Winning Over Sales with a Predictive AI Solution


Aside from counting closed deals, there’s a lot about selling that is difficult to quantify or measure. High performing sales people are accomplished relationship builders and empathetic listeners. Additionally, they are experimenters – regularly creating and testing theories based on the bits of information they’ve squirreled away in their brains while researching prospects or during interactions with potential customers. Undeniably, there’s both art and science, and a lot of moving parts.

For this reason, there has been an explosion of technology offerings developed with the intention of making sales people more effective. The oft-cited Marketing Technology Landscape Super-Infographic (2018) – almost comical in its size – features literally hundreds of offerings targeted at sales specifically. With sales having to surmount ever-climbing quarterly goals, technology solutions are often brought in by management to help augment or amplify a salesperson’s efforts – promising more efficiency and effectiveness – claims that often make sales professionals a bit wary.

Salespeople have a right to be skeptical, as many of these software solutions don’t bring the value they promise or, importantly, help make their lives easier. The thing is, sales professionals have often developed their own methodologies for reaching their targets, and now they’re being asked to use something completely new and untested to achieve even higher goals. Now factor in the timeline that they’re being judged on – just three short, unforgiving months in a quarter – and put yourself in their shoes. Think of it like trying to convince Serena Williams to switch to a new racket right before Wimbledon and see how that goes over. Even if it might work better, now is not the time to introduce something new, when the stakes are so high. And for salespeople, let’s face it – the stakes are always high. Unless you’re selling beer at Oktoberfest, you’ve got to hustle to hit your quota.

And sales is a confidence game. To be good, you need to be confident in yourself, your strategy, your tools, and what you’re selling. Effective sales people, who have well-formed and thought-out methodologies are confident in their logic about who they should be targeting and when. Intuition and experience have served them well. Whether conscious or not, a skilled salesperson has numerous rules-based models in their head for what makes a good prospect.

This being said, there are faults in these mental models, no matter how successful the salesperson. There just isn’t enough capacity in the human brain to process the reams of data required to put the entire puzzle together. To paraphrase Donald Rumsfeld’s famous quote, there are known knowns, known unknowns, and unknown unknowns.

In the era of big data, we can home in on those known unknowns and unknown unknowns. Here we delve into AI and, more specifically, machine learning, a technology that enables us to mine vast pools of data for patterns that reveal what comprises an ideal customer. Prospects can then be ranked according to a predictive model.

After creating a predictive model using MintigoAI, sales occasionally views the results with skepticism. Sometimes the results aren’t congruent with a salesperson’s beliefs and experience. Perhaps they see a low ranked prospect who is very similar to a company they sold to before. This is where it’s important to think probabilistically. Think of it like the weatherman predicting a 70% chance of rain tomorrow. That doesn’t guarantee rain, but it’s a high-enough number for most of us to pack our umbrellas and consider precipitation inevitable. What it really means is that on seven out of ten days with conditions like they are, it would rain. So, pardon the pun, but with A-ranked prospects you’re a lot more likely to make it rain than C-ranked ones. So, if you had one-thousand prospects in your database, but you only had time to reach out to one-hundred of them, you’d have a lot more success focusing on A-ranked accounts.

But remember, timelines matter, especially with matters of probability. Results stabilize over time. Even if you flip a coin ten times and get heads every time, your chance of getting tails on the next flip are still 50%. Along those lines, when implementing a predictive model remember it can take time to show it’s working.

For sales, this often isn’t convincing enough. Back to the timeline issue, they live their work lives in quarters – and you have to be sympathetic to this. They’re being asked to think long term, but being judged on the short term. So how do you prove value? You apply the predictive model backwards. Look over the last year’s closed-won accounts and see how they sort-out in the predictive model. If the model is using good data and working properly, you should see a disproportionate number of A and B ranked accounts in the closed-won pile. During that exercise you can see potential future performance in past results.

While this should help mitigate most of the skepticism, this is the time for management to show that they’re bought in as well. To take data-driven sales to heart an organization really needs to commit to a minimum of six-months to see that the predictive models are working as expected. Now is not the time to jack up quarterly targets even higher, but rather align with sales in this new approach, even potentially adjusting metrics like numbers of meetings and opportunities. Those numbers may actually fall, but despite the lower volume, they should be of overall higher quality. Ultimately, this will mean more wins for sales, and more growth for the organization.

Doug Ramsay


Doug is a marketing manager at Mintigo, where he works on digital advertising, website optimization, design, and content creation. In his free time he loves traveling and taking short weekend road trips in Northern California, seeking out beautiful vistas, tasty vinos, and opportunities to go stand-up paddle-boarding.