Measuring forecast accuracy

Anybody who says forecast accuracy isn’t important has never worked at a publicly traded company.

That’s like a meteorologist saying their weather forecast doesn’t need to be accurate.

Or a publicly traded CFO telling investors that they are also surprised by an earnings miss.

Unfortunately, many FP&A teams don’t prioritize their forecast accuracy.

The main reason FP&A forecasts aren’t accurate (aside from senior leaders changing it to fit a narrative):

They don’t take time to measure the accuracy.

But first, a story…

Early in my career, my company’s centralized FP&A team would send my boss a monthly report card on how accurate my forecast was.

What sounds like a pain was actually a blessing in disguise.

I studied the forecast evaluation criteria, reverse engineered it, then optimized for it.

Within a year, my forecast went from bottom tier to top tier in the company.

All because I was hyper-focused on improving the accuracy.

I learned that measuring my forecast accuracy helped me to see where I was falling short. And then I’d work on improving each forecasted metric until it was within an acceptable range.

I want you to have the same focus on improving your forecast accuracy by using a simple tool each month.

Below is the tool I use to measure forecast accuracy:

Forecast evaluation tool

I use this tool to measure the accuracy of my forecast for each individual metric (feel free to steal it):

Let's pretend the view above is a forecast evaluation for monthly sales.

Along the columns, you can see I have each calendar month. And down the rows, I have each forecast month (with actuals at the top and the % variance between actuals and forecast at the bottom).

Input your forecast picks into the tool

I pasted the sales output from my forecast model in the row titled ‘Jan Fcst’ in the top section called ‘Forecast’.

In January 2023, I was forecasting the months of Jan-Dec. You can see my year-end pick was 537.

Fast forwarding to the end of the year, sales came in at 532 (you can see that in the row with blue cells at the very top). This means my January forecast was only off by 0.9% for the full year. You can see that in the first row of the bottom section titled ‘Accuracy’ in the ‘Jan Fcst’ row.

How to use the tool to measure accuracy

Use the ‘Accuracy’ section of the tool to identify where your forecast consistently misses. Any variance that is more than 5% off will show up as red.

The cool thing about this view is that you can see if your forecast for the full year got more accurate as the year went on. Mine clearly didn't since the months towards the end of the year had larger misses than the months in the beginning of the year.

You can also use this section to see where results might have done something unexpected. In Oct and Nov it looks like results dropped below all of my forecast expectations. This could signal that our forecast model isn’t considering something unique about the business in Oct and Nov.

In Summary:

It’s important to have a system for measuring your forecast accuracy.

Keep it simple, but measure monthly to look for hot spots.

How will you take action on this?

See you next Saturday.

Whenever you are ready, there are a few ways I can help you:

1. Join the waitlist for Next Level FP&A, the course teaching you to grow your career by mastering the critical skills I used to go from Analyst to Director in 8 years.

2. Check out The FP&A Flywheel, the course teaching FP&A professionals at small and medium sized businesses best practices typically reserved for the highest performing companies.

3. Join The FP&A Lab where you get ongoing access to my courses, continuing FP&A education, and mentorship.

4. Looking to sponsor this newsletter? Hit reply to this email and let me know!

 Say hi 👋 on LinkedIn