Arijit Sengupta is the CEO and founder of BeyondCore (now a part of Salesforce.com), which offers tools that empower business users with enhanced analytics and augmented intelligence, allowing them to benefit from data-driven insights.
I used to believe that when knowledgeable people said something was impossible, it really was impossible. I learned to have a healthy dose of cynicism and not take things at face value. When people are extremely confident about what they’re talking about, when they use phrases like “Everybody knows this,” or “This is how our business actually works,” there’s a very high probability they are wrong. The most unquestioned assertions are the sources of the biggest opportunities. I learned the importance of letting the data speak.
One of the occasions that helped me learn this lesson was three years ago when we were building BeyondCore’s Einstein Data Discovery, which is now part of Salesforce. The product automates data analysis; you just point the software at a data set, and it comes back and tells you what’s going on with it.
We were talking to this customer who had a product-return problem. [The company] was shipping things to [its] customers, and the customers were saying they were not getting the products on time. For instance, they were supposed to get the product in 24 hours and instead got it in 72 hours.
So we went in and analyzed the data, and the pattern we found was a specific shipping hub in Texas; whenever the [company shipped] from that hub, there were major delays.
When we told our customer the story, it said, “You guys are completely wrong. That’s impossible. That hub only ships to other hubs, and those hubs ship to other hubs and those hubs ship to customers. This is our storage repository. And there’s no way of shipping to customers from there.”
We spent a month looking at our algorithms, trying to figure out if we did something wrong, if there was a mistake in the product. It almost shook our confidence a little bit; this was in the early days of the product, and we thought, Maybe automated systems don’t quite work.
Then we manually did all the work and found that the software actually was right.
We went back to our customer and said, “Look, you said you never shipped from this hub, but here are thousands of cases where you [did].”
What [had] happened was, in its enterprise resource planning, [the company] had made it impossible to ship something from that location. But people always break the rules. Somebody would call up a friend in the hub and say, “Could you do me a solid and ship this product out because my customer is yelling for it?”
And even when the ERP system wouldn’t allow it, the [hub employees] would find a way, and the product would be shipped from the hub.
The most unquestioned assertions are the sources of the biggest opportunities.
Now when we find a pattern that a customer says is impossible, our entire focus goes onto that pattern. People believe tribal knowledge, but the data does not lie.
Quite often, by focusing on one thing that everyone thought was impossible, you can create tremendous value for the customer. It’s flipping the equation from what most people do.
When a customer says, “I like these three things, and the fourth thing I absolutely hate,” most people will focus on the three things the customer liked. Instead, we focus on the thing they thought was wrong. That’s because the things they said they like are things that they already know and that align with their beliefs about their business.
You don’t create value for the customer when you relate back to them things they already knew about their company. You create real value when you change their tribal knowledge. Without the intelligence to back it up, knowledge is just bias.
Everyone talks about artificial intelligence, but I talk about augmented intelligence. You can use that to find out about problems, but then you need to translate it back into something people can work on. Machines can look at millions of possibilities and rank them to create a stack of things to [examine]. A human then looks at that stack and figures out what it means.
You need the AI because if you start with a human, you just get information biases. If you just use the AI, if never translates into domain-specific information that is factored into the data. AI takes the bias out of the equation.
I’ve learned that when [people say] something is impossible, rather than take them at their word, let the data speak. Someone saying something is impossible should be an invitation to dig further.