Research

As we all know, research is extremely complex. We often fall into our own bias traps through cherry-picking. Data sources aren't always trustworthy, and the information may be outdated or irrelevant.

Nevertheless, gathering comprehensive data is valuable—as long as you don't fall into a rabbit hole of endless research instead of solving the actual problem.

Our Methodology

The methodology we used to find the data was:

• We used ChatGPT, in Deep Research mode - using the GPT 4.5 model.
• We carefully reviewed every line of output and examined each referenced data source.
• We repeated this process a few times. When there were discrepancy in the ranges, we chose the conservative (lower) estimate.

Results & Validation

We were pleased with the results given that:

• The data spans approximately 15 years, with sources ranging from 2006 to 2025.
• The data across multiple sources aligned at a macro level. While specific numbers varied (as expected), the general figures were consistent across different sources.
• The data aligns with Juan's 40+ years of experience and common industry knowledge. For example, while experts might debate whether a refinery generates $2M or $5M per day for oil companies, no one would rationally argue that refineries lose $100 per day when idle.
• The data reveals an unmistakable trend: projects are consistently delayed. The evidence supporting this conclusion is overwhelming, regardless of how we might adjust the specific numbers.

If you're skeptical about the data's validity due to "AI hallucinations," we encourage you to review the evidence—you'll find every point is well-supported and justified.

If you have more accurate data, we would love to hear from you.

That being said, here is the data and references we went through: