An 8-person commercial real estate firm sees more deals in a year than it will ever close — and almost all of that intelligence evaporates. The broker blast you screened and passed on in March, the seller who wanted 20% too much, the market where cap rates moved before anyone else noticed: that judgment lives in one person’s inbox and one person’s memory, and it leaves the building when they do. A proprietary deal database is the discipline of capturing that judgment in a structured, queryable form your whole firm — and, increasingly, your AI tools — can draw on. The instinct is to treat it as something only institutional shops need. The opposite is true. At 8 people, your institutional memory is at its most fragile, and the cost of building the habit is at its lowest. This is the case for doing it now.
The short answer
Build a proprietary deal database as soon as your firm screens more deals than any one person can hold in their head — which, for most acquisition and brokerage shops, is well before you hit ten people. The database is not market data you rent; it is the record of what your firm saw, underwrote, and decided, including every deal you passed on and why. Its value compounds: each deal you log makes the next one faster to judge, and after a year you own a searchable history of your own market that no subscription sells. The reason to start at 8 people rather than 80 is that the habit is cheap to build when volume is low and painful to retrofit when it is high — and the memory you fail to capture now is gone for good.
What a proprietary deal database actually is
The phrase gets confused with two things it is not, so start by drawing the lines.
It is not a market-data platform. CoStar, Crexi, and their peers sell you comps and ownership records about the whole market — data you rent, on their terms. That is a different asset with a different decision behind it, one we work through in detail in when a market-data subscription is enough and when you need your own data layer. A proprietary deal database holds something no platform can sell you: your own firm’s judgment on specific deals.
It is not a CRM contact list. A CRM tracks people and relationships. A deal database tracks deals as structured records — asset type, submarket, ask price, your underwritten value, the cap rate, the debt assumptions, who sourced it, the date, the decision, and the one field that matters most and almost never gets recorded: why you passed. A contact record tells you that a broker exists. A deal record tells you that in Q1 you underwrote their industrial listing to a 6.8% cap, they wanted a 6.0%, and you walked — context worth more than gold the next time that same asset trades.
At its simplest, a proprietary deal database is a single structured source of truth for every opportunity that crossed your desk, live or dead. Deal-management platforms built for larger teams describe exactly this outcome — logging both active and inactive deals to “maintain a complete record of opportunities evaluated,” which lets a firm “build institutional knowledge and analyze sourcing trends, target markets, and strategy over time” (Dealpath). The insight is right. The assumption that you need an enterprise platform to capture it is what this piece exists to correct.
Why 8 people is exactly the right size to start
The standard framing treats a proprietary deal database as a luxury that only large firms can afford. Reverse it. The smaller you are, the more of your institutional knowledge is trapped in individual heads, and the more catastrophic it is when one of those heads walks out the door.
At an 8-person firm, there is no analyst bench, no research team, no documented process quietly recording what the firm has learned. When your one acquisitions lead leaves, they take a year of screened deals, seller psychology, and submarket feel with them — and their replacement starts from zero. A larger firm survives that departure because redundancy is baked into its size. You do not have that cushion. The database is your redundancy.
There is a volume argument too, and it cuts in your favor. Acquisition teams routinely review well over a hundred opportunities for every deal they actually close; disciplined shops reject the majority within the first hour on market, pricing, or asset quality (Requity). A large firm generates so much flow that capturing it becomes a real operations project. At 8 people your flow is high enough that the pattern-value is real, but low enough that one person can log a deal in ninety seconds as it comes in. That window — enough signal, low enough volume to capture by hand — is exactly where a lean firm sits and an institutional one does not. Starting now is not premature. It is the only time the habit is genuinely cheap.
The compounding asset your inbox throws away
Email is where deal intelligence goes to die. A broker blast arrives, someone skims it, replies or doesn’t, and the whole thing is unrecoverable inside a week. The deal you passed on leaves no trace of why. Six months later the same building comes back at a new price and nobody remembers the last round.
A structured record breaks that cycle, and the value accrues in three directions at once:
- Faster judgment. When a new deal arrives, you compare it against your own history instantly — what similar assets in that submarket actually underwrote to, what sellers there have accepted, where you have been burned. That is a real edge on speed, which for a lean firm is often the difference between winning work and losing it, a theme we develop across the deal-analysis playbook for lean CRE teams.
- A pass list that becomes a pipeline. Every deal you declined at the wrong price is a future opportunity at the right one. A queryable record of “passed, too expensive, spring 2026” is a follow-up list your competitors do not have, because theirs is scattered across inboxes.
- Proof of your own market read. Over a year, the database becomes evidence — of where cap rates moved before the reports caught up, of which sources send you real deals versus noise. That is proprietary intelligence about the markets you own, the same edge we argue lean firms should build rather than rent throughout the small-firm AI manifesto.
None of this requires more deals. It requires not throwing away the intelligence in the deals you already see. The compounding is the point: a database with one deal in it is worthless, and a database with four hundred is an asset no amount of money can buy off the shelf, because it is a record of your firm’s decisions.
Why this is now an AI question
For most of CRE history, “keep better records” was good advice that firms ignored without much penalty. That has changed, and the reason is AI. A general assistant like ChatGPT, Claude, or Microsoft Copilot can now screen a deal, draft a preliminary underwrite, or summarize a market — but only against the context you give it. Point it at a raw broker email and it produces generic output. Point it at your structured deal history and it produces output that reflects your firm’s judgment.
An assistant with no memory of what your firm has underwritten cannot tell you “this ask is 15% above what you paid for three comparable assets last year.” A structured deal database is what supplies that memory. It turns a generic model into something that reasons the way your firm reasons — because it is grounded in your firm’s actual decisions. The practical mechanics of feeding that history into a screening workflow are what we walk through in inside a small shop’s AI-augmented deal pipeline, and the discipline of consistent inputs is exactly what makes the automation described in lessons from automating deal intake for an acquisitions team hold up in production.
This is why the database is no longer optional hygiene. It is the substrate AI screening runs on. Deloitte’s 2026 Commercial Real Estate Outlook — drawn from more than 850 executives across 13 countries — frames data-and-AI capability as the line separating firms pulling ahead from those falling behind (Deloitte). A firm that has spent a year structuring its deal flow can put AI to work on day one. A firm that has not is starting from an empty page while its assistant guesses.
What it costs — and why it isn’t an engineering project
The reason firms hesitate is a category error: they picture a $150,000 custom build. That number belongs to a different problem. A proprietary deal database, done right for a lean firm, starts as disciplined capture inside a tool you likely already pay for.
Here is the honest cost picture in market ranges:
| Approach | Typical market cost | What you get |
|---|---|---|
| Structured spreadsheet or database tool | A tool you already own, plus the discipline to fill it | A queryable deal record, if you enforce the habit |
| Purpose-built deal-management platform | Roughly a few hundred dollars per user per month | Pipeline structure, reporting, and audit trail built for the job |
| Custom automation on top (intake, enrichment, scoring) | Roughly $25K–$150K to build | Auto-captured, AI-scored deal flow with no manual entry |
Most 8-person firms should start at the top of that table, not the bottom. The first version of a proprietary deal database is a well-designed structured sheet with a fixed set of fields and a rule that nothing gets screened without a row. The platform tier — Dealpath and its peers — earns its cost once volume and team size outgrow a spreadsheet’s discipline, and those platforms are explicitly built so that “as your firm screens new deals, you can build a proprietary deal database of market intelligence to support future decisions” (Dealpath). The custom automation tier is where you land only after the manual habit has proven its value and manual entry has become the bottleneck.
The expensive version is not the one you build. It is the year of deal flow you fail to capture because you were waiting for the expensive version.
The objections, answered honestly
“We’re too small — this is overkill.” The opposite, as covered above: small firms have the most fragile institutional memory and the cheapest window to capture it. Overkill is buying an enterprise platform at 8 people. A structured sheet is not overkill; it is the minimum.
“Our CRM already does this.” It almost certainly does not. A CRM tracks contacts and communications, not deal economics and pass reasons as structured, comparable fields. If your CRM genuinely captures ask price, underwritten value, cap rate, and decision rationale per deal in a queryable form, you already have the start of a deal database — most do not.
“We don’t have time to log every deal.” Logging a screened deal is a ninety-second task if the fields are fixed and the habit is enforced. The time you spend is trivial against the time you lose re-underwriting a deal you already judged and forgot. The failure mode is not the ninety seconds; it is inconsistent capture, which produces a database nobody trusts.
“The data will be messy and incomplete.” Some of it will be, and that is fine. A deal record with the asset type, submarket, price, and pass reason is useful even if half the other fields are blank. Perfect capture is not the bar. Consistent capture of the few fields that drive judgment is.
How to start in 30 days
You do not need a vendor, a consultant, or a budget cycle to begin. You need a schema and a rule.
- Week 1 — define ten fields. Date, source, asset type, submarket, ask price, your underwritten value, cap rate, decision, pass reason, and a free-text note. Ten fields, no more. Put them in a structured sheet or a database tool the firm already has.
- Week 2 — set the rule. Nothing gets screened without a row. Assign one owner to enforce it. The rule matters more than the tool; a perfect tool with inconsistent entry is worthless.
- Weeks 3–4 — backfill and review. Reconstruct the last quarter’s deals from memory and inboxes while they are still recoverable, then hold a fifteen-minute weekly review of what came in and what you passed on. That review is where the pattern-value first shows up.
After the habit sticks, two upgrades are worth considering, in order. First, point a business-tier AI assistant at your accumulated records to draft first-pass screens against your own history. Second — only once manual entry is the bottleneck — commission automation to capture and score inbound deals, the custom-automation tier that lands in the $25K–$150K market range. The sequence matters: the discipline comes first, the tooling second, the automation last. A firm that inverts that order buys a platform and fills it with nothing.
FAQ
What is a proprietary deal database in commercial real estate?
It is a structured, firm-owned record of every deal your firm has evaluated — active and dead — capturing asset type, submarket, ask price, your underwritten value, cap rate, source, decision, and the reason you passed. Unlike a market-data subscription, it holds your firm’s own judgment on specific deals, which no vendor can sell you. Unlike a CRM, it records deal economics and decisions rather than contacts and communications.
Why would an 8-person firm need a proprietary deal database?
Because a small firm’s institutional memory lives entirely in individual heads and disappears when a person leaves. At 8 people you have no analyst bench or documented process backing you up, so the database becomes your redundancy. Your deal volume is also high enough that the patterns are real but low enough that one person can capture each deal by hand — the exact window where the habit is cheap to build.
How is a deal database different from CoStar or Crexi?
CoStar and Crexi are market-data platforms you rent — they sell comps and ownership records about the whole market on their terms. A proprietary deal database holds what your firm saw and decided: your underwrites, your pass reasons, your sources. One is rented breadth about the market; the other is owned depth about your firm’s own judgment. Most firms need both, for different reasons.
Can’t our CRM do this already?
Usually not. A CRM tracks contacts, relationships, and communications, not deal economics and pass reasons as structured, comparable fields. If your CRM captures ask price, underwritten value, cap rate, and decision rationale per deal in a queryable form, you have the beginnings of a deal database. Most CRMs record that a broker exists, not that you underwrote their listing to a 6.8% cap and walked.
How much does it cost to build a proprietary deal database?
The first version costs almost nothing beyond discipline — a structured sheet with fixed fields in a tool you already own. A purpose-built deal-management platform runs roughly a few hundred dollars per user per month. Custom automation that captures and scores inbound deals lands in the roughly $25K–$150K market range and is worth it only after manual capture has proven its value and become the bottleneck.
What fields should a deal database capture?
Start with ten: date, source, asset type, submarket, ask price, your underwritten value, cap rate, decision, pass reason, and a free-text note. The single most valuable and most-neglected field is the reason you passed — it turns declined deals into a future pipeline and records your firm’s judgment over time. Keep the field set small so capture stays fast and consistent.
How does a deal database improve AI deal screening?
A general AI assistant screens deals only as well as the context it is given. Fed a raw broker email, it produces generic output. Fed your structured deal history, it can flag that an ask sits 15% above what you paid for comparable assets last year and reason the way your firm reasons. The database is the memory layer that grounds AI screening in your firm’s actual decisions rather than the open internet.
Isn’t logging every deal too time-consuming for a lean team?
Logging a screened deal takes about ninety seconds when the fields are fixed and the habit is enforced — trivial against the cost of re-underwriting a deal you already judged and forgot. The real risk is not the time per deal; it is inconsistent capture, which produces a record nobody trusts. One owner enforcing one rule solves that.
When should we move from a spreadsheet to a real platform?
Move when the spreadsheet’s discipline breaks down — typically when deal volume or team size makes manual consistency unreliable, or when multiple people need concurrent, audited access. A purpose-built deal-management platform earns its cost at that point. Below it, a well-designed structured sheet with an enforced entry rule is enough, and jumping to a platform first usually produces an expensive database filled with nothing.
What’s the biggest mistake firms make with deal databases?
Waiting for the perfect system. Firms delay because they picture an expensive custom build, and in the meantime a year of deal flow evaporates from inboxes. The intelligence you fail to capture now is gone for good. Start with ten fields and one rule this month; upgrade the tooling later, once the habit has proven its worth.
Key takeaways
- A proprietary deal database is your firm’s own structured record of every deal seen, underwritten, and passed on — not rented market data, and not a CRM contact list.
- At 8 people your institutional memory is at its most fragile and cheapest to capture; the database is the redundancy a small firm otherwise lacks.
- The compounding value is faster judgment, a pass list that becomes a pipeline, and proof of your own market read that no subscription sells.
- AI deal screening is only as good as the deal history you can feed it; the database is the memory layer that makes an assistant reason like your firm.
- It is not a $150K engineering project — the first version is a structured sheet with ten fields and a rule that nothing gets screened without a row.
- Start the discipline first, add a platform when volume outgrows the sheet, and commission automation last; inverting that order buys an empty database.
Not sure what your firm’s deal flow is already telling you — or what you’re throwing away each week? A short, free AI-readiness assessment maps how you screen deals today and shows where a structured record would pay off first. Book your free AI-readiness assessment → and we will size the smallest version worth building for your firm.
Dirk Jan van Veen, PhD
