Sales Forecast Dynamics
The technology needed to fatten a sales pipe is not the garden-variety spreadsheet or even BI -- each is inherently backward-looking. The ideal technology should mimic the way a human brain learns, by evaluating past scenarios and applying the learning to current situations to arrive at a true forecast. The technique in question is machine learning for total revenue intelligence.
Take a look at almost any sales software on the market, and you'll find the vendor's claim that the product accelerates the sales process. It's a given. For years, we've been trying to achieve this acceleration, and we've been succeeding. The only problem with this approach is that it can't go on forever. It's a form of expecting infinite growth based on limited resources.
In this case, the limited resource is the human brain. The degree to which we can realistically expect to accelerate a sales process is directly dependent on our ability to accelerate the human decision-making process -- and that is not infinite.
Acceleration was easy once. The railroad and telegraph accelerated selling by reducing the time between touches when information could be exchanged. The car and the telephone did the same thing, and so did the fax machine, but with it a competing factor comes into play.
Someone has to pick up and deal with a fax in order for it to outpace a phone call, and the same is true for email, texts and tweets. These are all asynchronous technologies, though, meaning they can be answered at leisure -- or in the case of a business deal, when the customer has something to say, which might need to wait for an internal meeting or other deliberation.
So, it's clear that acceleration is reaching an asymptotic limit. We got the first 80 percent without much difficulty, but the last 20 percent represents significantly diminishing returns. That's cold comfort to anyone with a pipeline and a forecast to deliver on, but it doesn't have to be the end of the story.
Skinny Is Risky
Let's consider two real-world pipes -- not the sales variety but the plumbing kind. Flow in cubic units is defined by an equation with some constants like Pi; however, it basically multiplies pipe diameter by velocity of the fluid, though it's an inverse proportion. In other words, you can get the same flow from a fat pipe running slowly as you can from a skinny pipe running faster.
Now, apply this to the sales pipeline (an aptly chosen term). Since forever, we've looked at the pipe and never considered that the diameter could be increased; we've only sought ways to increase flow velocity -- hence, the extravagant vendor claims. However, if our ability to affect velocity has reached a limit, then we need to consider how to make the pipe fatter.
It's a good and useful question to ask. Historically, we haven't tried to make the pipe fatter, because it requires a great deal of data and computing power. More deals mean more data, and there is also an asymptotic limit to the number of deals a person can keep in mind.
Don't think about the sales rep alone -- consider the plight of the sales manager. A manager with eight reports, each tracking 50 deals in various stages, has to ride herd on 400 deals. That's a lot. Under these conditions, sales update meetings are more like reminder meetings, and who recalls the status of each deal last week or last month? So how do you compare and assess progress?
No wonder that as the quarter rolls on, the pipelines narrow. They get skinny -- and very risky if one or more deals don't come in, because that's all that's manageable. They don't have to get skinny and risky if we can apply compute power to track all the deals, though.
While we're at it, let's not simply and passively track what is. Instead, let's use compute power in the form of machine learning to develop probabilities and scenarios for deal closes. Doing that provides us with the fatter pipe that we need in order to match the somewhat slower and eminently realistic nature of the real buying process.
Machine Learning for TRI
The technology needed to do this is not the garden-variety spreadsheet or even BI -- each is inherently backward-looking. The ideal technology should mimic the way a human brain learns, by evaluating past scenarios and applying the learning to current situations to arrive at a true forecast.
This isn't perfect, and it will not by itself generate more closes. It will only assign probabilities to specific deals, assuming your current process is diligently applied.
Think of what that means, however. If you can get a good idea of the likely result of the track you are on, you also can begin to tinker with the inputs to see if changing approaches can materially affect the outcome.
The technique in question is machine learning for total revenue intelligence as practiced by vendors like Aviso and others. In addition to enabling managers to get a better handle on the whole pipeline and thus fatten it, machine learning also can help you to selectively upgrade individual deals to get some of them across the finish line.
This is not the same as accelerating individual deals. Still, by increasing the quality of deals in the pipeline -- that is, by fattening it -- you gain the ability to throttle deals when it makes sense while improving flow, which for the bottom line amounts to the same thing.