Sales forecasting is hard. For proof, you need look no further than the 2018 CSO Insights Sales Performance study, which reported that on average a little over 46% of all forecasted sales deals actually resulted in a win (never mind the timing).
Even the top performing sales organisations did only marginally better – at just under 54% forecasting success rate on a deal-by-deal basis. Now, there are obvious reasons why accurately predicting the outcome of every complex buying process is fraught with difficulty.
But it’s hard to avoid concluding that we ought to be able to do better. And I’m going to suggest that one of the reasons that organisations struggle to do better is down to simple statistical naivety…
It’s become a cliché: the convention that CRM systems assign a percentage to each stage in the sales pipeline. Those percentages invariably rise with each succeeding stage. And those percentages are typically used to make judgements as to where a deal is likely to close or not.
Now, if you’re in a very high volume, short sales cycle and relatively low transaction value environment, and if you carefully analyse the actual outcomes of opportunities at each stage in a way that is statistically significant, I’ll grant you that it can be possible to make a reasonably accurate prediction of the likely overall revenue outcome.
But I’d suggest that there is absolutely no chance of using the same crude “average probability” principles to accurately forecast outcomes on either a deal by deal or overall revenue achievement basis in the relatively low volume, long sales cycle and high transaction value environment that is typical of the clients I work with.
The maths simply doesn’t add up!
And if we zoom out and take a “big picture” perspective of allthe vendors that are competing for a given opportunity, the maths becomes even more ludicrous. Let’s say that 3 shortlisted vendors are in the final selection phase. Let’s imagine that each of the vendors has established a 50% probability for that stage in the process, and that the same opportunity is sitting in each of those vendor’s forecasts at that percentage.
And now let’s factor in that CSO Insights finding that fewer than 50% of those forecasted deals are actually going to buy anything. Multiply those three factors together assuming that each vendor has an equal chance, and the actual average probability of getting an order isn’t 50%, it’s somewhere around 16%. And that assumes that using the same average for every deal at a given stage is a good idea, and that every vendor has an equal chance of winning – two unsupportable assumptions.
Flaws in the system
Please allow me to enumerate just a few of the things that are wrong with the idea of assigning the same fixed percentage to all deals that are at the same stage:
Let’s start by acknowledging the common confusion as to whether the chosen percentage figure is an attempt to assess the probability that any opportunity in that stage is likely to close, or whether it simply (and probably more accurately) reflects how far the opportunity has progressed through the sales cycle.
If the percentage isintended to reflect probability rather than progress, then how have we come to establish the percentages? Did we (as is far too often the case) simply inherit the default percentages that were pre-configured by the vendor of our CRM system? Or have we made a genuine attempt to measure the actual success rates from each stage? And have we monitored and adjusted those average percentages to reflect changes over time in our actual success rate?
But in truth, unless we have the benefit of relatively high opportunity counts, even having an accurate average success rate is still going to cause uncomfortable variations in the actual outcomes. That’s because our chances of winning are not just influenced by the stage the deal has reached.
Key predictive factors
I’ve found that three factors are particularly important in making better judgements about our true chances of winning any given opportunity:
- The type of opportunity: is it a brand-new customer, a new project in an existing customer, or an upsell or upgrade to an existing project? For reasons that should be obvious, our chances of winning (and our confidence in close dates) will differ dramatically depending on this factor
- The probability the customer will do anything: given that half of all forecasted projects actually decide to do nothing, what is our confidence that the customer will place an order on anyone? This factor is closely related to the perceived urgency of the project compared to all the other investment options the prospective customer is considering
- The probability the customer will choose us: it is vanishingly rare for all the finalists to have an equal chance of winning. One of the vendors almost always has some level of advantage. The rational basis for our assessment might include: Did we shape their vision, or respond to their RFP? Have we engaged all the key stakeholders or are we reliant on one individual? Have we established (and has the customer acknowledged) any unique differentiators or is the playing field actually rather level?
If we can accurately determine where we stand with regard to these three dimensions, based on evidence rather than hope or supposition, then our forecast is likely to be far more accurate at both an individual opportunity and aggregate pipeline level. There is one more key factor – our confidence in the projected close date – which I wrote about here.
A more intelligent approach
So how can we break away from the tyranny of standard stage forecast percentages? I really like the approach taken by Membrain – which can either be implemented as a standalone CRM or as a module within salesforce.com.
Rather than a single fixed percentage, each stage is represented by a “collar” or range – a minimum and maximum percentage. And then each individual opportunity is individually scored according to a handful of the most relevant factors – such as the ones above.
The chosen percentage factors can then be validated and where necessary adjusted by identifying the key predictive variables and measuring the actual outcomes – Membrain also includes all the required analytics.
It’s a much more intelligent approach, and if you’ve ever been frustrated by the limitations of the conventional pseudo-probabilistic approach to forecasting, you’ll find it a breath of fresh air (and a welcome dose of reality).