The Best Forecast Methodology

The best FP&A/Finance professionals I’ve ever worked with have 1 thing in common:

The are relentlessly focused on the ‘why’:

• Why did revenue increase this month?

• Why did our expense forecast miss?

• Why did we capitalize more than what accounting told us?

Without knowing it, they are actually working on building a world class forecast.

When you ask ‘why’ you are really looking for the drivers.

And when you understand the drivers of every P&L account in the business, you have a forecast that is air-tight.

It’s why I’m such a fan of driver-based forecasting as the north star.

It forces you to get in the details, understand what’s really going on, and then building models that help the business tell their story.

Aside from having a more accurate forecast, driver-based forecasting gives you the tools to analyze and storyline why your forecast missed down to the specific business activity.

Let’s jump into driver-based forecasting and why I’m such a fan (when used correctly):

What is driver-based forecasting?

The best way to learn driver-based forecasting is by thinking about your personal budget (you should really use fina.money).

[Let’s use your monthly retirement plan contributions as an overly-simple example]

A non-driver-based way to forecast your retirement plan contributions would be to look at your historical monthly contributions in dollars and project that going forward.

If last month was $400, then next month will be $400.

The downfall of this method?

• What happens if your contributions in dollars change from what you expected?

You won’t (easily) be able to explain why your contributions changed from what you expected by simply doing a trend pick.

Now let’s do the driver-based methodology…

Using basic logic, your retirement contributions are driven by 2 different things:

1. How much income you make each month

2. What % of your income you invest each month

Each of those is the driver.

They are both knowable data points that allow you to calculate the result that might otherwise be a mystery.

Driver-based forecasting is simple when you think about it in these terms.

But when you flip it into the business world, it can become a lot more complex.

Think about what drives retirement contributions from the company’s perspective…

• How many people participate

• The average participation rate

• The average salary of those participating

And in a large corporation with multiple business units, product lines, and revenue streams - this will be much more complex.

Why driver-based forecasting is better:

Why should you pivot to driver-based forecasting? The answer lies in its core advantage: relevance.

Relying on historical data to make future picks isn’t bad - but it can be limited in organizations that are rapidly growing and changing (what company isn’t right now?!).

Driver-based forecasting ensures you are leveraging relevant data based on what actually impacts the business, rather than leveraging out-of-date data and information.

It’s the definition of moving from reactive to proactive. Rather than leveraging historical data, you can leverage future expected data (more on this in the next section)

Here’s a few practical benefits for companies if you are trying to get your CFO onboard to do more driver-based forecasting:

• Increased Accuracy: By focusing on real-time drivers, forecasts become more reflective of current and future conditions.

• Enhanced Strategic Planning: Aligns forecasts with strategic business objectives, enabling better decision-making.

• Resource Optimization: Helps in identifying areas of inefficiency and redirects resources to high-impact areas.

You might be wondering ‘but aren’t I still just making trend picks on the drivers rather than the final number?’

When you are dealing with knowable drivers, you can easily work with your business partners to get projections for those drivers (more on that in the next section).

But in a lot of cases you are making guesses on drivers that are not knowable.

Things like average retirement participation rate is not knowable - it’s impossible to know what all your employees are going to collectively do with any certainty.

But the beauty comes in your ability to analyze at the driver level - not just forecast.

Take the example below of a company’s retirement plan contributions for the month:

We forecasted at the driver level that resulted in $1.47M in contributions, but the actuals came in $260k short.

Why?

Because we forecasted at the driver level, we can quickly determine the attribution of which driver caused the variance.

It’s clear that the average participation rate dropped below expectation and caused our forecast miss. This allows us to study that specific metric and ask what might have caused this.

How to implement driver-based forecasting:

Hopefully the above example begins to spark some thoughts for how you can implement driver-based forecasting.

But the real key to effective driver-based forecasting is picking the right metrics and drivers, gathering the relevant data, and developing the model.

1. Identify Key Drivers:

   Pinpoint the most influential factors affecting your business performance.

   It serves no purpose to spend effort building a driver-based model for a metric nobody cares about.

   Spend time understanding what P&L items are both important and difficult to forecast. This will give you a great place to start.

   Pick 1 P&L account to understand each month (less is more when going deep).

   Within each of those P&L items, think critically about what causes them to change each month. Every business is different, but a good place to start is asking ChatGPT ‘what are the main drivers of [key metric] in [industry]?’

2. Data Collection and Analysis:

   Gather historical and expected future data for each identified driver in the last step.

   You are likely going to come across 2 different types of drivers: knowable and not knowable.

   We already covered the second one above, but for any drivers that are knowable, you’ll want to work with your business partners to understand what they expect to do in the future with those drivers.

   For example, marketing spend is a core driver for online sales in B2C companies (nike, amazon, etc.). And your marketing team probably knows exactly how they are spending the next few months of budget.

   Every month you should be meeting with them to build your online sales forecast based on how they are spending their budget.

3. Model Development:

   Create forecasting models that incorporate these drivers.

   This might mean an extra row in your FP&A corporate model or an entirely different side model. It really depends on the complexity of the drivers you’ve chosen

   I recommend for anything not knowable that you avoid building anything too complex then work over time to continue improving.

Like I’ve mentioned throughout, the real benefit is analyzing why your forecast missed.

So make sure you spend enough time each month studying your picks and improving your understanding of the business.

In Summary:

Driver-based forecasting gives you so much more than a better forecast - you can more clearly articulate why your forecast was off.

It’s a huge credibility builder.

Just make sure to not over-build driver-based forecasts so you can move quickly every time you need to update your forecast.

 Say hi 👋 on LinkedIn