The “Gut Feeling” Trap: A Success Story in Data Driven Forecasting

Many years ago, at the company I was working for, I was the Account Manager for the most important client they had.

Back then, data was not seen as critical as it is today. To be honest, even today, many people still make decisions based on “gut feeling.” Discussions were constantly revolving around external factors—things that you cannot really control.

The Core Problem

Going back to the main point: Because the forecast was not accurate, the Buyer was reluctant to place a big order. They were used to ordering only for the next week to stay safe.

When I took over, I saw right away that this was the core problem. But I could not simply complain to my client about why they behaved like that. If I were them, I would do the same. When you are in this type of situation, you need to understand why the other person behaves the way they do.

In this case, it was pretty obvious. The forecast was not accurate. It was not based on data but on gut feelings. The end result? Many sales were lost due to a lack of available units. At the same time, ordering only for the next 4 weeks created a massive problem for the Operations Department. They needed much bigger orders to plan their production lines. But that was not possible.

Are you seeing all the problems that the lack of data brings?

Why Excel Was Not Enough

I created a plan. I needed to provide a forecast in such a strong way that the other person felt willing to place an order for the entire quarter.

For that, I spent a significant amount of time analyzing the data. The data tells you many things; you just need to invest the time to dive in.

However, as there was a lot of data to analyze, it was not possible to use Excel. It would simply crash. When you have too much data, it is smart to use Python to dig in and do data discovery. It allowed me to handle millions of data points instantly and see the “truth” that Excel was hiding.

I use the word “invest” because, in my experience, taking the time to build a robust Python analysis has really paid off in all my projects.

The Solution: A Dynamic 50-Variable Model

I created a forecast for each individual SKU. It considered 50 distinct variables. And everything was very dynamic. This was done on purpose.

Let’s say my data indicated that sales for next year would increase by 5% based on specific trends, but my client said, “No way, it will increase by only 3%.” No problem. I simply changed the input from 5% to 3% right there in the meeting. This way, my client felt that he was building this forecast with me. It was a strategic way to remove all his constraints, one by one.

The Climax

When I went to Headquarters to make the forecast presentation, the Director and their full team were there. They expected the discussion to be as it always was: based on gut feeling, experience, and things that were not data.

When I presented the file with all 50 metrics and the logic behind why they needed to buy X amount of dollars, they could not believe it. They said that, finally, they had real data in front of them.

I presented only one SKU to show exactly how I calculated the forecast for that particular item. I wanted them to see that I had dedicated considerable time to building this. I could be wrong, but that forecast was super strong and considered every variable.

Once I presented that one SKU, I challenged them: “Tell me one random SKU, and I will show you my forecast.”

This time, they could clearly see how I arrived at the number. The trust was built instantly.

The Result

To make a long story short, it was the first time in a very long time that this client placed an order for the full quarter.

No one could believe it. At the same time, this had many beautiful implications for the Operations Department. Finally, they could get a big chunk of units to start producing.

We started a new circle where every metric got better and better. And the client was very happy. All this was done thanks to making good use of the data and making decisions based on it.