How do you know a demand forecast is telling you the truth, and not just a good story?

It's a fair question — and one that most energy research houses, including EA at points in the past, have struggled to answer convincingly. Here's how EA's demand team closed that gap.

The challenge

Demand is a black box — and forecasts can drift without anyone noticing

Unlike supply or refining, demand is difficult to measure directly across the industry. That makes it easy for a demand forecast to quietly become narrative-led rather than data-led — shaped more by the prevailing story about where the market is headed than by continuously tested evidence.

It's a trap EA's own demand team is candid about having fallen into in the past. The fix wasn't to trust the forecast less — it was to give it something agnostic to be tested against.

The approach

An algorithmic nowcast, used as a continuous sense-check

EA's demand nowcast is a genuinely agnostic, algorithmic system. It draws on data points across the board and dynamically switches between underlying models to find the best fit as new information arrives. It does not know what EA's published forecast says, and it is not trying to match it.



• Forecast, historical forecast snapshot, and nowcast trend are plotted together — by product, region, and country.
• Every week, the team reviews the largest deltas between forecast and nowcast, and publishes them transparently across the business.
• A gap is treated as a question to investigate, never an automatic verdict on who is "right".
• In development: showing the nowcast's own evolution as a range, to reveal how stable or volatile it has been in a given period.
  

The result

In the early stages of the US-Iran conflict, the review surfaced a 2.7 million barrels/day gap between EA's forecast and the nowcast — roughly half of it traceable to a specific divergence in the China forecast alone. That's not a story about EA being wrong. It's a story about EA having a system that would catch it if it were — and being transparent about it internally, every single week, whether the gap is large or small. After a review process, we lowered our Chinese demand forecast amid a weaker refinery runs profile.

Why it matters

This is what "not just commentary" actually looks like

The market sometimes assumes energy research is built on analysts talking to contacts and writing up what they hear. This is the opposite of that: a systematic, quantified, company-wide discipline that tests EA's own assumptions every week against an agnostic, algorithmic check — and publishes the result internally regardless of whether it's flattering.

Want to see how this shows up in your balance?

Get in touch to talk through EA's demand nowcasting and how it feeds into published research.