I don’t know anyone who was ever great at sales forecasting. This is not to say that it can’t be done but it’s a hard problem, like forecasting the weather. Today they can tell us within a few degrees the weather for many days in advance because we’ve invested in many data gathering and analysis tools like satellites, computer modeling, and analytics.
It’s not that weather forecasting has been perfected though. Instead, forecasters have been able to bring the forecast down to a range of probabilities through the use of all that data and models for how the oceans and atmosphere, the jet stream and other major parts of the earth work. So we typically see something like a 30 percent chance of rain and a temperature range of about 4 degrees and those ranges are enough to enable us to make plans like going to a picnic or not, or remembering to carry a hat or umbrella.
Sales forecasting has none of this going for it at the moment in part because while a sales forecast is a much smaller problem than the weather five days hence, we still do it much as we did a hundred years ago. Too often that means a senior person reviews all the deals, applies his or her experience and follows a gut instinct. By the way, that person is known affectionately as a HIPPO because the decision was the result of the Highest Paid Person’s Opinion. How’s that for science?
But what would the forecasting process look like if we had as much science employed as a local TV weather person? Like the weather person, the sales manager would have a model based on many iterations of patterns that account for all kinds of variables including deal stages and their inputs, sales rep and that person’s track record, the product and its adoption patterns and much more.
If the interest is in total revenue intelligence, the model would also include inputs from the current customer base including propensity to churn, cross-sell and upsell opportunities, and customer lifetime value. To manage all this there would also be powerful analytics to process all the data and the result would not be a single answer such as a dollar number or the determination of whether or not you’ll make your target.
The output would be a set of ranges. Just like the weather forecast you don’t know until the forecast is in the past, so there’s no such thing as 100%. But also, like a weather forecast you don’t need absolute certainty, you simply need to know the probability of rain on Saturday so you can figure out if you want to golf.
Right now human behavior plays a big role in sales forecasting. We have a number to shoot for and often we try to figure out ways to make it regardless of soundness of our reasoning. So we assume deals are going to close even if it means we have to ignore telltale signs of trouble until it’s too late. But imagine using modeling and analytics to evaluate your company’s position in each deal. The model can tell you the warning signs because with analytics it can tell you how closely any deal fits with the model’s known history of success. As with weather forecasting there’s no value judgment just a probability of rain. Managers still need to apply their reasoning. But armed with this kind of knowledge sales people up and down the organization can evaluate scenarios constructed to make their deals match up with the ideal.
Here’s a trivial example. A company I know has a sales process, which is a model itself, which at a certain point has the rep meeting with a key decision maker. If the key decision-maker is not onboard they know the chances of getting the deal done are greatly reduced, so the meeting is an important gating factor. Yet, time after time, reps skip this step. Perhaps the coach in the deal reassures the rep that the decision-maker has already approved, or maybe the decision-maker doesn’t have time to see the rep, or any of a hundred other reasons.
But it might be hard for management to discover this, if for instance, the rep simply indicates on the forecast that the deal is at forecasting stage. A manager with 5 reps each working 50 deals in various stages might not have the time to go through every piece of deal data so the mistake would get through but it wouldn’t improve the forecast one bit.
Now, take a more scientific approach. The sales process stage is one of the variables in the model that the company uses to evaluate deals. More importantly, the business applies analytics to the deal data rather than expecting managers to review all of it. Analytics would have no trouble spotting the incomplete deal stage and would down grade the forecast appropriately. A report would then show the variance between the forecast and the model.
If you apply this logic to every deal in the forecast then you have that range of probabilities that weather people rely on to tell you about Saturday’s conditions. More importantly, and unlike a weather report, the forecast is also prescriptive because it shows you how you can improve it. Get that meeting with the decision-maker if you want to close the deal!
Mark Twain once said, “Everyone complains about the weather but no one does anything about it.” If you’re tired of complaints about sales forecast accuracy, consider building an accurate model and applying analytics. It works for the weather.
A couple of industry heavy weights have pooled their talents and assembled $8 million in funding resources to launch Aviso, a new kind of analytics company this week. K.V. Rao, co-founder and CEO, and Andrew Abrahams, co-founder and CTO, are Ph.D. scientists taking on risk assessment for modern enterprises. Rao, has a string of successful tech startups including founding Zuora, the subscription commerce company. Abrahams recently retired from JPMorgan Chase where he was “Firm-wide Head of Quantitative Research and Model Oversight, reporting to the firm’s CRO,” according to Jeff Yoshimura, the company’s CMO. Believe me that’s just the tip of two rather large icebergs.
These two talents raised an A series led by Shasta Ventures; First Round Capital; Cowboy Ventures; Bloomberg Beta; Subrah Iyar, founder and CEO of WebEx; Roger Sippl, founder and CEO of Informix and Vantive; Dave Hersh, founding CEO of Jive Software; and Ron Gill, CFO of NetSuite.
So what makes Aviso special?
First off, I like the way they were in stealth mode for two years while they made version one bullet proof. They wouldn’t even give me a briefing under NDA. Second, their initial customer list reflects a bunch of thinkers that want more advanced analytics and are willing to roll up their sleeves to work on a good idea. They include: RingCentral (RNG), Saba Software (SABA), FireEye (FEYE), Damballa, Replicon, and Zuora.
Third is the approach to the product itself. While most analytics packages work with a limited dataset that has to be refined and scrubbed, Aviso takes a different approach. Reflecting the founders’ interest in a broader range of influences on business results, Aviso uses a portfolio model that can just as easily interrogate company-wide data as it can also use market data about general economic trends before it serves up its results.
So, for example, if I understand what’s going on, a sales forecast should be primarily based on sales data that a company’s sales team collects. But it ought to also reflect what’s knowable about the general economy, but often not included. If for instance, the economy is heading for the exits, say job creation is down, interest rates and inflation are up, and just for fun, new housing starts are going sideways, how secure do you think your 80 percent probability sales orders are six weeks before the close of the quarter? More importantly, which of your deals conceals hidden risk? Hmmm? That’s the kind of scenario that is totally resistant to sleeping pills.
As you might expect, gathering all of that outside data might import a great deal of noise into your analytics process and noise can be the enemy of a strong signal. This has been the bane of data scientists since the earth cooled. But it’s also a problem the founders are quite familiar with — you don’t run Quantitative Research and Model Oversight at JPMorgan Chase unless you have a strategy for dealing with noisy data.
Rao tells me that back analysis of old data plus the work they did in beta shows great promise and I believe him. The immediate goal is to focus on revenue forecasting with what they are calling Total Revenue Intelligence to provide insights into all types of revenue streams in an organization. This tells me Aviso will be useful in both conventional and subscription companies but most especially in the years ahead in companies that are supporting a hybrid business model consisting of their traditional business and a new-fangled subscription business.
Think about it, one of the hardest things in business is changing your business model especially from something that brings in whales to something that uses a big net to capture sardines. Wall Street doesn’t like downside revenue surprises of the kind that early subscription businesses can provide. So a portfolio strategy that identifies risk across a company’s whole revenue stream makes a lot of sense. We’ll see if Aviso is the tool to provide it. So far, the early promise seems to be panning out.