Sales Forecasting -- by Reps, at Least -- Is Dead
As sales and finance teams build their revenue projections for 2015, sales forecasts are under added scrutiny. That shines a spotlight on one of the great paradoxes of sales: Optimism and lateral thinking are the hallmark of a “closer” and the chief attributes of a lousy forecaster.
But if companies praise reps for aspiring to all that optimism and lateral thinking, it’s disingenuous to put those same reps in the penalty box for calling the wrong number.
Some would argue that forecasting is a necessary evil, since it drives accountability. But quotas and variable compensation take care of that. Others claim that sales reps’ first-hand deal knowledge is critical to the forecasting process. A CSO Insights report put that misconception to rest when it found that 54 percent of the deals forecasted by reps never close.
In most cases, companies are better served by letting reps sell and leaving forecasting to a new set of technologies driven by data science. Indeed, in our experience, customers have a lot to gain by replacing manual forecasting processes with an approach that focuses on predictive analytics.
In fact these processes achieve an average 82 percent forecast accuracy on a deal-by-deal basis (versus the 46 percent CSO Insights reported) and over 95 percent accuracy in the aggregate (versus the industry average of 76 percent). As a result, sales teams get back an additional two-and-a-half hours of selling time (the average time spent forecasting according to Sirius Decisions). Here are three reasons why:
1. An unbiased point of view
The benefit of forecasts driven by data science is that they aren’t distorted by the inherent biases that come with a human inputting data. Case in point: A company we know of recently deployed a set of predictive algorithms to determine which deals its sales reps would close. After a few days, the company concluded that the models had to be inaccurate. Indeed, the models indicated that the reps had a better chance of winning deals when they were competing against other vendors versus those where no competitors were in the race at all.
But when the company analysts dug deeper into the specific deals, they learned that the models were right, after all. It turned out that most of the reps who stated there were no competitors being considered did so because they hadn’t done their homework. They were unaware of competitors, budgetary factors and the customer’s decision-making process. Had this key factor not been uncovered, the team's members would have gone on thinking that a lack of competition had actually hurt their chances of closing deals.
2. An ability to anticipate what’s ahead
Zendesk is an excellent example of a company that has put the power of predictive analytics to work to drive forecast accuracy. Its business has a relatively short sales cycle, so many of the deals closed by the end of the quarter aren’t even in the pipeline when the quarter kicks off.
That there's no visibility into what ultimately closes is the kind of thing that keeps sales executives up at night. Fortunately, predictive analytics addresses the issue in a way with which manual analyses can't compete. The best predictive models project the revenue that the current pipeline will yield by determining which deals will close, and when. Then these models go a step further by projecting the revenue derived from deals that companies haven’t yet identified (sound too good to be true?).
By analyzing historical sales cycles, current deal velocity and external macro-economic factors, these models can produce results that make even the best forecasters envious.
3. An ability to constantly adapt
Once customers have deployed a predictive model, they often ask when the consultants will need to come back to retune it. “Never” is my answer. Today, the best predictive solutions rely on machines learning to retune themselves. To do this, algorithms constantly evaluate the correlation between the individual deal attributes, such as geography, industry or product, and the desired outcomes, such as win rates and close dates.
When new correlations emerge, they’re factored into the predictions automatically. So, for example, if a company decides to move into an international market, an attribute such as geography may suddenly transition from immaterial to highly relevant. Contrast the ever-evolving nature of such algorithms to the static quality of spreadsheet-driven models. It’s not uncommon for an analyst to set up such a model and then not revisit it for multiple quarters. In an environment where the rate of change is constantly accelerating, machine-learning is one of the most compelling aspects of modern data science.
Advancements in the field are forcing us to revisit a fundamental sales assumption -- that sales professionals should own the forecast. If predictive analytics deliver better results in less time, perhaps it’s time to hire a few predictive models so that reps can get back to doing what they do best: closing.
Justin Shriber is vice president of products at C9, where he leads product management, product marketing and strategy. For two decades Shriber has helped companies accelerate growth and profitability by building strategies that align marketing, sales and service with customer needs. At Oracle, he headed Oracle’s CRM OnDemand organization, and at Siebel he was one of the early pioneers of the cloud, leading product teams responsible for delivering the first generation of SaaS applications.