Too many companies pour millions into predictive analytics projects only to watch them stall, underdeliver, or disappear altogether. Kyle Johnstone has seen it happen again and again in industries from finance to manufacturing. It’s why he takes a different path. For him, AI isn’t about chasing the latest tech fad; it’s about solving real business problems that move the needle. After leading more than 50 successful AI initiatives, he has learned exactly what turns a big idea into measurable results and what sends it off the rails.
What Predictive Analytics Actually Means
Most people see predictive analytics as a shiny new invention. Kyle Johnstone knows better. “AI has been around since the 70s,” he says. “We had tools like Lisp, Smalltalk, and a handful of others. We used to call them expert systems.” At the time, only elite organizations such as the military and Wall Street firms could afford the massive computing power required. A turning point came in 1996 when IBM’s Deep Blue defeated chess champion Garry Kasparov. That victory relied on the same basic algorithms companies still use today, just with far less processing power. “The math has not changed. The algorithms are the same. The statistical models are identical,” Johnstone explains. “The difference is that now we have the processing power to run them faster, at larger scale, and with far greater accuracy.”
Identifying Problems Before Solutions
Johnstone’s workshops often start with a deceptively simple exercise that exposes why so many analytics projects fail. He asks executives to write down what was bothering them about work when they woke up that morning. “When you got up today, what issues were you thinking about as they relate to work? What’s on your mind? Let’s talk about that,” he tells them.
This approach cuts through the technology hype and goes straight to real business challenges. It might be declining sales, underperforming marketing campaigns, or bottlenecks in the supply chain. “First, we’re going to create the premise, then we’re going to do the testing, and finally we’re going to prove it out,” Johnstone says. His advice is to tackle one problem at a time instead of trying to solve everything at once.
Ensuring Data Accuracy First
Here is where many companies stumble. They have the latest software and the biggest datasets, yet their results still fall flat. “The algorithms do not lie. It is math. But if they do not have the right data to work with, the results will be skewed and inaccurate,” Johnstone says. His teams always start with baseline tests using historical data before applying models to new scenarios. They may also bring in external sources such as weather patterns or census data to improve accuracy. But there is one rule that never changes. “I have no leeway with algorithms and mathematics. Those numbers have to be accurate.”
While industry research shows that 85 percent of AI projects fail, Johnstone’s team has maintained a perfect track record. The reason is not cutting-edge technology or magical algorithms. “Sometimes, in the first two or three weeks, we say this is not going to work. Do not invest in it. But here are some other options you might consider,” he explains. Many challenges do not actually require AI at all. “It is amazing what comes out of some of these projects,” he says. Often the real issue is employee training, broken processes, or sloppy data entry habits. His team focuses on finding the right solution rather than forcing AI into every situation.
Forget the Technology, Focus on Results
Despite all the marketing hype around different platforms and tools, Johnstone keeps things simple. “Do not worry about tools. The technology is there. It has been there for decades. The systems work. The algorithms are proven, the math works,” he says. Whether a company uses Python, Databricks, or another platform depends on cost and the skills the team already has. The real challenge lies elsewhere. “I get wrapped around the business problem. That is where I want to focus,” he explains. “If I have this unclear concept and I do not have it nailed down, you are going to get an unclear answer and it is never going to be successful.”
Setting up predictive analytics is only the beginning. Market conditions change, customer behavior shifts, and economic factors evolve. “Once I deploy it, I still have to monitor it because we have something called Data Shift,” Johnstone says. A model trained on last year’s data might give terrible predictions for this year’s market. His advice for anyone starting out is to approach the process with patience and flexibility. “You really have to go into this with an open mind and not be so stuck on AI or trying to force fit an AI solution.” For Johnstone, success comes from solving business problems, not from chasing the latest technology trends.
Follow Kyle Johnstone on LinkedIn for proven strategies to make AI projects succeed.
