Forecasting is not always about predicting one number for one point in time.
In many businesses, forecasts are needed at different time levels. You may need a daily forecast for operations, a weekly forecast for planning, and a monthly or yearly forecast for longer-term decisions.
These forecasts are connected. And making sure they stay connected is where temporal hierarchies can help.
What are temporal hierarchies?
A temporal hierarchy is a way of looking at the same time series at different time frequencies.
For example, a business may have:
· Daily forecasts
· Weekly forecasts
· Monthly forecasts
· Quarterly forecasts
· Yearly forecasts
Each level gives a different view of the same data.
Daily data can show short-term changes. Monthly data can make longer-term trends easier to see. Both can be useful, depending on the decision you are trying to make.
The challenge is making sure these forecasts agree with each other.
Why does consistency matter?
Imagine that your yearly forecast is 120,000 units.
You then create monthly forecasts, but when you add them together, they give you 126,000 units.
Both forecasts may look reasonable on their own. But they cannot both be right.
This can create problems when different teams use different forecasts.
The operations team may work with daily numbers, while management looks at monthly or yearly numbers. If these numbers do not match, people may end up making decisions based on different views of the future.
Temporal hierarchies help solve this problem by keeping forecasts at different time levels consistent.
Different time levels show different patterns
One reason to use temporal hierarchies is that different patterns can be easier to see at different frequencies.
Daily data captures details that other levels miss, such as weekday patterns, holidays, and promotions. But it can also be noisy, which makes longer-term trends harder to see.
When the same data is aggregated to a weekly or monthly level, much of that noise averages out. Trends become easier to see, but the daily detail is lost.
This does not mean that monthly forecasts are always better than daily forecasts.
It means that each level can provide useful information.
Temporal hierarchies allow us to use that information together. The daily forecast can benefit from the trend that is clearer at the monthly level, and the monthly forecast can benefit from the detail in the daily data. In practice, combining levels this way often makes forecasts more accurate, not just more consistent.
Forecasting and reconciliation
The main technical idea behind temporal hierarchies is forecast reconciliation.
First, forecasts are produced separately ateach level. Then reconciliation adjusts them so that they add up correctly.
Going back to our earlier example: the yearly forecast says 120,000 units, while the monthly forecasts add up to 126,000. Reconciliation does not simply pick one and discard the other. It combines them, giving more weight to the levels that have proven more reliable. The result might be a yearly forecast of 123,000, with monthly forecasts that add up to exactly that.
The same principle applies across the hierarchy: days add up to weeks, and months add up to quarters and years. Weeks and months are a special case, since a week can start in one month and end inthe next. So both are built from the daily level instead.
The result is a coherent forecast:different levels of the hierarchy are aligned and describe the same future.
Why is this useful?
Temporal hierarchies are useful when a business needs forecasts fordifferent planning horizons.
An operations team may need daily forecasts. A supply chain team may need weekly ones. Management may only look at monthly or quarterly numbers.
Instead of treating each forecast as a separate problem, temporal hierarchies connect them. Each team gets the level of detail it needs, everyone works from the same view of the future, and each forecast benefits from whatthe other levels can see.
How does this fit into Indicio?
Indicio uses temporal reconciliation to help businesses forecast at different time levels while keeping them consistent.
The goal is simple: different views of the forecast should still tell the same story.
Rather than choosing between a detailed forecast and a higher-level forecast, businesses can use both.
This can make forecasting easier to use across teams and planning horizons.
The takeaway
Forecasting at a single time level gives you one view of the future.
Temporal hierarchies allow you to work with several levels, such as daily, weekly, monthly, and beyond, while keeping them connected.
For businesses that need forecastsfor both short-term operations and longer-term planning, this can turn several separate forecasts into one coherent, and often more accurate, view of the future.
