Final Thoughts
December 11, 2025
Aluminum Supply-Demand Forecasts: Accuracy versus Precision
Written by Greg Wittbecker
As a 40- year veteran of the aluminum sector, I have done my share of supply and demand forecasts.
In my role as Vice President of Industry Analysis at Alcoa, I delivered quarterly updates that painstakingly compared our internal top-down/bottom-up estimates with third-party analysts such as CRU.
Our track record was directionally correct most of the time. However, we rarely nailed “the” number.
That naturally led to some “back-trading” by leadership asking, “What did we miss?” In a potential career-ending moment, I said “I am not sure any of this matters.”
That elicited the expected response: “What are you talking about? It’s your job to forecast these balances accurately, and we need to inform the market accordingly.”
At that point, I was reminded of one of the few principles in statistical analysis that stuck with me from my college days. An acceptable margin of error (ME) in much of statistical work typically falls between 1% and 5%.
Applying ME to the Primary Aluminum Market
I was given some sage advice many years ago by Joe Viland, who at the time was the President of Wabash Alloys, the predecessor of today’s Real Alloy.
Joe loved my passion for crunching the numbers on supply versus demand, but cautioned: “Don’t confuse accuracy and precision. You can be accurate, but you will die trying to be precise.”
When we begin to apply the ME to the primary aluminum market, we are dealing with some LARGE numbers and asking groups like CRU not only to be accurate, but to be laser-precise in their forecasts:
- World Production: ~ 74,000,000 metric tons (rounded)
- World Consumption: ~ 74,060,000 metric tons (rounded)
- World Net Balance: ~ (67,000) metric tons = deficit
Source: CRU Aluminum Market Outlook June 2025
This very well-crafted and thoroughly scrubbed deficit estimate by CRU represents the equivalent of 0.00091% of world production.
Statisticians would argue that a 1% ME would allow for an estimate of 740,000 tons to be acceptable. CRU, and their competitors, do give themselves that kind of slack. They are attempting to defy statistical probability and be precise.
It is for this reason that they are inevitably back-traded, just like I was at Alcoa, for getting the numbers “wrong”. It is also why these forecasts must be dynamic and updated quarterly, if not monthly.
There are simply too many moving parts in the market to achieve “precision”.
Recent supply-side examples applied to the June Balance
- Century’s Iceland smelter lost its electrical transformers in October 2025, resulting in the loss of 65% of the plant’s output for up to one year. That could represent up to 208,000 metric tons of annualized production loss. That would triple the size of their June 2025 deficit.
- South32’s MOZAL smelter in Mozambique is threatened with total closure at the end of the first quarter of 2026 due to the inability to secure a new power contract. This smelter has a nameplate capacity of 580,000 tons per year but is now expected to produce only 214,000 tons from July 1, 2025 through June 30, 2026. This would mean a reduction of production of roughly 235,000 tons for the period April-December 2026. That’s about twice the current deficit projection – and if that loss is added to the Iceland disruption, global supply would fall by 443,000 tons.
These two supply-side excursions would’ve made the June 2025 (67,000) deficit obsolete. CRU could not possibly have foreseen the Iceland problem. The South32 problems at MOZAL have been discussed before, and some caveat about this possibility should have been expected.
Why This Matters
Respected analysts like CRU have a thankless task in being asked to forecast supply versus demand.
The supply side in aluminum is “normally” less volatile than the demand side. This is the opposite of my early career in grain trading, where grain supply was the source of volatility and demand was relatively constant (we all got to eat, after all!)
Increasingly, you are seeing fewer listed companies making supply-demand forecasts themselves because of the volatility in the variables.
They fall back on CRU and others to buffer them from criticism of over- or under-shooting guidance.
From CRU’s perspective, more and more provisions are being made for variability in demand and the timing of production increases or possible excursions. Some analysts now publish bear-, base-, and bull-case scenarios, and believe this is the right approach.
So the next time you are tempted to rant about why the supply-demand estimate is “off,” keep in mind that statisticians would say these estimates are attempting the impossible.
Stick to accuracy, not precision. It’s better for your health.
