Versioning Your Models: Building an Audit Trail of Thinking

Investors update their valuations frequently. A company reports earnings, management revises guidance, a competitor emerges, or broader market conditions shift. You revise your assumptions and recalculate intrinsic value. A month later, your intrinsic value estimate is $10 higher than three months ago. Was that increase justified by improving business fundamentals? Or did you become overconfident? Most investors can't answer this question because they don't track how their assumptions and valuations have changed over time.

Implementing version control in your spreadsheet models solves this problem. Each time you update assumptions materially, you save a timestamped version capturing the state of your thinking at that moment. Over months and years, you build an audit trail showing the evolution of your analysis. This serves three critical purposes: it helps you calibrate whether your estimate changes are grounded in fundamentals or sentiment, it documents the reasoning behind assumption shifts, and it creates a feedback loop that improves your forecasting accuracy.

Versioning doesn't require complex software or advanced technical skills. Spreadsheet-based approaches range from simple (manually saving dated copies of your model) to sophisticated (using spreadsheet revision history features and change tracking). The key is discipline: committing to consistent versioning practices and reviewing them periodically to extract lessons.

Quick definition: Valuation model versioning is a practice of regularly saving timestamped copies of your spreadsheet (or using built-in change-tracking features) that capture assumptions and intrinsic value estimates at specific points in time, creating a historical record of how your analysis has evolved.

Key Takeaways

• Versioning creates accountability for your estimates by documenting when and why you changed them
• The simplest approach is saving dated copies (TICKER_Valuation_2025-12-15.xlsx); more advanced approaches use spreadsheet revision history
• A "changelog" tab within your spreadsheet documents each major assumption revision, the reason for the change, and the resulting impact on intrinsic value
• Comparing versions reveals whether estimate changes were driven by business fundamentals or mere sentiment drift
• Regular reviews of your version history (quarterly or semi-annually) help calibrate forecast accuracy and identify recurring bias patterns
• Spreadsheet version control differs from software version control; it doesn't require external tools for basic practice, though cloud platforms offer helpful features
• Documenting not just what changed, but why, is what transforms version history into genuine learning
• Over years of versioning, you build empirical data on your forecasting accuracy, which informs your conviction levels on future estimates

The Case for Versioning: Why Your Memory Isn't Enough

The human memory is excellent at many things but terrible at precise quantitative recall. You remember that you were bullish on a stock, but can you remember what you assumed about its growth rate six months ago? You recall that your intrinsic value estimate increased, but was it because fundamentals improved or because you gradually increased your growth assumptions without consciously deciding to?

Versioning forces precision. When you save a timestamped copy, you're creating a snapshot. When you review that snapshot months later, you see exactly what you assumed at that point. More importantly, when you compare snapshots over time, you see the trajectory of your thinking.

Consider this scenario: In March, you estimate a company's intrinsic value at $45 based on 15% annual growth and 25% operating margins. In June, after a strong earnings beat, you revise to $50. Did the company improve? Maybe, but it depends on what specifically changed. If revenue growth is actually accelerating to 18% and margins to 27% based on new evidence, the $5 increase is justified. If growth is still 15% but you simply increased your confidence that you should assume 15% (rather than your prior 12%), the increase is intellectual drift, not analysis improvement.

Versioning lets you distinguish between these scenarios. Review your March and June versions side-by-side, and the truth is visible.

Simple Versioning: Dated File Naming

The simplest versioning approach requires no special software. Each time you update your model significantly, save a new copy with a date in the filename:

AAPL_Valuation_2025-03-15.xlsx
AAPL_Valuation_2025-06-22.xlsx
AAPL_Valuation_2025-09-10.xlsx
AAPL_Valuation_2025-12-15.xlsx

Use a consistent naming convention: TICKER_Valuation_YYYY-MM-DD.xlsx. This ensures files sort chronologically when sorted alphabetically. The date format YYYY-MM-DD (ISO standard) is international and unambiguous.

Store all versions in a single folder per company. When you want to compare your analysis from six months ago to today, you can open two versions side-by-side.

The discipline is this: don't save new versions constantly. Save a new version quarterly (after earnings) or when major assumption changes occur (guidance shift, significant news, competitive developments). This keeps the number of versions manageable while capturing the important evolution points.

The Changelog Tab: Documenting Your Thinking

A more sophisticated approach adds a "changelog" or "version history" tab within each spreadsheet. Rather than comparing multiple files, you maintain a table within your current model documenting major revisions:

Date	Reason for Change	Growth Rate	Op. Margin	Discount Rate	Intrinsic Value	Notes
2025-03-15	Initial analysis	15%	25%	8.0%	$45	Post Q4 earnings
2025-06-22	Q2 beat, raised guidance	18%	26%	8.0%	$52	Revenue growth accelerating
2025-09-10	Macro headwinds	16%	24%	8.5%	$46	Raised WACC due to rate environment
2025-12-15	Q3 guidance, margin pressure	14%	23%	8.5%	$42	Growth deceleration becoming clear

Each row captures a major version. The columns show what changed and, most importantly, why. The "Notes" column is where your learning lives. Why did you raise the discount rate in September? Because the broader market environment shifted. Why did you lower growth in December? Because the company's guidance decelerated, signaling competitive or execution challenges.

This changelog accomplishes several things. First, it creates a searchable history. Six months from now, you can review why you revised assumptions and learn from them. Second, it documents your thinking contemporaneously, when it's fresh, rather than trying to remember later. Third, it creates accountability; when assumptions change, you're forced to articulate the reason.

Maintain the changelog in chronological order (oldest to newest, or newest to oldest—be consistent). Update it each time you make material assumption changes. Review it quarterly to understand your thinking trajectory.

Using Spreadsheet Revision History

Modern cloud-based spreadsheets offer built-in revision tracking that simplifies versioning without manual file management.

Google Sheets includes version history out of the box. In the menu, select File > Version History > See Version History. Google Sheets displays every change to the spreadsheet, along with timestamps and the user who made the change. You can click on any previous version to view the spreadsheet's state at that point in time. You can also restore the spreadsheet to a previous version if needed.

This is powerful because it's automatic. You don't have to remember to save a dated copy; Google Sheets does it. However, Google Sheets' revision history shows every small edit, including typos and minor tweaks. Your version history might contain 500 revisions in a month, with most being inconsequential changes. You need discipline to identify which versions matter—the ones corresponding to material assumption changes.

Excel Online (browser-based Excel) offers similar version history called "Previous Versions" or "Version History" (name varies by Office 365 update). Access it from the File menu. It captures versions less frequently than Google Sheets (typically daily snapshots rather than every keystroke), making it slightly less granular but easier to scan.

Excel Desktop (non-cloud version) doesn't have built-in versioning. You must use external tools (like OneDrive version history if you store files there, or Git for serious version control) or implement manual versioning with dated file copies.

For individual investors, Google Sheets' built-in version history is often the easiest approach. Combined with a changelog tab documenting which versions matter and why, it provides powerful tracking without overhead.

Advanced Versioning: Git for Serious Investors

Some sophisticated investors use Git (a software version control system) to track spreadsheets. This is overkill for most investors but powerful if you're managing multiple models or collaborating with others.

Git works by tracking changes to text-based files. Spreadsheets are complex binary files (or XML-based in newer versions), but you can export them as CSV files or track the underlying data separately in text format. Each time you commit changes to Git, you create a permanent snapshot of the model with an attached message explaining what changed and why.

Using Git requires comfort with command-line interfaces or a graphical client (GitHub Desktop, GitKraken). It's not beginner-friendly. For most investors, the simpler approaches (dated file copies, Google Sheets revision history, or a changelog tab) are sufficient. Git becomes valuable if you're maintaining models over years, collaborating with others, or want forensic-level tracking of every assumption change.

Quarterly Retrospectives: Learning from Your Versions

Saving versions is only valuable if you review them. Schedule quarterly retrospectives where you examine your valuation history and extract lessons.

Pull up your four most recent versions (roughly quarterly intervals). Compare the latest version to the version from one year ago:

• How much has your intrinsic value estimate changed?
• Which assumptions shifted, and what drove those shifts?
• Did the business actually improve, or did your forecast change for other reasons?
• Were your assumption changes anticipatory (you were ahead of developments) or reactive (you changed your view after the market did)?
• Are there recurring bias patterns (e.g., you consistently raise growth assumptions after positive news, even when growth hasn't actually changed)?

This retrospective isn't just record-keeping; it's a feedback loop that improves future estimates. If you realize you have a pattern of being too optimistic about margin expansion, you can calibrate more conservatively next time. If you notice you always underestimate disruption, you can build more margin of safety into terminal values.

Some investors maintain a separate "investment log" or "decision journal" documenting major decisions and their outcomes. Versioning complements this. Your decision journal says "bought XYZ at $40, thesis was margin expansion." Your versioning history shows whether you were right about margins.

Collaboration and Multi-User Versioning

If you share your valuation model with a spouse, financial advisor, or investment club, versioning becomes important for collaboration. Google Sheets excels here; multiple people can edit simultaneously, and version history preserves all changes.

Set up clear conventions: one person owns the base case, another reviews and suggests changes. All changes are captured in history. You can see who made which changes and when. If someone updates the growth rate assumption, the timestamp and user are visible. This prevents confusion about who changed what.

Some investors use comments within Google Sheets for deeper collaboration. A user can select a cell and add a comment suggesting a different assumption. The model owner sees the comment, considers it, and either accepts or rejects the suggestion. These comments become part of the version history, creating a record of discussion.

Archiving and Storage

Over years, you accumulate many versions. Organize them by company in separate folders. Within each company folder, keep all versions of that company's model. This creates a searchable archive.

For security, ensure versions are stored safely. Google Sheets versions are stored in Google's cloud and backed up automatically. Excel files stored on OneDrive similarly benefit from cloud backup. If you use local files on your computer, back them up regularly to an external drive or cloud storage.

Some investors delete old versions periodically to save storage space. Before deleting, export important historical versions as PDF snapshots for archival reference. Your oldest versions (from 3+ years ago) become less relevant, but they're valuable for long-term learning about your forecasting accuracy.

Real-World Example: Learning from Version History

Consider an investor who has tracked Apple's valuation over three years:

Year 1, Q1: Intrinsic value $125, growth assumption 12%, margin assumption 27%, discount rate 7.5%

Year 1, Q4: Intrinsic value $140, growth assumption 15%, margin assumption 28%, discount rate 7.5% (reasoning: iPhone sales strong, services revenue accelerating)

Year 2, Q1: Intrinsic value $138, growth assumption 14%, margin assumption 28%, discount rate 7.8% (reasoning: growth expectations reset after weaker guidance, raised WACC due to rate environment)

Year 2, Q4: Intrinsic value $155, growth assumption 16%, margin assumption 29%, discount rate 7.5% (reasoning: margin beat triggered optimism, new products showing strength)

Year 3, Q1: Intrinsic value $145, growth assumption 13%, margin assumption 27%, discount rate 8.0% (reasoning: macro headwinds, margin guidance disappointed, raised WACC)

Reviewing this history, the investor notices: When Apple beats, growth and margin assumptions increase. When it misses or guidance disappoints, they decrease. This is reactive analysis, not anticipatory. The investor realizes the valuation is being driven by recent earnings surprises rather than fundamental business changes.

This realization is valuable. Going forward, the investor can implement discipline: avoid adjusting growth assumptions immediately after earnings surprises. Let two quarters of data pass before revising long-term growth assumptions, which reduces reactivity.

Without version history, this pattern is invisible. With it, the pattern becomes obvious, and the investor can improve future estimates.

Common Mistakes in Versioning Practice

Mistake 1: Versioning Without Documentation

You save dated file copies, but you don't document why each version was saved. Years later, you have ten versions but can't remember what distinguished version 3 from version 4. Always document the reason: "Post earnings revision," "Management guidance shift," "Market conditions changed." The version history is only useful if it's interpretable.

Mistake 2: Versioning Too Frequently or Too Infrequently

Versioning every day captures noise. Versioning annually captures too little. Quarterly versioning (after earnings) or when major assumption shifts occur strikes the right balance for most investors. This keeps the number of versions manageable while capturing important evolution.

Mistake 3: Changing Multiple Assumptions Simultaneously

You update your model and change growth, margins, discount rate, and terminal value assumptions all at once. When you review the version history, the intrinsic value changed significantly, but which assumption drove it? Always change one or two assumptions at a time. Document each change separately. This isolates the impact of different assumption changes.

Mistake 4: Not Reviewing Versions

The most common mistake is saving versions but never reviewing them. This defeats the purpose. Commit to quarterly retrospectives. Pull up your version history quarterly, compare recent versions to versions from one year ago, and extract lessons. This transforms versioning from busywork into genuine learning.

Mistake 5: Letting Old Versions Become Outdated

Early versions of your model used inferior assumptions (you learned better methods). You're tempted to delete them because they're "wrong." Don't. Old versions capture your thinking at the time. They're not current estimates, but they're valuable historical records. Archive them but keep them.

FAQ

Q: How often should I create a new version?

A: After quarterly earnings reports or when major assumption changes occur (significant news, guidance shift, competitive developments). Don't version daily or weekly; this creates noise. Quarterly is a good baseline for most investors.

Q: Should I include multiple scenarios (bear, base, bull) in versions?

A: Yes, if you model scenarios. Version all three scenarios' intrinsic values. This lets you see whether your base case changed but your bull/bear cases remained similar, indicating confidence in the range, or whether entire scenarios shifted, indicating broader reconsidering of the business.

Q: How long should I keep old versions?

A: Keep all versions indefinitely if storage space permits. They become less relevant after 3+ years, but they're valuable for understanding your long-term forecast accuracy. Cloud storage is cheap; local storage is less so. Archive very old versions to external media if needed.

Q: Should I version competitor comparisons or just DCF models?

A: Version your primary valuation model (DCF, dividend discount model, etc.). If you also track relative valuations (peer multiples), include them in the changelog tab as context ("peer P/E average: 18x, our assumption: 16x"), but don't create separate versions for each. One primary version per company, with supporting metrics noted in the changelog.

Q: Can I use spreadsheet version history if I'm the only user?

A: Yes, absolutely. Even for solo analysis, Google Sheets' version history is valuable. You can click through old versions to see what you assumed at different times. Combined with a changelog tab documenting major revisions, you have robust tracking without technical overhead.

Q: What if I made a mistake in an old version?

A: Don't delete or correct it. Old versions should be immutable—they represent your thinking at that time, errors and all. If you notice an error, document it in your current version or changelog. "Year 2 version had error in WACC calculation (should have been 7.8%, was 7.5%); corrected in current version." This transparency is valuable.

Related Concepts

Decision Journals — A personal record of investment decisions and their reasoning, updated over time with outcomes. Versioning complements decision journals by providing quantitative tracking of how assumptions changed.

Feedback Loops — The broader practice of gathering data on outcomes and using it to improve future decisions. Quarterly version retrospectives create feedback loops that improve forecasting.

Calibration — The practice of testing whether your confidence levels match actual accuracy. If you assigned 70% confidence to something and it happens 70% of the time, you're calibrated. Versioning and retrospectives help build calibration data.

Investment Policy Statements — Some investors document a written investment policy stating their approach and decision rules. Versioning history helps validate whether you're actually following the policy or gradually drifting.

Summary

Valuation models evolve as companies change and new information emerges. Without systematic tracking, you lose the ability to learn from your own estimates. Versioning solves this by creating a timestamped record of how your thinking has evolved.

Implement one of three approaches: dated file copies (simplest), a changelog tab within your spreadsheet (moderate sophistication), or cloud-based revision history (easiest if using Google Sheets). The approach matters less than consistency. What matters is discipline—actually saving versions at meaningful intervals and reviewing them regularly to extract lessons.

Quarterly retrospectives are the key practice. Pull up your version history, compare estimates over time, and ask: Where was I right? Where was I wrong? What bias patterns do I exhibit? How can I improve future estimates? This transforms versioning from record-keeping into genuine learning that compounds over years of investing.

An investor who can trace the evolution of their thinking—seeing when and why they adjusted growth assumptions, when they raised discount rates, when confidence in estimates increased or decreased—is better equipped to understand their own forecasting accuracy and make better future estimates.

Next Steps

With version history documenting how your thinking has evolved, the next step is making your models accessible to others. Learn how to implement safe sharing practices that protect your analysis while enabling collaboration with trusted advisors or investment clubs.