How I Built a Bottoms-Up Budget Model in 40 Minutes Using Claude

Every marketing team has a version of this problem: the budget planning season arrives, someone needs a model by Friday, and you spend the next three days doing a job that is 20% thinking and 80% formatting, cross-referencing, and explaining the same logic in four different versions of the same spreadsheet.

I have done this cycle more times than I can count. You gather inputs from five stakeholders, reconcile conflicting assumptions, rebuild the model three times because the structure keeps changing, and eventually deliver something that looks polished but took far longer than the insight it contains actually required.

Last quarter I tried something different. I sat down with Claude, described a budget planning brief in plain language the same way I would explain it to a smart colleague, and iterated in conversation until I had a fully structured bottoms-up budget model — channel allocation, CAC targets, pipeline waterfall, and a narrative layer explaining every number. The whole session took 40 minutes.

This is the complete account of how that worked, what I actually said, what came back, and the reusable prompt structure anyone can take and apply to their own budget cycle.

First: What a Bottoms-Up Budget Model Actually Is

A bottoms-up budget model works from outcomes backward to spend — rather than starting with a budget number and dividing it up. You start with the pipeline target, work out what conversion rates are required at each stage, calculate the lead volume you need, and then back-calculate what each channel needs to spend to generate that volume at an acceptable cost.

Done properly, it produces a model where every budget line is defensible from first principles: you are not allocating 30% to paid search because that is what you did last year, you are allocating it because 30% generates X leads at Y CAC which contributes Z to the pipeline target. That is the model that wins budget conversations.

It is also the model that takes the longest to build manually — because you are simultaneously managing the maths across multiple channels, the assumptions for each one, and the narrative that explains why the numbers are what they are. Three separate cognitive tasks, usually happening in one spreadsheet at the same time.

What the Old Process Actually Looked Like

The 40-Minute Session — Step by Step

Here is exactly how the session ran. Every prompt is real. Every output structure is what came back. I have kept the detail high because the replicable part of this is not the concept — it is the specific way you structure the conversation.

What the Model Output Actually Looked Like

The Reusable Prompt Structure — Take This and Use It

What Claude Gets Right — and What You Still Own

This is not a “AI replaces the marketer” story. It is a “AI eliminates the work that was never the valuable part” story. There is an important distinction, and it matters for how you actually use this in practice.

Why This Matters Beyond the Budget Model

The budget model is one use case. The principle behind it — verbally brief Claude the way you would brief a smart colleague, then iterate until the output matches what you need — applies across almost every structured marketing deliverable: campaign briefs, channel audits, competitive analyses, board deck narratives, pipeline reviews.

The pattern is the same in every case. The thinking is yours. The structure, the calculations, the scenario modelling, the first-draft narrative — Claude handles that layer. What you get back is time to think about whether the strategy is right, rather than time spent making the spreadsheet look like a spreadsheet.

That is a meaningful shift. Not because it is faster — though it is dramatically faster — but because the cognitive bandwidth it frees up is exactly the bandwidth that produces better strategic decisions.