Stop asking for the price forecast. Ask for the price probability.

After years of building forecasting models, we have seen the same pattern again and again. A precise price forecast lands in a budget, the market moves the other way, and trust in forecasting takes a hit. The problem is rarely the model. It is the format of the answer.

Below are the honest pros and cons of both approaches, and when to use each.

The point forecast: one number, one answer

A point forecast says something like: “Copper will trade at $9,800 per tonne in three months.” It is the format most organisations know.

Pros

• Easy to communicate. One number fits in an email, a slide or a budget line.
• Plugs straight into planning. Finance, procurement and sales can all use the same figure.
• Simple to evaluate. Accuracy metrics like MAPE or RMSE are well understood.

Cons

• False precision. A single number hides how uncertain the future really is.
• Almost always wrong. In volatile markets like metals, the exact price rarely lands. Each miss erodes trust.
• No view of risk. It says nothing about the chance of a sharp spike or a crash.
• Hard to act on. If the forecast is off by 4%, should you have bought or waited? The number does not say.

The probability forecast: how likely is a move?

A probability forecast says: “There is a 70% chance copper rises more than 5% in the next three months.” It describes the odds, not a single outcome.

Pros

• Matches real decisions. Buyers decide when to buy, hedge or wait. A probability crossing a threshold is a clear trigger.
• Honest about uncertainty. It shows both the likely move and the risk around it.
• Built for volatility. It stays useful even when prices swing hard.
• Works as an alert. “Probability of a 5% rise just passed 75%” is a signal you can act on today.
• Measurable over time. Calibration can be checked: do 70% calls come true about 70% of the time?

Cons

• Harder to read. Many people hear “70%” as “it will happen”.
• Needs good data. Reliable probabilities require long history and careful calibration.
• No single number for the budget. Planners still need a price to put in the spreadsheet.
• Trust builds slowly. A single outcome cannot prove a probability right or wrong. It takes a track record.

Side by side

	Point forecast	Probability of % change
Answers	What will the price be?	How likely is a move, and how big?
Best for	Budgets and planning	Buy, hedge and timing decisions
Shows uncertainty	No	Yes
Easy to explain	Very	Takes some education
Handles volatility	Poorly	Well
How it is judged	Error vs. actual price	Calibration over many calls

Which one should you use?

Use both, but for different jobs.

• Planning and budgeting: a point forecast, ideally the median of a probability range.
• Purchasing and hedging: probabilities. Set a rule, such as “lock in volume when the chance of a 5% rise passes 70%”.
• Risk reporting: a range, for example “80% chance the price lands between $9,200 and $10,400”.

The strongest setup combines them: a probability band with a central price inside it. Planners get their number. Buyers get their signal. Everyone sees the risk.

The takeaway

No one can tell you the exact price of steel, copper or aluminium three months from now. Anyone who claims to is selling certainty that does not exist.

What a good model can tell you is the odds. And for a buyer, knowing when the odds have shifted is worth far more than a precise number that will probably be wrong. Next time you review a forecast, ask for the probability, not just the price.

Related

• Why nobody can forecast the exact steel price, and what to track instead
• Order timing for metal buyers
• Copper price forecast and prediction for purchasing teams