Using AI to Boost Sales Forecast Accuracy
Ever since there have been sales, merchants (and, later, sales executives) have tried to accurately forecast sales and identify the prospects most likely to make purchases.
It’s a noble purpose. After all, accurately forecasting sales would help ensure that the business had the right amount of inventory in precisely the right place at precisely the right time. Knowing who is most likely to make a purchase would help ensure that companies are devoting precious sales resources to the right targets and avoiding those that have no intent to buy.
These efforts started with intuition and gut instincts, then became more refined as technology evolved. Now artificial intelligence is among the latest technologies companies are using to accurately forecast sales by determining which prospects have a better chance of converting to customers.
But artificial intelligence can’t do anything for companies if they don’t do some initial legwork, warns Christian Wettre, senior vice president and general manager of Sugar Platform at SugarCRM.
“To have a sales forecast, you need to have a sales methodology. AI will use the sales methodology in a model,” Wettre explains. “You need to have that documented, and you need to have your people trained. You need a commitment to actually engaging with the CRM and the system where AI is applied and using it and [deciding] whether you apply AI or not.”
Additionally, management has to ensure that salespeople are applying the company’s sales methodology and validate that sales expectations are based on some known metrics (such as the numbers and levels of engagement with buy-side decision makers), according to Wettre.
Customer engagement is one of the best predictors of potential prospect conversion, according to Wettre. So to get the biggest lift from AI, organizations should document engagement with prospects, paying particular attention to the breadth of the engagement, the levels of the people who are engaged, where in the sales cycle engagement is occurring, the channel used in the engagement, and more.
“One thing that people typically get wrong about sales forecasting is that they think it is about calling a number,” says Frank Dale, Salesloft’s senior vice president of product management. “That is certainly important. But sales forecasting is as much about building an action plan as it is about calling a number. That’s often missed when you talk to people in sales forecasting.”
Many organizations make sales forecasts without sufficient action plans to hit their numbers, according to Dale. “Forecasting can’t just be an analytical exercise. It has to be deeply connected to action,” he asserts.
To come up with good forecasts, companies need a disciplined cadence, reviewing the sales pipeline, updating the key forecast elements, closing dates, the confidence percentage, the committed deals, and similar metrics, Wettre says.
“At the end of the day, all sales forecasts are basically just abstracted out of estimates of what’s going on between the buyer and the seller,” Dale adds. “You have to have visibility into what goes on between the buyer and the seller—all of the interactions, whether phone calls, meetings, text messages, emails.”
But many organizations lack such visibility, according to Dale. “If you don’t have visibility into the relationship between the buyer and the seller, it’s hard for you to build an accurate sales forecast.”
SO WHAT’S AI TO DO?
AI based on the right data helps reduce much of the time needed to make accurate sales forecasts, says Brett Weigl, Genesys’s senior vice president and general manager of digital and AI. “AI-powered virtual assistants can help sales reps by automating repetitive and error-prone tasks that come with engaging leads. They can also help qualify leads, answer questions and follow-ups as needed. Once a lead is qualified and ready to talk, reps take over. AI assistants free up sales reps for better focus and give them the insight they need to do what they do best—closing deals.”
AI-powered engagement tools equip businesses with predictive technology that helps convert customer interest into sales, Weigl adds. “This enables agents to intervene at the right moment with an offer, resources, or help to encourage customers on their journey in an insightful way. Over time, machine learning continues to adjust and improve the engagement model as the technology engages with more prospects and customers.”
AI can also quickly identify where different sales are in the funnel—whether they are on track, off track, and why. Human sales managers, particularly those with sizable sales teams, can’t make such assessments anywhere near as fast, Dale says. “Machines are uniquely qualified to be very efficient in doing that. So often when you see AI getting used in forecasting, it tends to be marketed as kind of a dark art. The reality is that when you lift the hood, AI is taking human intuition, testing it with data, and then applying it consistently to a long list of deals.”
Though AI can aid with the speed and accuracy of sales forecasting, companies can’t simply invest in the technology to improve sales forecasting, Wettre also warns. “You just don’t invest in the technology by itself and think it’s going to solve your problems without some level of understanding. That’s a trap you can fall into: deploying AI and thinking things are going to get better. You need to have an understanding of how that technology works. If you understand the basic principles of AI, then it you’ll be set up for success.”
AI helps with correlating the different factors involved in the sales methodology, such as product, geography, and size of the deal, according to Wettre. “There are dozens and dozens of factors that all have a correlation to your sales process. AI simply calculates the correlations at all times and compares them against [the methodology] in a way a human never could. Even a new person straight out of college assigned to the sales territory can benefit from that model of correlation. That takes decades for an experienced salesperson to learn intuitively.”
However, there are still misconceptions about what AI can and cannot do, according to Dale. “We are still not quite out of the ‘AI is going to do magic’ phase. A lot of [AI providers] promise a lot of different things, but it’s not clear what they can actually deliver.”
NEW OFFERINGS
In September, Salesloft introduced Salesloft Rhythm, a patent-pending signal-to-action engine designed to help sellers more accurately hit their sales forecasts. The AI-powered solution, which will be available sometime next year, will take all signals from the Salesloft platform and provide APIs to take signals from partners like Drift, Highspot, or Ironclad. These signals will be ranked using an AI model to produce a prioritized list of actions, according to Dale. For example, if a buyer responds, calling that buyer goes to the top of the queue because that’s the most important thing right now. Triggers from data and AI models are other kinds of signals that can suggest actions based on previous outcomes. Each action item Rhythm prioritizes comes with an explanation as to why the action is important and what the likely outcome will be.
That is an important component in any AI for sales forecasting implementation, according to Dale, who says AI has to offer value for companies and the salespeople who will use it. “Some stuff out on the market demos really well but then doesn’t solve the problem that people actually have,” he says.
A major problem for many in sales is repetitive work, namely the tracking and generation of prospecting emails, according to Dale. When AI does it, “that output is very valuable, because it helps you understand what is working and what is not,” he states.
Additionally, by letting AI do the work, salespeople can focus on closing deals, Dale says further. Even the best sales leaders can track only so many deals at a given time, so without AI to free up their time, they will miss deals that were winnable but slightly off track.
Sales leaders also need to be certain that AI is explainable: “Whenever you’re working with AI, you need to be able to explain to [human salespeople] why you are automating a task or the AI is making a certain recommendation,” Dale advises. “At the end of the day, sales is a human profession. We need humans to say whether or not a recommendation makes sense.”
And it helps to keep the focus on sentiment, something that Ensco Connect, a hospitality-focused automation and CRM platform provider, has found to be most effective to determine which prospects have a higher chance of conversion, according to Sasha Barak, its marketing manager.
“Our most used AI feature is sentiment recognition in customer conversations,” Barak says. “This is mostly relevant for selling to past guests or increasing the inquiry-to-booking conversion. Identifying and prioritizing communication to happy and extremely happy guests allows increasing re-marketing and sales results for hotels and vacation rentals.”
That kind of insight is particularly helpful today for the travel and hospitality industry, which was hit especially hard by the global COVID-19 pandemic, Barak notes.
Five years ago AI wasn’t broadly used in the sector, but the challenges of recent years have given space for innovation and expanded use. Now, companies in the sector can use AI to identify and prioritize clients with highly positive sentiment for the sales team, and prioritize clients with negative sentiment for the customer support team, Barak says.
THE BIGGEST MISTAKES WITH AI
But for all that AI can do, it’s important to know its limits, experts agree. A big mistake that some companies make when using AI to predict prospect conversions and forecast sales is to trust without verifying, according to Wettre. Bad data will also impact other departments in the organization.
Dale argues that companies need to question the validity of, and their visibility into, the data they’re using to make a sales forecast. They also need to review how accurate previous sales forecasts have been and adjust future assumptions based on the accuracy of past predictions.
“The biggest mistake I see businesses make is that they haven’t considered their data operations and data hygiene before implementing AI,” agrees Natalie Furness, founder and CEO of RevOps. “With poor data entering the AI system, they are only able to gain insight from poor data quality, which leads to poor data results.”
AI for sales forecasting risks being labeled as ineffective due to poor data hygiene more so than because of problems with the AI models themselves, Furness says. “In the future, we will see AI models which will be able to be trained to identify data hygiene issues, but from what I have seen so far, AI is currently very much based on the data input from the company, without any consideration for CRM data health.”
That is only one reason that the human element is still essential, according to Wettre. “You can’t just sit back and let AI do the work; humans still need to do their part to get the value out of AI.”
Similarly, humans need to be on the watch for unintentional bias, which can lead to missed sales opportunities and, in industries like financial services, potential problems with regulators. AI might, for example, overlook certain customer profiles, even though they could be interested and financially able to afford a product or service, due in large part to insufficient data about certain market segments, Wettre points out. There can be bias in the models, or bias in the marketplace. Humans need to work with the technology to eliminate any such biases, he states.
And then, too, “critical thinking is important,” Wettre says. “AI is now increasingly being included in college curriculums, business education, and technical education. College students are coming into the workforce with a basic understanding of AI that is real promising for the future.”
And Wettre expects to see sales leaders who are more knowledgeable on how to best deploy AI for the largest impact, and predicts that the process for building sales models will be greatly simplified. “When we first started with AI, models were built by hand and by experimentation,” he explains. “Over the last few years, models have been self-building, making them faster to launch, with a wider breadth of applications.”
And that is likely to continue, making artificial intelligence smarter and smarter as time goes by.
Phillip Britt is a freelance writer based in the Chicago area. He can be reached at spenterprises1@comcast.net.