When Steve Jobs launched the iPhone in 2007, many pundits were less than impressed. Some said that its unusual shape made it unwieldy. Others thought that it was too expensive. Still others remarked that all the extra software made it a poor choice for its primary function — making phone calls.

But part of Jobs’s genius was his ability to recognize patterns that others couldn’t. Executives at Xerox, for example, didn’t see much potential in the Alto, but he built the Macintosh based on it. When music players seemed like a dead end, he reimagined them with the iPod and transformed the industry.

The problem with patterns is that it’s so devilishly hard to tell the good ones from the bad. What may look like a promising pattern is often out of context or incomplete. Sometimes, we think we see a pattern that isn’t really there. That’s what makes innovation so difficult, we can never validate new patterns by looking backward, we can only test them going forward.

The Importance of Context

In 2001, Fabio Rosati left his lofty position as Global Chair of Strategic Consulting for Capgemini to join a troubled startup called Elance. The idea behind the company was a familiar pattern, the disintermediation of a labor intensive industry by technology. In this case, the aim was to automate the market for freelance workers.

Unfortunately, the pattern was a false one. While it seemed to work well in the context of job boards for full-time positions, for some reason it just didn’t work for freelance jobs. So Rosati imagined a completely new pattern. It seemed to him that large enterprises, many of whom dealt with thousands of vendors, could use technology to keep track of them all.

It turned out to be a very good idea and today, vendor management software is a big business. Elance was soon profitable, but before long Rosati saw another familiar pattern emerge. The success of the vendor management business was beginning to attract stiff competition from the likes of SAP and Oracle. He decided to sell Elance’s software business.

However, in his five years running the vendor management operation, he noticed other patterns. The problem with many freelance contracts isn’t finding people to do the work, but creating a successful engagement. With the remaining staff at Elance, he relaunched the freelance marketplace, but this time instead of focusing on making matches, the platform was designed to create for successful engagements.

The idea took off and Elance grew at an astounding pace. Later, it would merge with its rival, oDesk, to create Upwork which is today a massive enterprise, encompassing 12 million freelancers, 5 million clients and $1 billion in annual freelancer billings.

Familiar Patterns Gone Astray

As the Elance story shows, being able to recognize important patterns is key to innovating effectively. The problem is that just because we recognize a pattern doesn’t mean that it’s worth pursuing or even that it’s really there at all. In fact, there is an entire branch of mathematics dedicated to identifying when distinct patterns arise from random points.

Consider the case of Coke executives in the early 1980s. They had been humiliated by the Pepsi Challenge, a blind taste test that showed consumers preferred their competitor. So the company developed a new formula that focus groups said they loved. The result was New Coke, one of the greatest marketing disasters in history.

Not to be outdone, in 2010 Pepsi dropped its Super Bowl ad spots and invested $20 million in Pepsi Refresh, a social platform that awarded grants to good causes. Pepsi’s social KPI metrics soared, but in business terms it was an unmitigated disaster. A Harvard case study showed that sales dropped 5 percent and, for the first time in 20 years, the brand fell to third in its category.

Both companies made essential the same mistake. They recognized a pattern that was out of context. A blind taste test told Coke’s marketers nothing about the loyalty customers felt about its branded product and while many applauded Pepsi’s social efforts, they did little to spur sales for a brand built on sugar water and good times.

The Power Of Embedded Patterns

Jim Allison spent his career learning the patterns of the immune system and made some key contributions himself. Yet when he imagined a new pattern, he soon learned how difficult it is to break an old one that’s been deeply ingrained in an industry.

Allison’s idea was that the human immune system was capable of recognizing cancer cells, but because of the way immune system works our defenses are turned off too soon. It seemed to him that if he could just turn off the “molecular brakes” for a while, the body’s own defenses could kill off cancer cells in the body.

He performed some studies on mice and they were hugely successful. So he began flying around the country to sell his idea to pharmaceutical companies, but all of them refused. They had seen this pattern before, lost a ton of money, and were unwilling to go down the same road again. “It was depressing,” Allison told me. “I knew this discovery could make a difference, but nobody wanted to invest in it.”

After three years pounding the pavement, a small biotech company, Medarex, invested in Allison’s idea and today, cancer immunotherapy is recognized as a miracle cure and has saved the lives of thousands of terminally ill patients who once would have no hope. Medarex was sold to Bristol Myers Squibb in 2009 for $2.4 billion.

You Can Only Validate An Idea Going Forward

Think about Steve Jobs and Appl for a minute and you will probably recognize the pattern and assume I mispelled the name of his iconic company by forgetting to include the “e” at the end. But I could have just have easily been about to describe an “Applet” he designed for the iPhone or some connection between Jobs and Appleton WI, a small town outside Green Bay.

The point is that we can only validate patterns going forward, never backward. That, in essence, is what Steve Blank means when he says that business plans rarely survive first contact with customers and why his ideas about lean startups are changing the world. We need to be careful about the patterns we think we see. Some are meaningful. Others are not.

Another important point is that recognizing a valuable pattern is necessary, but not sufficient to create a business. Fabio Rosati’s ideas about vendor management were only the start of what made Elance successful. It was later moves, such as offering certification and training for freelancers, building private talent clouds for companies and countless others that made the company a powerhouse.

Finally, we need to take a more Bayesian approach to strategy, where we don’t expect to always get it right, but to become less wrong over time. A flash of insight will not make you Steve Jobs, it just means that you’ve recognized a pattern that may or may not really be there.

This article originally appeared at DigitalTonto

Greg Satell is a popular speaker and consultant. His first book, Mapping Innovation, is coming out in 2017. Follow his blog at Digital Tonto or on Twitter @DigitalTonto