An evidence-based approach to improving sales forecast accuracy

An evidence-based approach to improving sales forecast accuracy:from Building Scalable Businesses: The Inflexion-Point Blog by Bob Apollo

I was chatting recently with Thomas Oriol (a Director at Nimble Apps Limited) when he mentioned some recent research they had conducted amongst their customer base into the relative effectiveness of various forecasting methodologies. Thomas was kind enough to offer to contribute a guest article summarising their findings. I think you will find the results very interesting. Over to you, Thomas...

"Sales forecasting is a major issue for B2B companies. Because they often lack the thousands of data points that statistical forecasting techniques require, they tend to rely on judgmental methods (such as weighted pipeline or forecast categories).

But with increasingly complex sales processes, sales reps and managers are bound to miscalculate opportunity amounts, closing dates and closing probabilities. Typically, respondents to CSO Insights’ 2011 Sales Performance Optimization Survey indicated that less than half their sales opportunities closed at the amount or in the timeframe predicted...

Giving up on forecasting, however, would be doubly mistaken. In addition to the obvious resource planning benefits of accurate sales forecasts, recent research by Aberdeen Group shows a clear link between forecasting best practices and sales performance.

Looking for effective B2B forecasting methods

For these reasons (and for product development purposes), we recently tested multiple B2B forecasting techniques on aggregate SalesClic client data. Our sample included:

• 12 sales teams
• in the US, UK and Asia
• in the software, electronic equipment and financial services industries
• managing structured sales pipelines (i.e. following a step-by-step sales processes)
• with sales cycles of 75 to 250 days
• totalling 144,817 closed opportunities

We tested 8 forecasting methods, ranging from the basic (the standard weighted pipeline) to the sophisticated (various types of decision trees). These techniques, as well as our methodology and results, are described in greater detail in this SalesClic white paper.

Interesting patterns

We found two interesting sales forecasting/performance patterns in our data.

1: Closing dates are always optimistic

Initial closing dates were optimistic for 10 teams out of 12 in our sample. On average, it took 22% longer than initially expected to win an opportunity for our sample teams.

Such optimism is an absolute disaster in a B2B context, where sales forecasts are very sensitive to closing dates. For sales managers and sales operations managers, monitoring closing dates is the clear priority.

Not surprisingly, the bias increases with the length of the sales cycle. In our sample, we found a closing date error of 16% for the team with the shortest sales cycle, and of 109% for the team with the longest sales cycle.

2: Losing takes longer than winning

In our sample, two thirds of closed-lost opportunities were lost after the closing date initially expected, and losing an opportunity took an average 1.7 times longer than winning one.

Stagnation in the pipeline is clearly a bad omen for opportunities, and B2B companies have much to gain from detecting “stuck” opportunities as early as possible.  

Winners and losers

So how did the methods we tested fare?

Judgmental forecasts are not reliable - really

Our research confirms what most sales managers already know: the simple weighted pipeline forecasting technique (declared amounts, declared closing dates and declared closing probabilities) is useless. It was the worst performing forecasting method in our test.

Leveraging historical data helps

Traditional CRM software is bad at leveraging your historical data for decision support purposes. That is unfortunate, because our research shows that replacing sales rep and manager judgment on closing dates and closing probabilities with historical averages increases forecast accuracy by up to 35%.

Sales reps and managers are sitting on a treasure trove of forecasting information. At a minimum, they should use it to inform their judgments - ignoring it is a crime!

Sophistication pays... up to a point

The sophisticated linear predictors we tested did not perform well. This is disappointing but not surprising since sales pipelines can have widely different shapes, and linear equations are ill-equipped to deal with such irregularities. This is a typical difference between B2B and B2C forecasting.

Decision trees, however, proved quite efficient: compared to the simple weighted pipeline, they increased forecast accuracy by an average 46%. We have started working on the implementation of a nimble version of this technique in SalesClic.

A surprising winner

But the best forecasting method in our test was a surprising one: the “daily closing rate”. Here is an example of how it works:

• Suppose your team has closed €300K over the past rolling 3 months
• That is a daily average of €3.3K
• Suppose that you are 30 days into the current quarter (1/3 of the quarter)...
• ...and that you have closed €50K so far
• Your “historical daily closings” forecast for the quarter is: €50K + €3.3K x 60 days = €250K

This technique improved simple weighted pipeline forecasts by a staggering 53%. Not bad for such a simple calculation - although we acknowledge that there may be challenges in applying it to sales pipelines with a relatively small number of large opportunities. We have nonetheless decided to add it to SalesClic’s “forecasting box”.

What should you do?

Our research suggests immediate ways for B2B companies to improve sales forecast accuracy:

• Pay close attention to “pipeline dynamics” (opportunity time-to-wins, pipeline stage durations, closing probabilities by pipeline stage) and use that information for forecasting purposes
• Calculate “daily closing rates” and use that information for forecasting purposes

Based on our experience with SalesClic clients, we would also recommend:

• Averaging forecasts derived from multiple forecasting methods
• Using sales simulations when confronted to “elephants” (large opportunities that could make of break your forecast)

Again more information on the research underlying these comments is available in the corresponding white paper."

About the author

Thomas Oriol is a Director at Nimble Apps Limited, the Dublin-based publisher of SalesClic. SalesClic is a sales pipeline visualization, analysis and forecasting solution for B2B companies. SalesClic integrates with Google Apps, Highrise and Salesforce.
Prior to founding Nimble Apps, Thomas spent 10 years advising European technology companies on mergers and acquisitions. Thomas holds a MS in Political Science from Sciences Po Paris, an MBA from ESSEC Business School and an MPhil in Economics from Sciences Po Paris - and can be reached at t.oriol@nimble-apps.com and @thomasoriol on Twitter.