I keep seeing the same forecasting problem in growing businesses, and it is rarely the spreadsheet formula everyone wants to blame.
The forecast breaks because the business has not agreed what the key numbers actually mean.
That sounds almost too basic to be worth writing about. But it is one of those boring finance problems that quietly leads to expensive decisions.
Where the forecast starts to drift
A sales leader says the pipeline is strong. Finance asks whether that means signed contracts, verbal commitments, weighted CRM opportunities, quotes sent, or work that has already started but has not yet been invoiced. Operations says capacity is tight, but their version of booked work includes jobs that might move by three weeks.
The founder hears “we are on track” and makes a hiring or stock decision. Then the month closes and everyone is surprised.
The model did not fail. The definitions underneath it were never clear enough to support the decision.
This is common in mid-market businesses where finance is being asked to become more commercial. The CFO or finance manager is expected to give better forward visibility, but the operational systems were often built for activity, not forecasting. The CRM tracks sales behaviour. The job system tracks delivery. The billing system tracks invoices and revenue rules. Each system may be reasonable on its own, while still producing numbers that do not line up cleanly.
One practical example: I have seen teams debate forecast accuracy for weeks before realising that “revenue forecast” meant three different things in the room. One person meant expected invoice value. Another meant work performed. A third meant signed scope, regardless of timing. Nobody was being careless. They were using the same word for different business events.
Once you see that, the fix becomes less glamorous and much more useful.
Start with the definitions page
Before rebuilding the forecast, define the business events that matter. Which items count as committed revenue? Which ones are still probable? When does a job move from pipeline to backlog? Which event changes the forecast: signature, purchase order, deposit, resource allocation, delivery start, delivery completion, invoice issue, or cash receipt?
You do not need a hundred definitions. You need the ten or fifteen that drive decisions.
- What are the forecast categories?
- Which source system owns each number?
- What event moves an item between categories?
- Who can override the number, and why?
- How often does finance review the assumptions?
- What happens when sales, operations and finance disagree?
This page can feel painfully simple. But it often improves forecast trust faster than another layer of formulas.
The aim is not to create a policy manual. It is to make the definitions clear enough that two capable people would classify the same transaction the same way. If they cannot, the dashboard will eventually become a negotiation instead of a decision tool.
Then automate the right thing
This is where AI and automation can help, but only after the business has done the thinking. An automation can pull CRM opportunities into a forecast. It can flag changes, compare actuals to forecast, and show where assumptions are drifting. AI can help summarise movements and spot patterns.
When the underlying events are vague, though, automation just makes the confusion arrive faster and look more polished.
That is the uncomfortable bit. Better finance systems do not remove the need for judgement. They make the judgement visible.
If I walked into a business with a forecast nobody trusted, I would not start by asking for the most complex model. I would ask for the definitions page. If there is no definitions page, that is usually the first deliverable.
Founders and CFOs do not need perfect visibility. They need a forecast they can test without everyone arguing about the language. Once the definitions are clear, the model can improve. Reporting can improve. Automation can improve. The conversations also become sharper because people are finally looking at the same version of reality.
If your forecast keeps missing, pause before you rebuild the file. Ask whether the business has agreed what the forecast is actually measuring. That conversation is not as exciting as a new dashboard, but it is usually where the real progress starts.
If this is the kind of finance systems problem you are trying to untangle, Think Numbers can help you get the definitions, workflows and reporting lined up before you automate the mess.