The proptech build vs buy question is usually asked backward. A CRE firm decides it wants a data platform, then debates building the whole thing versus buying the whole thing, as if it were one decision. It is not. A data platform is a stack of layers, and the right answer is almost always to build one of them and buy the rest. The layer worth building is the one that encodes how your firm underwrites, screens, and thinks, because that is your edge. The layers worth buying are the ones every firm needs and no firm competes on: storage, extraction plumbing, and infrastructure. Getting this line wrong is expensive in both directions.
Key Takeaways
Build vs buy is not one decision. Build the layer that encodes your firm's edge; buy the commodity layers every firm needs and no firm competes on.
The Standish Group's 2020 CHAOS report found 31 percent of software projects succeeded, 50 percent were challenged, and 19 percent failed. Building the wrong layer walks into those odds for no strategic gain.
Ongoing maintenance of custom software runs an industry-benchmarked 15 to 20 percent of the original build cost every year, before any new feature is added.
What a CRE firm should own is its data model and workflow logic. What it should almost never build is document extraction infrastructure, a data warehouse, or OCR.
The right question is not "build or buy the platform." It is "which layer is our edge," and build only that.
What Does "Build a Data Platform" Mean?
Building a data platform means constructing a stack of distinct layers: infrastructure and storage, a data warehouse, an extraction pipeline that turns documents into structured data, a canonical data model, and the workflow applications on top. Treating these as one buildable thing is the first mistake, because each layer has a different build-versus-buy answer.
The layers differ in one decisive way: how much of your competitive edge they carry. Storage and infrastructure carry none. Every firm needs a place to keep files, and no principal ever won a deal because of superior object storage. A data warehouse is a solved commodity with mature vendors. Document extraction is hard and general, which is exactly why it is a category others have spent years solving. But the canonical data model, how your firm defines a qualified deal, what belongs in your buy box, how you weight tenant credit, and the workflow that runs your screening, those encode judgment specific to your firm. That is the layer where building can create durable advantage, and the layer where buying a generic tool forces your process into someone else's mold.
Layer | Build or buy | Why |
Infrastructure and storage | Buy | Commodity, zero competitive edge |
Data warehouse | Buy | Mature vendors, solved problem |
Document extraction pipeline | Buy | Hard, general, a solved category |
Canonical data model | Build | Encodes your firm's definitions |
Workflow and screening logic | Build | Encodes your firm's judgment |
What Are the Real Odds and Costs of Building?
The real odds of building are worse than most firms assume, and the costs continue long after launch. The Standish Group's 2020 CHAOS report found that 31 percent of software projects succeeded, 50 percent were challenged, meaning late, over budget, or short on features, and 19 percent failed outright. Building a layer you did not need to build spends those odds for no strategic return.
The cost that surprises firms is not the build. It is the maintenance. Industry benchmarks put ongoing maintenance of custom software at 15 to 20 percent of the original build cost per year, before a single new feature is added. Work a rough example: a custom extraction pipeline that costs 400,000 dollars to build carries roughly 60,000 to 80,000 dollars a year to keep running, which over five years is 300,000 to 400,000 dollars of maintenance on top of the original 400,000. That is the price of owning a commodity layer. It buys no edge, because document extraction does not distinguish your firm from any other, and it commits engineering headcount, which most CRE firms do not have, to keeping the lights on rather than improving how the firm decides.
What Should a CRE Firm Build, and What Should It Buy?
A CRE firm should build the thin layer that is uniquely its own, its data model and workflow logic, and buy every layer beneath it. The buy list is long and cheap in strategic terms: storage, warehouse, and the proptech tools that handle extraction and infrastructure. The build list is short and expensive to get wrong, so it should stay short.
The test for what to build is a single question: would two well-run CRE firms want this layer to work identically. If yes, buy it, because you are not competing there and building it only means maintaining a worse version of something a vendor already runs. Document extraction is the clearest example. Every firm wants a lease read accurately, so the requirements converge, which is why it is a category, not a differentiator. Now ask the same question of your buy box, your screening criteria, your deal-scoring model. Two good firms would want those to work differently, because they reflect different theses and different risk appetites. That divergence is the signal to build. The discipline is to keep the build list to only those layers and resist the temptation to build the plumbing around them.
Build the layer that encodes how your firm thinks. Buy every layer that only needs to work. Firms lose years building the plumbing and never get to the part that was theirs.
When Does Building the Whole Platform Make Sense?
Building the whole platform makes sense in a narrow set of cases: when your firm is large enough to fund and staff a permanent engineering team, when no vendor serves your specific requirements, and when the platform itself is part of the product you sell rather than a tool you use internally. For most operating CRE firms, none of these hold.
The scale threshold is the hard one. A platform is not a project that ends. It is a permanent obligation, subject to the 15 to 20 percent annual maintenance benchmark and the 69 percent challenged-or-failed base rate from the CHAOS data. Carrying that obligation requires a standing team, and a standing engineering team is a fixed cost that only pencils at size or when software is the business. A vertically integrated firm building proptech to sell it is a different animal from an operator who needs clean data to underwrite faster. The operator's edge is in the decisions, not the infrastructure, so the operator should own the decision layer and rent everything else. Building the whole platform to avoid a subscription is a false economy: it trades a predictable operating cost for an unpredictable capital project with better-than-even odds of disappointing.
Frequently Asked Questions
Should a CRE firm build or buy its data platform? Neither as a whole. Build the one layer that encodes your firm's edge, your canonical data model and workflow logic, and buy the commodity layers every firm needs and no firm competes on: storage, data warehouse, and document extraction. Framing it as one all-or-nothing decision is the core mistake.
What is the true cost of building custom CRE software? Beyond the build, industry benchmarks put ongoing maintenance at 15 to 20 percent of the original cost per year, indefinitely. A 400,000 dollar build implies roughly 60,000 to 80,000 dollars a year to maintain, and the Standish Group's 2020 CHAOS report found only 31 percent of software projects succeed, with 50 percent challenged and 19 percent failed.
What should a CRE firm never build itself? A firm should almost never build document extraction, a data warehouse, OCR, or infrastructure. These are general problems every firm shares, which makes them solved categories with mature vendors. Building them yields no competitive edge and commits scarce engineering effort to maintaining a commodity.
Conclusion
The proptech build versus buy question has a cleaner answer once you stop treating a data platform as one thing. It is a stack of layers, and only one of them carries your firm's edge: the data model and workflow logic that encode how you underwrite, screen, and decide. That layer is worth building. The layers beneath it, storage, warehouse, and extraction, are commodities every firm shares, and building them buys no advantage while committing you to the odds and the maintenance drag that sink most software projects.
The firms that get this right stay narrow on what they build and generous on what they buy, so their engineering effort compounds on the decisions that are theirs. The firms that get it wrong spend a year building plumbing, carry the maintenance forever, and never reach the layer that was the point. Build vs buy was never the question. Which layer is your edge is the question, and the answer is almost never the plumbing.
Related
What Belongs in a Buy Box, and How AI Screens 500 Deals a Week