Plan next quarter with the ranges left in

Budget season produces a lot of single numbers. "This plan delivers £24m next half." The number goes into a deck, the deck goes to finance, and the number quietly becomes a target.

When the quarter lands at £22m, nobody remembers that the forecast was never that certain. It just looks like the plan failed.

What an honest forecast includes

• A range, not just a number. "Most likely £24m, probably between £22.5m and £25.5m" sets expectations correctly from the start.
• A range that widens further out. Next month is easier to forecast than next spring. A forecast whose range stays the same width for six months is hiding something.
• Carryover from past spend. Spend from the last few weeks is still working. Ignore it and the first weeks of the forecast are wrong before the plan starts.
• Stated assumptions. Price changes, promotions, competitor activity, the economy. A forecast is only as good as these, so write them down where everyone can see them.

Comparing plans honestly

Most planning compares the middle numbers: plan A forecasts £24.1m, plan B forecasts £23.3m, so A wins. But if both ranges are wide and overlap heavily, the data can't really tell them apart.

The better question is: what's the chance plan A beats plan B? That's calculated by simulating both plans many times from the same model and counting how often A comes out ahead. "A beats our current plan in about 96% of simulations" is a strong case. "In about 55%" is a coin toss dressed up as a decision.

Presenting it to finance

1. Lead with the decision, then the range: "We recommend plan A. It most likely adds about 3%, and very likely beats the current plan."
2. Show the assumptions that move the answer most, and what happens if they're wrong.
3. Say what you'll check. Agree to compare actual weeks with the forecast range as they arrive.

Then check it

Each week, see whether actuals land inside the forecast range. If they do most of the time, the forecasts deserve trust. If they keep missing on one side, something in the model or the assumptions needs fixing, and you've found out early rather than at the year-end review.