuck.

“Instead, approach it with an 80/20 mindset. What are the 20% of your variances that make up 80% of the impact?” Focus on that ratio to get to your insights.

3. Use your forecast analysis to inform business partners

So you’ve compared your forecast to your actual business performance and analyzed the results. Now, you get to use this analysis to make your recommendations to the executive team.

“That’s the ultimate goal, taking your analysis and forecast and making concrete recommendations about what the company should do. It makes finance fun and rewarding.

Business partnering

Finance teams can have a huge impact on the business, but it’s important that they translate their raw data into something concrete. So how can FP&A team members turn their data into useful recommendations? 

Christian identified two ways for finance teams to involve key stakeholders.

Root cause analysis 

“When doing your variance analysis between your forecast and your actuals, you need to dig. Keep digging, look at data, and go to business partners and ask follow-up questions. 

"The tricky part is: how do you know you’ve found the root cause? 

You know you’ve done enough questioning, or found the root cause, when what you’ve found suggests an explicit action that will address the issue.”

Scenario planning

“Scenario planning allows you to make decisions and plan ahead of time. This improves the quality of decision making.  

“Imagine that you expect Q3 revenue will be $5 million. But of course it won’t be exactly that. It will be more or less. If it’s a lot less, then you have to rush to create an emergency plan to catch up in the next quarter.  

“When you collect inputs from your business partners to update your forecasts, ask for both the worst case and best case. Take all these inputs, put them into your model, and decide, along with the leadership team, where to take conservative, aggressive, or middle routes. Then you get the expected case scenario. 

If the worst case happens, what are you going to do? Scenario planning allows you to have these conversations ahead of time, and then you can make concrete recommendations because you already have all the data.”

Forecasting best practices

Forecasting is multidimensional. It’s not a one-and-done exercise, and there are different strategies and practices that will help your team get better returns.

Use multiple forecasting methods

To yield the best and most accurate forecasting results, run multiple forecasting methods in parallel. 

“Run your normal top-down forecasting method, which is likely to be more aggressive due to executives’ goals. 

“At the same time, ask your business teams to share their bottom-up with you. Bottom-up usually tends to be more conservative, because the teams who have to deliver against this forecast will want to over-deliver and exceed expectations.

“Compare the two results, and study the difference. This will open a conversation about how and why this difference exists, and you’ll be able to find a way to get to the desired result.

“In addition, run a proper driver-based forecast. Estimate how much each of your investments will result in top-line and bottom-line. At first, your accuracy may be lower. This is because you need to iterate; every time you run it, you learn something and your accuracy increases.”

And Christian’s final tip for better accuracy? If your company has the resources, try a data science-based approach as well, where you use algorithms to extrapolate the future from the past data. 

Strive for simplicity

No matter what, there will always be an element of uncertainty, whether it’s due to economic or geopolitical factors. This uncertainty is tough for companies to deal with, but there are ways to account for it in forecasting. 

To counteract uncertainty, “...many companies increase the complexity of their forecasting models. But the more assumptions you build into the model, the worse the accuracy. You’re stacking errors on top of each other, and then they multiply. 

“When the uncertainty increases and the accuracy decreases, take a step back and see if you can simplify your model. Take out some assumptions.

“This forces you to slow down and analyze which business drivers have the biggest impact or are the most helpful to predict future performance.  

“Once you narrow it down to just a few assumptions and actuals, it’s much easier to iterate and tweak the model. When there are too many assumptions, it’s hard to see what the driver or reason is. Start simple, then add layers of complexity, so you can see how each layer affects the final result.”

Agree on frequency

There are multiple options for timing your forecasting, depending on your company’s needs. You will need to decide as a team how often to forecast and the period it covers. 

Rolling forecasts vs. traditional methods

“The traditional way of forecasting is to forecast the period until your fiscal year ends. So right now we’re in February, if your business year ends in December, then you have 10 months left in your forecast. But once you get into October or November, then you only have two or three months of forecast.  

A rolling forecast is different in that you always forecast a set period of time, 12 months or 16 months for example.

It has the advantage that your annual planning gets a lot easier, because you don’t have to start from scratch. You already have your forecast, and then you can just dive in. It also makes it easier for the business teams to plan the first quarter.”

Resetting forecasts

How often you forecast depends on your company’s situation. Some companies reset their forecast every month, others once a quarter. 

“It comes down to two things:

1. What’s the need of your company to keep updating the forecast? If you’re a public company, quarterly isn’t enough. You need to know before that quarter’s over, how you’re trending, if you’re meeting Wall Street’s expectations. If you’re a private company and you don’t have ambitions in terms of getting additional investors, maybe you don’t need to forecast as frequently. 

2. What do you want to get out of it as the finance team? Would increasing frequency help you make better recommendations or get better data from the other departments? If yes, then do it more frequently.”

Using AI in forecasting

Artificial intelligence is top of mind for finance leaders, and could bring real gains in forecasting efficiency and accuracy. But first you must understand how it works, and the limits of what the technology can do. If there's any sort of variance, you can't ask the model why.

Christian introduced a few concepts about machine learning that finance professionals should know to help address this problem:


“What’s exciting about machine learning models is that they can act as if the past hasn’t happened. Feed the model with information from last year, it would forecast that last year, and then you can compare it to what actually happened. Run the model and it’s a great way to see how things are going. Ask your data scientist about backtesting results.”

Feature selection

‘Feature’ in data science speak, is an input. You could think of it as garbage in, garbage out. To figure out what your ML model is doing and how to explain why it may be off, is not by looking at what the model is doing, but rather looking at the inputs. A machine is very patient, you can put in hundreds of different variables and change the inputs. By playing around, you can increase the accuracy.” 

Correlation vs. causation 

“The machine correlates data but it doesn’t mean causation. Just because two movements track closely with each other, doesn’t mean that one causes the other. There are two ways of looking at it, to decide whether there's really a causation there.”

Knowing a bit about data science will help finance professionals get the most out of machine learning for forecasting.

Will AI replace FP&A or corporate finance teams?

AI will definitely play a role in shaping the finance team of the future, but it probably won’t replace humans in finance entirely.

“Manual tasks, such as copy and pasting or creating a report, will be replaced by AI. But there will always be a need for someone to analyze the data and decide what matters for the business.” 

Another important part of finance that still needs humans? Interpreting the results and then going out and influencing people.

“Focus on your communication, leadership, and analysis skills. These skills will help you to get to those root causes, develop insights, and make concrete recommendations. Those will be the most useful skills in the future.”

More FP&A expertise

CFO Connect would like to send a big thank you to Christian for sharing his forecasting expertise with the community in this webinar. 

To learn more about anything that Christian discussed in the webinar, you can find his FP&A Bootcamp training HERE

Join CFO Connect!

CFO Connect is a global community of finance leaders. Gain access to exclusive events, connect with like-minded peers in a private Slack group, and receive curated content for finance professionals like you.

Apply for free