Think Build Implement Repeat
London, UK +44 7367 067226
WhatsApp FOLLOW f in X
  1. Home
  2. Blog
  3. Hierarchical Forecasting: Making the Numbers Reconcile
AI & Machine Learning

Hierarchical Forecasting: Making the Numbers Reconcile

Forecast by branch and by region separately and they will disagree. How reconciliation works, which level to forecast at, and what to show each audience.

Updated 2 min readBy SpiderHunts Technologies

Free estimateNo obligation

Get a free estimate

Tell us what you need. A senior engineer reads every enquiry.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →

Quick answer — TL;DR

Forecasts produced independently at different levels will not sum correctly, which destroys trust before anyone looks at accuracy. Reconciliation forces consistency; forecasting at the level with the strongest signal and distributing down is usually the pragmatic choice.

The credibility problem

A business forecasts total revenue for the board, regional volumes for operations and branch-level numbers for the managers. Each is produced by whoever owns that view, and none of them add up. The first hour of every planning meeting then goes on arguing about which is right.

This is a structural problem rather than an accuracy one. Independent forecasts at different levels of the same hierarchy have no mathematical reason to agree, and they will not.

Bottom-up, top-down, or both

There are three broad approaches, and the right one depends on where your data is strongest rather than on organisational preference.

ApproachHow it worksSuits
Bottom-upForecast each branch, sum upwardBranches with distinct, stable patterns
Top-downForecast the total, split by shareNoisy branch data, stable shares
Middle-outForecast at region, split down and sum upRegional signal stronger than either end
ReconciledForecast every level, then adjust to fitWhere all levels carry real information

Bottom-up appeals to managers because it respects local knowledge, but noise at branch level accumulates. Aggregate series are almost always smoother and easier to forecast, which is why top-down often wins on accuracy at the total even when it feels less rigorous.

What reconciliation actually does

Reconciliation takes forecasts made at every level and adjusts them so the hierarchy adds up, weighting each level by how reliable it has proven. In effect the levels lend each other information - a noisy branch borrows strength from its region.

The practical benefit is not only arithmetic consistency. Because each level contributes, the reconciled numbers are often more accurate than any single approach, particularly for the smaller, noisier units at the bottom.

Choosing the level to plan at

Before reaching for reconciliation, ask whether every level needs a forecast at all. Plenty of hierarchies have one level where decisions are actually made, and the rest are reporting views.

  • If stock is held centrally and shipped on demand, the branch forecast may not drive any decision.
  • If each branch orders independently, branch-level accuracy matters directly and deserves the effort.
  • If staffing is rostered regionally, region is the level to optimise and branch numbers are informational.

Forecasting only where a decision depends on it is cheaper and produces less to argue about.

Presenting it so people trust it

Once the numbers reconcile, show the same set everywhere. The moment two versions of the same figure circulate, confidence goes and people revert to their own spreadsheets.

Include the accuracy history at each level, so a branch manager can see how reliable their own number has been. That does more for adoption than any amount of methodology explanation, and it surfaces genuine local knowledge where the model is consistently wrong. Our piece on why staff override predictions covers what to do with those overrides.

Numbers that do not add up get replaced by spreadsheets, whatever their accuracy.

FAQ

Frequently asked questions

The questions readers ask us after this guide.

Still have a question?

Ask us directly — a senior engineer will get back to you.

Ask about your project

Is bottom-up or top-down more accurate?

Top-down is often more accurate at the top because aggregate series are smoother, while bottom-up respects local differences. Reconciliation is a way of getting some of both.

Do we need every level forecast?

No. Forecast where a decision depends on the number and treat the other levels as reporting views derived from it.

How do we handle a branch that opened last year?

Treat it like a new product with no history - borrow from comparable branches and adjust, then switch to its own data once there is enough.

Can reconciliation fix a badly wrong forecast?

No. It makes levels consistent with each other, not correct. A poor forecast reconciled is still a poor forecast.

Keep reading

More on AI & Machine Learning

Start here

Want machine learning project details from us?

Tell us what you are trying to predict and roughly what data you hold. We will come back with an honest view on whether machine learning is the right tool, what the work would involve and a realistic cost range. If a spreadsheet would do the job, we will say so.

  1. You tell us what you needTwo minutes on the form, or a message on WhatsApp.
  2. A senior engineer reviews itAnd comes back with questions, a realistic range and an honest view on fit.
  3. Free 30-minute scoping callWe talk through scope, options and a realistic estimate — with no obligation.
Free estimateNo obligation

Talk to someone who builds this

Send a short brief and we will come back with an honest view and a realistic range.

Takes under a minute. We never share your details.

  • Free consultation
  • No commitment
  • NDA on request

Prefer to talk? Book a free 30-minute call →