Testing Claude for Excel: Can It Build a Real Financial Model?

Claude for Excel: A Financial Modeling Challenge

Introduction to Claude for Excel

• Claude for Excel enables users to create comprehensive financial models using basic prompts, raising questions about its legitimacy compared to traditional methods.

• The video aims to compare the effectiveness of Claude in building a financial model against the creator's own expertise, structured into four key sections.

Getting Started with Financial Modeling

• The presenter mentions that templates used in the demonstration are sourced from their Excel add-in, Model Wiz, which simplifies creating financial models and offers beautiful dashboards.

• The first task involves uploading an existing income statement and balance sheet to see if Claude can generate a model incorporating historical data and projections.

Building the Three-Statement Model

• After uploading the file, the presenter instructs Claude to generate a financial model that includes actual figures from January 2024 through December 2025 and projects until December 2027.

• Claude confirms understanding of the request and begins creating necessary tabs, including a drivers tab for historical and projected values.

Analyzing Outputs and Projections

• As Claude processes the information, it displays code behind its actions; it uses various driver types like sales growth percentages for calculations.

• Initial outputs show hard-coded values in error checks; however, as projections are added, formulas begin populating across both income statements and balance sheets.

Error Detection and Adjustments

• Notably, Claude identifies errors in net income calculations and recognizes that net income resets every January—demonstrating advanced analytical capabilities.

• It corrects references within formulas automatically when detecting discrepancies between columns during projection updates.

Finalizing the Model

• As completion nears, Claude auto-fits columns and summarizes findings on the drivers tab while ensuring no months contain errors—showcasing its efficiency in finalizing complex financial models.

Analysis of Financial Model Adjustments

Overview of Balance Sheet Adjustments

• The speaker notes that the model has successfully frozen the balance sheet, indicating a thorough check on its accuracy and functionality.

• Retained earnings and net income are highlighted as key accounts dependent on the income statement; the model uses prior month's values to project these figures.

• The cash flow statement is not initially included; instead, cash is calculated as a plug based on asset and liability differences.

Cash Flow Statement Creation

• The speaker decides to create a dedicated cash flow statement, linking it directly to the balance sheet for accurate cash tracking.

• Complications arise in calculating net differences between assets and liabilities, emphasizing the need for careful data management in financial modeling.

• A circular reference error is detected during checks, with a $2 rounding difference noted as acceptable within an error threshold.

Insights from Model Outputs

• The model indicates significant cash burn due to operating losses projected to lead to negative cash by mid-2026, highlighting potential financial risks.

• Changes made in the income statement (e.g., adjusting growth rates) reflect accurately across linked statements, demonstrating effective integration within the model.

Investment Impact Analysis

• An example investment of $10,000 in computers results in an expected cash outflow reflected correctly in the cash flow statement.

Comparison of Modeling Approaches

• The speaker contrasts their method of unpivoting income statements and balance sheets into tables for easier reference against Claude's approach which uses hard-coded values.

• A drivers tab allows for dynamic updates based on selected averages or assumptions, showcasing flexibility in projections compared to static models.

Error Checks and Assumptions Consolidation

• Emphasis is placed on consolidating all assumptions onto a centralized drivers tab for efficiency; however, some projections remain isolated within specific tabs.

• The speaker expresses concern over rigid projection methods used by Claude’s model and suggests that more adaptable approaches should be considered.

What Happens When Balance Sheet Values Are Changed?

Understanding Errors in Financial Models

• The speaker questions the implications of changing a balance sheet value (e.g., assets, liabilities) without any visible errors, highlighting potential discrepancies.

• A hard-coded value in projections can mislead users into thinking historical balances are accurate when they may not be, emphasizing the importance of dynamic data.

• The lack of error notifications across different tabs can lead to missed mistakes; the speaker notes that errors are only indicated on one tab and could easily go unnoticed.

• The speaker describes their method for setting up error checks within their model, which consolidates various checks into one tab for better visibility.

• Trust in one's own financial model is emphasized, alongside appreciation for robust error checking mechanisms.

How Does Claude Forecast Revenue?

Challenges in Revenue Forecasting

• Revenue forecasting is complex due to varying business models; the speaker lacks a specific template and relies on inputs from decision-makers.

• The speaker provides details about their SaaS app's revenue model, including pricing ($300 per license), customer acquisition methods, and average customer lifespan (18 months).

• Claude demonstrates an understanding of SaaS revenue models by incorporating churn rates and customer acquisition channels without explicit instructions from the user.

• Claude successfully creates a revenue build tab that captures essential inputs like price per license and churn rate while linking back to the drivers tab.

• Color coding is used to indicate changes made by Claude; green signifies linked data from separate sections, showcasing common practices in financial modeling.

Error Detection During Revenue Build

Identifying Issues with Projections

• Claude performs final error checks during revenue build completion but encounters issues related to accounts receivable that were not initially communicated by the user.

• Circle references are flagged as problematic by both Claude and the speaker; this indicates a need for careful structuring of formulas to avoid calculation errors.

• An unexpected $10 discrepancy is identified as significant rather than trivial rounding; this highlights the importance of precision in financial modeling.

• Despite impressive capabilities, there are concerns about overestimations in projected customer acquisitions that would likely raise flags if presented externally.

Creating a Custom Revenue Model

Inputting Variables for Accurate Forecasting

• The speaker outlines their approach to inputting variables such as license revenue and sales quotas into their financial model for clarity and accuracy.

• A ramp-up period for new sales representatives is detailed, showing how gradual increases in productivity can be modeled effectively.

Headcount Forecasting and Financial Modeling Insights

Overview of Headcount Management Tools

• The speaker expresses satisfaction with the tool "Quad," noting its ability to generate useful outputs from limited instructions, indicating a strong adaptability for various business models.

• Emphasizes that headcount is typically the largest operating expense, making accurate forecasting crucial for financial management.

Headcount Forecasting Process

• The speaker uploads data regarding projected hires into Quad to assess its integration capabilities within their financial forecast.

• Quad analyzes the headcount file, linking salary details and total headcounts back to the income statement, showcasing its analytical process.

Evaluation of Quad's Performance

• While Quad successfully pulls some data, it only captures certain columns and relies on hardcoded values for gross salaries and projections.

• The speaker notes limitations in Quad’s functionality, such as lack of prorating for mid-month hires or terminations, which could lead to inaccuracies in forecasts.

Comparison with Manual Methods

• The speaker contrasts Quad's output with their own method that utilizes formulas for precise calculations of gross salaries and departmental expenses.

• Highlights the importance of including all cost types (e.g., payroll taxes, health benefits), which are essential for comprehensive forecasting but were overlooked by Quad.

Error Checking Mechanisms

• Discusses implementing error checks in their manual model to ensure accuracy; this includes verifying that total spend matches departmental breakdowns.

• Demonstrates how errors are flagged across tabs when discrepancies occur, emphasizing the need for robust error-checking systems in financial modeling.

Headcount Route: Who Won?

Overview of Headcount and Dashboard Creation

• The speaker reflects on the simplicity of the headcount route, expressing surprise at its straightforwardness despite its basic template.

• A call to action is made to create a dynamic KPI dashboard that summarizes financial model information for management, investors, or board members.

• Confidence is expressed in outperforming Claude due to extensive experience in building engaging dashboards linked with their tool, Model Wiz.

Dashboard Features and Comparisons

• The speaker showcases their KPI dashboard, highlighting features like period comparisons against prior periods or budgets.

• Custom formatting options are praised for enhancing readability (e.g., using K for thousands and M for millions), along with easy snapshot sharing capabilities.

Evaluation of Claude's Dashboard

• As Claude’s dashboard comes together, the speaker notes its ability to self-QA but critiques its visual appeal compared to their own design.

• While acknowledging Claude's functionality in pulling data, the speaker finds it visually unappealing and prefers their version for presentation purposes.

Final Thoughts on Financial Models

• The speaker expresses a biased preference for their financial model due to better schema and error checks while recognizing some strengths in Claude's revenue build.

• Concerns about maintaining an advantage over evolving technology like Claude are voiced; collaboration rather than replacement is emphasized as a strategy moving forward.

Future Considerations

• The speaker contemplates future developments of AI tools like Claude and expresses intent to closely monitor advancements while improving personal skills.