The state of forecasting in today’s businesses is such that only 1% are able to achieve 90% forecasting accuracy 30 days out.
I think we can all agree that’s sub-optimal, but why is this the case? In this post, we’ll talk at a tactical level about the nine most-common mistakes that might be preventing you from accurately predicting your future business performance.
As you read through this list, consider which ones apply to you. Out of those that apply, which impacts your planning and performance most? What have you tried to correct it?
Here are the nine most-common forecasting mistakes to avoid:
1. Lack of a systematic and process-driven approach.
This is the number one forecasting sin. Many of the mistakes below could be solved by taking an unbiased and replicable, process-driven approach to forecasting, however, a minority of business has achieved a truly scalable and automated approach to forecasting (which should ideally include a complete view of all influencers and drivers affecting all business functions).
2. Taking the easy way out by peanut-butter spreading.
In the absence of a process-driven approach, or the computational power required to build forecasts at a very granular level, most staff will default to evenly distributing one high-level forecast across different dimensions of their business.
An easy analogy to see where this creates a world of trouble is if you’re out to dinner with a group of friends, and you order water and an appetizer while they order steak and wine. Then, when the bill comes, they ask you to split it evenly. Does that seem like the best approach to you?
3. Expecting to aggregate poor forecasts into one good forecast.
Many enterprise forecasting approaches call for functional leaders to prepare their own, independent forecasts, so they can then be aggregated by finance. The thinking is those who are closest to the numbers today are most likely to know what reality will be tomorrow. While it’s true a functional owner may be able to spot false assumptions faster than someone removed from day-to-day operations, this doesn’t necessarily equate to higher-quality forecasts (see below).
4. Massaging data to fit a preferred storyline.
Forecasting is meant to provide a view of what your performance will be in reality. It should be ruthlessly pragmatic and matter-of-fact. However, this in conflict with the natural incentive we as staff have when forecasting the metrics we’re also responsible for attaining.
As an operations manager, I might be tempted to sandbag delivery times to manage expectations. As a customer success manager, I might forecast a higher churn rate to anchor the targets for my variable compensation at a more attainable level.
5. Relying on tribal knowledge over statistics.
While massaging a storyline might be a deliberate choice, it’s easy to let bias creep into our process without realizing it. Sales may assume headcount is the greatest predictor of revenue attainment because that’s how the forecast has always been made. The next sales director might inherit that model and also assume it’s correct, thereby propitiating a legacy of bias.
In other words, forecasts are often created through a particular method for no other reason than, That’s the way it’s always been done.
6. Misleading, spurious correlations.
This is another silent killer. We’ve all heard the saying, Correlation isn’t causation! and yet, whenever we see two parallel lines moving in tandem or a tight group of colored dots, our first reaction is to assume causation. Even when basic statistics are used, it can be incredibly difficult to establish true causation.
For a comedic take on this reality, check out how the divorce rate in Maine correlates to per-capita margarine consumption, how computer science doctorates correlate to arcade spending, and other hilarious spurious correlations.
7. Different functional groups applying different methods.
As an offshoot from a prior misstep mentioned, when cross-functional leaders are tasked with creating their own forecasts without a common source of truth and shared process, different methodologies applied to different datasets at different levels can not only make for divergent predictions, it ignores the interdependencies each area of the business has on the other.
If marketing is forecasting lower lead volume, yet sales is telling support it’s going to land more customers than the annual plan calls for, it can result in real confusion (and very awkward leadership meetings).
8. Inability to efficiently procure and structure data.
The first step to creating a forecast or modeling a decision scenario is, obviously, preparing data. But if you ask anyone who’s recently created a forecast, they will tell you identifying the right sources of information, getting access to it from other departments, cleansing and structuring the data, and doing it all over to keep their model refreshed and updated can consume two to three full work weeks. Easy.
By this time, a window of opportunity or timeline to execute a decision may have passed you by.
9. Neglect of external data to inform long-range forecasts.
Another data-related challenge is the neglect of external datasets like weather, events, and economic conditions, which can improve the quality and accuracy of long-range predictions.
As a somewhat extreme example, if you’re running an airport hotel and there’s going to be two unprecedented blizzards within a single month, it would be career-enhancing behavior to ensure your revenue, staffing, and projected inventory all reflect this.
While these nine mistakes top our list, it could no doubt spiral to include dozens more. What’s another forecasting misstep you’ve seen play out in your business that we’ve not included here? What impact has this had on you and your team?
