Prompting Guide

While we trust Crunched will perform well “out of the box”, in this guide we’ll cover prompting tips to maximize the quality of Crunched outputs.

​

5 Key Principles

---

When interacting with Crunched, think of it as a brilliant but very new employee (with amnesia) who needs explicit instructions. The more precisely you explain what you want, the better Crunched’s output will be.

​

1. Be Explicit

Being explicit and providing more details helps Crunched understand your ask better and generates a more customized response with fewer errors.

​

What’s the Context of the Model?

Just like you might be able to better perform on a task if you knew more context, Crunched will perform better if it has more contextual information. Examples:

• The end goal of the task, or what a successful task completion looks like (i.e., market model with TAM, SAM, SOM built bottom up based on input XYZ)
• What workflow the task is a part of, and where this task belongs in that workflow (i.e., Working on a DCF model to estimate the valuation of XYZ company – need help stress testing assumptions)

​

Which Cell Ranges are Relevant?

Specify which sheet / cell ranges that are relevant for Crunched to:

1. Insert things in
2. Fetch from

Examples:

• “Error-check the output sheet” (it will still check for cross-sheet references)
• “Extend on the cash flow analysis in row 50-60, incorporating interest rate and debt payback from the ‘Debt schedule’ sheet to get to levered FCF”

​

Which Steps Should Crunched Take?

Use numbered lists or bullet points to better ensure that Crunched carries out the task the exact way you want it to.

​

How Should Crunched Do Your Task?

Prioritize telling Crunched about model structure. I.e., how should set up the model or the structure of the edits it should make. It can often be quicker to tweak assumption input cells than spelling out the assumptions in the chat. Simultaneously, Crunched would benefit from you as a power-user and professional modeler laying the foundation of how the task should be done. You can always ask Crunched to research and update assumptions in the next turn.

​

2. Make Verification Fast and Easy

Making verification easy and fast is the core skill that differ power users that are efficient and can actively use Crunched’s output despite occasional mistakes, vs. those that find it more tedious to verify long builds. Key is to make reviewing easy – think about it like sitting down with an analyst and walking through the model.

​

Ask Crunched to Make a New Column with Relevant Information

Example 1:

When linking from a raw data sheet with 100s of line items to a more consolidated input sheet, say: “…for each cell/line item in the ‘raw data’ sheet that you link to the output sheet, note which line item in the Input sheet you’re matching the raw-data financials with in a column to the right of the current raw data content”

Example 2:

Ask Crunched to take you through the analysis once you’re done: “Go through the analysis step-by-step. Do not do any edits, only walk me through and verify the calculations at the same time”

​

3. Split Multi-Step Tasks into Several Prompts

When working with complex tasks, Crunched can sometimes drop the ball if you try to handle everything in a single prompt.

​

Exception: Research Tasks

Crunched benefits from doing research and will provide more sophisticated and relevant outputs if you ask it to summarize its research in the chat before builds. Example:

“You are building a market model for Spanish online beauty products:

1. research the online beauty market in Spain and identify typical revenue drivers for an online beauty retailer
2. summarize your findings in the chat and decide on the bottom-up sizing methodology
3. create a market sizing for 2025-35 using your methodology”

​

4. Use Examples

Examples are your secret weapon for getting Crunched to work exactly as you need – it increases accuracy and consistency. Accurate examples = better performance, especially for complex tasks. Example 1: Trial Balance Categorization

When categorizing accounts in a trial balance, give a few examples of the right granularity for the desired categorization: “map all trial balances, use the following example mappings: …”

Example 2: Formula Pattern

When adding formulas in several distinct places: “Average the numbers in every sixth cell in the empty space of range B2:B100, e.g., B2 = AVERAGE(B3:B7), B8= AVERAGE(B9:13) etc.”

​

5. Use Crunched’s Built-In Features

When writing a prompt, you can leverage Crunched to quickly make them more detailed:

​

Prompt Optimizer

When writing a prompt, Crunched’s “improve prompt” suggestion will appear in the chat.

• Press the button or just press Tab on your keyboard
• Crunched will make a more detailed prompt based on your current sentence

​

Ctrl+L: Quick Cell Range Insert

Quickly insert the cell range you’d like to fetch data from / error-check / get Crunched to edit in: