A commercial real estate loan portfolio outgrows Excel at the point where the number of loans, lenders and conventions exceeds what one person can keep consistent by hand: usually between ten and twenty loans, always by fifty, and immediately if the portfolio includes agency floaters, hedges and lender-specific covenant tests. The spreadsheet does not fail because Excel is bad at math. It fails because the portfolio’s work is repetitive, and repetition by hand produces drift, error and dependence on the one person who built the model. The failure modes are specific, a replacement has a defined job, and LoanBoss is a teammate for Excel rather than a competitor.
Where the spreadsheet fails
Drift
Balances updated from servicer statements, sometimes. Rates updated at reset, usually. Amortization on floaters recomputed monthly, rarely. Within a year, the balance on the summary tab is not the balance the servicer has, and every downstream number inherits the difference. See floating-rate re-amortization and SOFR tracking.
Conventions approximated
The spreadsheet has one yield maintenance formula. The portfolio has six conventions with different reference rates, lookbacks and floors. The DSCR tab has one definition of NOI. The lenders have twelve. Approximation is invisible until the number is challenged. See real-time prepayment calculations and DSCR and debt yield tests.
Dates without owners
Extension notice windows, cap replacement deadlines, repair completion dates and step-downs are in the documents and possibly in a column. Nobody is alerted. See loan critical date tracking.
Provisions in prose
Cash management triggers, lender consent thresholds, release formulas and burndown milestones do not fit in cells, so they live in a memo or in memory. See the 400-field loan abstract.
Reports rebuilt
The SREO for each lender, the board debt summary and the compliance certificates are assembled from the model each quarter by hand. Every change is made in four places. See automating the SREO and debt summary.
Key-person risk
The model’s author is the only person who knows why cell G47 subtracts 3% of revenue. When they leave, the model becomes an artifact. A customer’s capital markets lead said that replicating the accuracy LoanBoss delivers would cost multiple FTEs and still carry key-person risk.
Error rate
Research on spreadsheet risk consistently finds material errors in a majority of operational spreadsheets. A loan model with forty tabs is not exempt.
A worked example: the model at year three
An owner built a loan model in 2023 with eight loans. Forty tabs, one per loan plus summary, maturity, covenant and hedge tabs. It was excellent.
By 2026 the portfolio is 31 loans. The model’s history:
- Loans added by copying a tab. Eleven of the 23 new tabs were copied from a fixed-rate agency tab and adapted for bank loans, bridge loans and one CMBS loan. The yield maintenance formula on the agency tab is on all eleven, including the CMBS loan that defeases and the bridge loans with step-downs.
- Floaters. Six floater tabs use a fixed amortization schedule with a manual rate cell. Balances differ from servicer statements by $1,000 to $40,000 each.
- Covenants. The covenant tab computes DSCR one way. Twelve lenders. The analyst keeps a separate “adjustments” workbook that is updated at quarter end from memory.
- Hedges. Cap values are typed in when a broker sends a mark, roughly every six months. Three caps have replacement deadlines in the next year; one is on the maturity tab, two are not.
- Amendments. Four loans were amended. Two amendments are reflected; the other two are in a folder.
- Reports. The board debt summary is a fifth workbook linked to the model. The links break when a tab is renamed, which happens when a loan is refinanced.
- The author. Promoted in 2025. The current analyst inherited the model with a two-page note.
Illustrative example:
| Area | The model in 2026 | Failure mode |
|---|---|---|
| Loan tabs | 31 loans; 11 of the 23 new tabs copied from a fixed-rate agency tab, including a CMBS loan that defeases and bridge loans with step-downs | Conventions approximated |
| Floaters | Six tabs on a fixed amortization schedule with a manual rate cell; balances off the servicer by $1,000 to $40,000 each | Drift |
| Covenants | One DSCR definition for twelve lenders; adjustments in a separate workbook updated from memory | Conventions approximated |
| Hedges | Cap values typed in from broker marks about every six months; two of three replacement deadlines on no tab | Dates without owners |
| Amendments | Four loans amended; two reflected, two in a folder | Provisions in prose |
| Reports | Board summary in a fifth linked workbook; links break when a tab is renamed | Reports rebuilt |
| Author | Promoted in 2025; the model came with a two-page note | Key-person risk |
The model is not wrong in any obvious way. It is wrong in twenty small ways that will surface on the day a lender or a buyer asks a precise question. Nobody in the firm can say which of the twenty. That is the failure mode, and it is structural.
What a replacement has to do
- Hold every provision as data, abstracted from the documents once and updated on amendment.
- Compute under each loan’s actual conventions: index, daycount, amortization type, prepayment, hedge, covenant adjustments.
- Take financials from the accounting system automatically. See Yardi, MRI and RealPage integrations.
- Own the dates with alerts and recipients.
- Produce the deliverables in the formats the team already uses.
- Give the data back to Excel for the analysis that belongs there.
Teammate, not competitor
The industry runs on Excel and will continue to. Deal underwriting, fund models, waterfalls and ad hoc analysis belong in a spreadsheet. What does not belong there is the repetitive maintenance of loan data and the recurring production of the same reports. LoanBoss focuses on automating those functions; the platform’s outputs export to Excel, and the reports the team already uses in Excel are rebuilt in the platform so they refresh automatically. A customer’s finance lead said it can be tough to quantify the benefits until you start using it, and once you are, you will not want to live without it.
Signs it is time
- A quarter-end close that includes two days of debt summary assembly.
- A prepayment number that was wrong by more than rounding.
- A covenant breach or cash sweep the lender noticed first.
- A missed extension notice or cap replacement.
- A portfolio that grew and a team that did not.
- One person who cannot take a vacation in the first two weeks of a quarter.
Frequently Asked Questions
Is this an argument against Excel?
No. It is an argument against maintaining loan data by hand. Excel remains the analysis tool.
What about a shared cloud spreadsheet with better controls?
It addresses versioning, not conventions, dates or provisions. The failure modes above are about what the spreadsheet contains, not where it lives.
How do we know the platform’s numbers are right?
Run one reporting cycle in parallel and reconcile line by line. See spreadsheets to platform without disruption.
What does the analyst do afterwards?
Analysis. Scenarios, refinancing strategy, hedge decisions. The role improves; it does not disappear. See cost and ROI versus hiring analysts.
At what portfolio size does Excel stop working for loan tracking?
Usually between ten and twenty loans, always by fifty, and immediately if the portfolio includes agency floaters, hedges and lender-specific covenant tests. The limit is the number of conventions one person can keep consistent by hand, not the arithmetic.
Key takeaways
- The loan spreadsheet fails on drift, approximated conventions, unowned dates, provisions stored as prose, hand-rebuilt reports and key-person dependence, not on arithmetic.
- The failure is invisible until a lender, a buyer or a servicer asks a precise question.
- Portfolios outgrow Excel between ten and twenty loans, and immediately if they include floaters, hedges or lender-specific tests.
- A replacement holds every provision as data, computes under each loan’s actual conventions, reads financials from accounting, owns the dates, produces the deliverables and gives the data back to Excel.
- Excel stays for underwriting, fund models and analysis. What leaves is the maintenance.
- The analyst’s role improves: review and analysis instead of assembly.
Related reading
- Spreadsheets to platform without disruption
- Single source of truth for CRE debt
- Lender reporting automation that keeps your Excel reports
- Best CRE debt management software 2026
- Loan abstract in the glossary
Automate things that should have been automated a long time ago. Keep Excel for the things it is good at.
Sources
- SoftwareAdvice, CRE firm survey on manual debt tracking (accessed September 2026)
- European Spreadsheet Risks Interest Group, spreadsheet error research
- Customer statements published on loanboss.com
- LoanBoss product philosophy, loanboss.com