Sales Gets a Machine-Learning Makeover

We live in a data-saturated world where a great many of our interactions with other humans happen online. It makes sense then that one of the most human of business activities — sales — is currently undergoing a digital renaissance. While the sales function has historically relied on metrics, today there is far more sales-centric data, and far richer data, than ever. It comes from social media, from website interactions, and from A/B tests, just to name a few.

To help make sense of all the available data and to improve sales effectiveness and efficiency, organizations are turning to machine learning. Smart machines are becoming trusted sidekicks in sales departments as they make opaque processes more transparent, provide analysis to inform decision making, and offload low-value tasks.

In our survey of executives at 168 large companies with at least $500 million in annual revenue, 76% of respondents said they are targeting higher sales growth with machine learning, the kind of artificial intelligence software that continuously learns from big data and optimizes recommendations in real time to sales staff. Moreover, more than two out of five companies have already implemented machine learning in sales and marketing.

Our research shows large companies are applying machine learning to sales processes along three dimensions, each of which adds algorithmic rigor to human intelligence and intuition, creating a dynamic new formula they hope will boost sales. The first dimension allows for a scientific approach — with data and clarity of process — in sales interactions. The second enables more data-driven experimentation within a sales and marketing environment. The third uses science to create more time to sell, that is, by automating the administrative tasks that get in the way of managing accounts, finding leads, and closing deals. Taken across these three dimensions, machine learning is creating speedier, more scientific processes for generating sales revenue.

Before machine learning came along, static databases, analytics from historical data, and experience and instinct steered execution — with performance improvements coming in set increments over time. With machine learning, real-time data can drive actions and process change along a continuous path. Hypotheses can be quickly formed, tested, and revised, enabling a new kind of workflow that can dramatically outperform previous ones. In our survey, 38% of respondents credited machine learning for improvements in their key performance indicators for sales — such as new leads, upsells, and sales cycle times — by a factor of 2 or more, while another 41% created improvements by a factor of 5 or more.

Bringing Science to Analyzing Social Cues

Historically in sales, a field rep might meet with a potential customer face-to-face and read non-verbal cues like nodding or frowning to determine his or her next move. But in a digital world — without the benefit of physical social cues — sales becomes an opaque process that can be difficult to deconstruct. If a prospect doesn’t work out, it can be hard to find the errors that could be corrected in future attempts.

But what if a salesperson could know with confidence when a potential customer is ready to buy? One company called 6sense offers a product that provides digital predictive buying signals to help sales professionals pinpoint the optimal time to approach prospects. By analyzing the online behavior of potential customers who visit a client’s site — as well as third-party data from a variety of publicly available sources, including social media — 6sense provides a better picture of interest and if and when a potential customer might be ready to buy.

The company analyzes website data on a large scale, using machine learning to fine-tune predictions. With the right data in hand, sales teams can identify prospects more quickly, while targeting sales pitches at the right time, with a higher likelihood of success. With more data about potential customers, sales professionals have the ability to test different approaches, spending more time fine-tuning pitches rather than chasing false leads.

Data-driven Sales Experiments

Machine learning can also enable more effective A/B website testing, eliminating the bottlenecks often associated with sales experiments. Fewer bottlenecks means more speed: One-third of our survey respondents say they have accelerated sales processes by double or more, while another one-third claim increases of 5 times or more. Adobe Target is one software tool that allows non-technical salespeople and marketers to quickly modify websites to deploy large numbers of A/B tests. Based on data from website interactions, machine-learning algorithms find and suggest the best content to tweak, as well as help sales and marketing staff validate assumptions after developing a test.

A startup called Optimizely uses machine learning to run A/B tests on pricing strategy. In an experiment with marketing firm Bizible, Optimizely integrated its experimentation software with Salesforce. The result was a dashboard that displays experimental variables — the original prices and the test prices — as well as information on lead, contact, case, etc. Software ensured that pricing was consistent across a range of IP addresses so that potential customers at the same companies would see the same prices. The test only ran 30 days, but the results were compelling. New higher prices led to fewer opportunities, but those opportunities brought in 25% higher value on average.

When intelligent automation is used to facilitate science within organizations, it provides opportunities to test new actions and processes that can lead to revenue growth. Machine learning can act like an assistant at the lab bench, logging data and suggesting new experimental approaches. It can illuminate previously opaque processes and free salespeople and marketers to build their own experiments with clarity and confidence.

Automated Science for Sales Efficiency

Another important aspect of machine learning is that it can optimize processes behind the scenes. More than nine out of ten companies in our survey said they agreed that machine learning is improving processes in real time without human intervention.

Algorithms can conduct automated science experiments with data as it arrives without any human intervention. In the case of sales, machine learning can minimize time spent on administrative tasks and eliminate steps that take time away from interactions with customers. The end result can be a significant reduction in sales cycle times.

Historically, many sales and marketing teams have attempted to increase efficiency with one-off tricks that prove difficult if not impossible to scale — writing their own macros or personalizing spreadsheets, for instance. Conversely, machine learning algorithms that automate administrative tasks or provide just-in-time customer predictions can be easily standardized and implemented across teams.

Gainsight, a company that offers software to manage sales and customer service more effectively, helped the online questionnaire service SurveyMonkey create automated alerts to ensure that all team members were up to date on renewals, invoicing, and upsell opportunities. Using Gainsight’s technology, SurveyMonkey cut the process time to send an invoice by about a third.

Another company called Anaplan is helping Hewlett-Packard reduce the time spent gathering sales data from a month to three days, effectively a 10-fold improvement. Instead of churning through month-old data, sales teams can make decisions with fresh, up-to-date analysis, allowing sales staff to spend time on higher-value tasks. Similarly, a machine-learning company called Aviso, working with an enterprise cloud company called Nutanix, can compress a 12-hour task of compiling sales reports into four minutes. That’s a 100-fold improvement.

Whether machine learning facilitates analysis, experimentation or automation, it provides real value for sales and marketing teams. In some cases, salespeople and marketers gain confidence and clarity in processes that were previously opaque, enabling a more systematic and consistent approach to client interaction. In others, machine learning runs the experiments behind the scenes by pruning processes and allowing salespeople to attend to higher-value tasks. While we remain in the early stages of bringing the full value of machine learning to bear in sales (and elsewhere in the organization), what’s clear already is that machine learning holds the potential to find significant hidden revenue where there were previously only marginal gains.