Spreadsheets stop working when the workflow they hold becomes too shared, too high-stakes, or too repetitive for a single file to carry safely — and there are seven signals that tell you the moment has arrived: two people editing different versions of the same truth, a model only one person can run, hours of manual re-keying between systems, formula errors that touch money or legal terms, slowness at volume, numbers nobody trusts, and a reporting cycle that turns into a fire drill. But outgrowing a spreadsheet does not mean you commission custom software. It means you pick one of three exits — buy a fit-for-purpose tool, wrap a thin AI workflow around the file you already have, or build custom — and the honest answer for most small commercial real estate firms is one of the first two. The mistake that costs the most is treating every spreadsheet failure as a build trigger. Most are not. This piece names the signals, then tells you which exit each one points to.
What “the spreadsheet stopped working” really means
A spreadsheet has not stopped working just because it annoys you. It has stopped working when it has become a liability — when the file now creates risk faster than it saves you work.
That distinction matters because it separates a problem you should tolerate from one you should pay to fix. A slightly clunky model that a fluent analyst runs cleanly once a week is fine. A rent roll that two property managers edit at the same time, producing two versions that quietly disagree about who owes what, is not fine. The first is friction. The second is a live error waiting to reach a tenant or an investor.
The reason firms get this wrong is that spreadsheets fail slowly and then all at once. Excel scales with your firm for years, absorbing more tabs and more macros, until the day a broken formula misstates a debt-service number in an offering, or the one person who understood the underwriting model leaves. The signals below are how you catch that curve before it catches you — and the goal is not to replace the spreadsheet reflexively, but to read what it is telling you and choose the cheapest exit that removes the risk.
The seven signals a CRE spreadsheet has failed
A spreadsheet is sending you a real signal, not just a complaint, when one of these seven patterns shows up in daily work. Each one describes a specific failure, and each points toward a different fix.
1. Two people, two versions of the truth. The moment more than one person needs to edit the same file at the same time, spreadsheets start to diverge. A rent roll emailed around as attachments, a deal pipeline kept in a shared drive that two brokers overwrite, a CAM worksheet with three dated copies in the folder — these are version-control failures, and they mean the file can no longer serve as your single record. This is the most common signal at a growing firm, and it is usually the least expensive to solve.
2. One person is the only one who can run it. When a model has grown so intricate that a single analyst is the only one who can operate or audit it, the spreadsheet has become key-person risk. Their vacation stalls underwriting; their departure takes institutional knowledge out the door. A file that only one head can hold is fragile no matter how elegant it is, and elegance is often what makes it unreadable to everyone else.
3. You re-key the same data by hand between systems. If someone on your team spends hours each week copying figures out of a PDF lease, a Yardi export, or a bank statement and typing them into a spreadsheet, the file is failing at its real job. Manual re-keying is slow, and every keystroke is a chance to transpose a number. This signal is about wasted labor and injected error at the same time.
4. Errors are reaching things that matter. A typo in a comps tab is embarrassing. A formula error that understates operating expenses in an offering memo, or a dragged-down cell that corrupts a debt-service coverage calculation, can cost a deal or invite a dispute. When spreadsheet mistakes start touching money, legal terms, or investor-facing numbers, the file has crossed from tool to liability. Industry studies of operational spreadsheets have long found that a meaningful share of complex sheets contain at least one material error; in a CRE context, “material” can mean a mispriced acquisition.
5. It has gotten slow and fragile at volume. A model built for ten properties groans at eighty. Sheets that take a minute to recalculate, break when someone inserts a row, or crash on open are telling you the data has outgrown the container. Volume is a legitimate signal — a workflow that ran fine at your old deal count can genuinely stop working at your new one.
6. Nobody fully trusts the numbers anymore. When your team has learned to double-check the pipeline total by hand, or caveats the rent roll with “assuming it’s up to date,” the spreadsheet has lost the one thing it exists to provide: a number you can act on without re-verifying. Lost trust is a quiet signal, but it is a decisive one, because an untrusted record is not a record.
7. Reporting has become a fire drill. If assembling the monthly investor report, the quarterly board deck, or the CAM reconciliation means someone blocks out two or three days to stitch tabs together by hand, the reporting layer has failed. The data exists, but getting a clean view out of it costs more than the view is worth. Recurring, high-effort reporting is one of the strongest signals that the workflow deserves a purpose-built tool.
None of these seven, on its own, dictates that you commission custom software. Each dictates that you do something — and the right something depends on which exit fits the signal.
Three exits, not one
When a spreadsheet fails, you have three ways out, ordered here from cheapest to most expensive. The discipline is to take the least costly exit that fully removes the signal, not the most impressive one.
Exit 1 — Buy a fit-for-purpose tool. For most of the seven signals, a product already exists that was built to solve exactly this, and it will cost a subscription rather than a project. Deal pipelines that two brokers overwrite belong in Dealpath or Apto. Rent rolls and CAM that diverge across copies belong in Yardi, AppFolio, or Buildium. Lease terms you re-key by hand belong in Prophia or Leasecake. Listing and marketing sprawl belongs in Buildout. Project-cost tracking belongs in Northspyre. When a category tool fits your process, buying beats building at almost any firm size — verify current features against each vendor’s own documentation, because proptech capabilities change every quarter. The full economics of when off-the-shelf proptech is genuinely enough are worked through in our buy-vs-build playbook for small CRE firms.
Exit 2 — Wrap a thin AI workflow around the file. Some signals are not about the container at all; they are about a slow, manual task attached to it. Re-keying lease terms, summarizing a stack of documents into a sheet, drafting the narrative around the numbers — these can be handled by a saved, tested prompt run in ChatGPT, Claude, or Microsoft Copilot, with a person checking the output, feeding a spreadsheet you keep. This exit costs little more than a subscription you likely already hold and requires no migration. It is the right first move surprisingly often, and it is the subject of our look at where duct-tape automation ends and a real build begins.
Exit 3 — Commission a custom build. Only when the workflow is central to how your firm competes, runs at real volume, and no off-the-shelf tool fits the way you actually work does a custom automation earn its cost — roughly $25,000 to $150,000 depending on scope, plus ongoing maintenance a firm with no IT department has to arrange. A build is the right exit least often, and it is the one every build shop will tell you to take. Whether your firm is even ready for one is a separate question, worked through in our build-readiness framework.
Matching each signal to the right exit
Each of the seven signals has a most-likely exit. The table below is a starting read, not a verdict — the honest choice depends on whether a product actually fits your process, which is why mapping the workflow comes before picking any tool.
| Signal | Most-likely exit | Why |
|---|---|---|
| Two people, two versions | Buy a shared-database tool | Multi-user records are a solved product category |
| One person can run it | Build or buy, plus documentation | Key-person risk needs a system others can operate |
| Manual re-keying | Thin AI workflow first | Extraction and entry are what AI tools do well now |
| Errors touching money/legal | Buy a validated tool | Purpose-built systems enforce checks a sheet cannot |
| Slow and fragile at volume | Buy a tool built for scale | The data has outgrown the container, not the process |
| Numbers nobody trusts | Buy, with a clean data migration | Trust returns with a controlled single source |
| Reporting fire drill | Thin AI workflow, then buy | Automate the assembly before replacing the whole stack |
Two rules keep this honest. First, a build appears in only one row, and even there a bought tool often wins — the custom exit is the exception, not the default. Second, before you act on any row, map the workflow end to end, because a signal often hides its real cause. A reporting fire drill can look like a tooling problem when the actual fault is that data enters three systems by hand upstream. Our workflow-mapping playbook walks through finding the true bottleneck before you spend on any exit.
The trap: rebuilding your spreadsheet as expensive software
The most costly mistake a small firm makes at this juncture is to hand a developer its broken spreadsheet and say “build me this, but as real software.” That does not fix the workflow. It encases every flaw in the process — the undocumented assumptions, the manual steps, the exceptions handled in one person’s head — inside a system that now costs six figures to change.
A spreadsheet that failed because the underlying process is unstable will fail again as custom software, only slower and more expensively, because the build inherited the instability. Porting a mess produces an expensive mess. The reason so many small-firm software projects stall is a scoping failure exactly like this one, examined in our piece on why most small-firm software projects fail and how scoping fixes it.
The way out of the trap is sequence. Get the process stable and documented first. Then take the cheapest exit — usually a bought tool or a thin AI workflow — and let it run against your real, messy data. Only if that thinner solution proves the workflow matters and still leaves a gap does a custom build earn its place, and by then you know exactly what it must do. Building your firm’s low overhead into an advantage rather than a liability is the whole argument of the small-firm CRE playbook.
A worked example: the rent roll that broke
Consider a hypothetical twelve-person property-management firm whose master rent roll lives in one shared workbook. Two managers update it, an owner pulls monthly reports from it, and lease terms are typed in by hand off PDF abstracts. Three signals are firing at once: two people editing one truth, hours of manual re-keying, and a monthly reporting fire drill.
The reflex is to commission a custom rent-roll system. It is the wrong first move. Signal one — divergent versions — is a solved product category, so the primary exit is buying a property-management platform such as AppFolio, Yardi, or Buildium that gives every manager one live record instead of dueling copies. That single change resolves the version problem and most of the trust problem with it.
The re-keying signal is separate, and a build is still not the answer. A thin AI workflow — a tested prompt that reads each lease PDF and drafts the structured fields a person confirms before they land in the system of record — handles the extraction for the cost of a subscription. And the reporting fire drill largely dissolves once the data lives in a platform that reports natively. No six-figure build appears anywhere in the honest version of this story; a firm that jumped straight to custom software would have paid the most to solve problems the market had already solved for a subscription.
Frequently asked questions
When do spreadsheets stop working for a real estate firm?
Spreadsheets stop working when a file becomes a liability rather than a convenience — when more than one person must edit it at once, when only one person can run or audit it, when errors start touching money or legal terms, or when getting a clean report out of it takes days. Annoyance is not the threshold; risk is. A clunky model a fluent analyst runs cleanly once a week is fine. A shared rent roll that two people overwrite into disagreement is not. The signal to act is when the spreadsheet stops giving you a number you can trust and act on without re-checking it by hand.
Does outgrowing a spreadsheet mean I need custom software?
No, and assuming it does is the expensive mistake. Outgrowing a spreadsheet means you pick one of three exits: buy a fit-for-purpose proptech tool, wrap a thin AI workflow around the file you already have, or commission a custom build. For most small CRE firms the answer is one of the first two, because the common failures — multi-user records, lease extraction, pipeline tracking — are already solved by products that cost a subscription. A custom build earns its roughly $25,000 to $150,000 only when the workflow is central to how you compete and nothing off the shelf fits it.
What are the signs a spreadsheet has become a business risk?
Watch for seven signals: two people editing different versions of the same file, a model only one person can operate, hours of manual re-keying between systems, formula errors reaching money or legal terms, slowness and fragility as your volume grows, a team that no longer trusts the numbers, and a reporting cycle that has become a multi-day fire drill. Any one of these means the spreadsheet has crossed from tool to liability. The presence of the signal tells you to act; which exit you take depends on whether a product fits your process or a thin workflow can bridge the gap.
Should a CRE firm buy proptech or build custom when spreadsheets fail?
Buy first, almost always. If a category tool — Dealpath or Apto for pipeline, Yardi or AppFolio for property management, Prophia or Leasecake for lease data, Buildout for marketing — already does the job at a per-seat price you can carry, buying beats building at any firm size. Build custom only when no product fits the process that differentiates your firm, the workflow runs at real volume, and you can staff the maintenance a build quietly transfers. Verify each vendor’s current features against its own documentation before you commit, since proptech capabilities change quarterly.
Can AI fix my spreadsheet problem without replacing the spreadsheet?
Often, yes. Many spreadsheet pains are about a slow manual task attached to the file, not the file itself — re-keying lease terms, summarizing documents into rows, drafting the narrative around the numbers. A saved, tested prompt run in ChatGPT, Claude, or Microsoft Copilot, with a person checking the output before it lands, handles those for little more than a subscription and no migration. This thin-workflow exit is the right first move surprisingly often, and it teaches you exactly what a future tool or build would need to do if the pain persists.
How do I know if it’s version control or a deeper process problem?
Map the workflow end to end before you decide. A signal often hides its real cause: a reporting fire drill can look like a tooling problem when the true fault is that data enters three systems by hand upstream. If the pain disappears the moment everyone works off one shared record, it was a version-control problem a database solves. If the pain persists even with a clean single source, the process itself is unstable — and no tool will fix an unstable process. Mapping first stops you from buying software that automates the wrong step.
Why do custom software projects fail at small real estate firms?
Most fail because a broken spreadsheet was rebuilt as software without fixing the process first. The build inherits the undocumented assumptions, the manual exceptions, and the instability that made the spreadsheet fail, then costs six figures to change. Projects also stall when the input data was messy, when no one at the firm could maintain the result, or when the workflow kept shifting so the system encoded a moving target. The fix is sequence: stabilize and document the process, take the cheapest exit that works, and only build what a proven thinner solution still leaves undone.
What does it cost to move off spreadsheets?
It depends on the exit. A fit-for-purpose proptech subscription runs from tens to a few hundred dollars per user per month, verifiable on each vendor’s pricing page. A thin AI workflow costs little more than an AI subscription you likely already hold, plus a few days of setup. A custom automation runs roughly $25,000 to $150,000 depending on scope, and that figure is a down payment because it carries ongoing maintenance. The cheapest exit that removes the risk is almost always the right one.
How many signals should be firing before I act?
Act on the first signal that has turned the spreadsheet into a genuine risk rather than an annoyance — you do not need all seven. A single formula error reaching an offering memo, or two managers overwriting a rent roll into disagreement, is enough to move that workflow off the file. Multiple signals firing at once usually point to buying a platform that resolves several at a time; fewer signals point to a narrower fix, often a thin AI workflow.
Where to start
The first question is not which platform to license or which developer to call. It is which of your workflows has actually crossed from friction into risk — and which of the three exits removes that risk for the least cost. Most firms that ask “is it time to build?” discover the honest answer is a subscription they can start this month or a thin AI workflow they can stand up in an afternoon, with a custom build reserved for the one workflow that truly earns it.
A free AI-readiness assessment produces that read: a short working session that looks at where your spreadsheets are failing, whether a product already fits, and where a thin AI workflow beats both buying and building — before you spend a dollar. It is built to talk you out of an expensive build you do not need as readily as toward one you do. Book a free AI-readiness assessment and get an outside score on your workflows before you commit to any tool.
Arthur Wandzel
