The institutional firms publish market research for a reason, and it is not vanity. A CBRE or JLL submarket report is a business-development instrument: it puts the firm’s name in front of every principal, lender, and broker who wants to know where rents and vacancy are heading, and it signals that the firm knows the market cold. Most 4-to-20-person firms assume that game is closed to them — no research desk, no six-figure data license, no analyst pool. It is not closed. A lean firm can produce a credible, publishable market report on a narrow slice of its own market, and the same work that makes the report publishable is the market thesis it should be underwriting deals against. The report pays twice. This piece gives you the framework to produce one without a research team, and draws the exact line where AI belongs in the work and where it must not.
Why a small firm should publish research at all
A published market report does two jobs for a lean firm, and the second one is usually the reason principals resist it. The first job is positioning. When a broker, a lender, or a prospective seller wants a read on where your submarket is going, a quarterly report with your name on it makes you the answer to that question. It compounds the same way a good broker relationship does: the firm that supplies the market view gets the next call. For a small shop competing against national brands, being the demonstrable local expert on one corner of the market is a sharper edge than trying to look bigger than you are.
The second job is discipline. To publish a defensible view of your submarket, you have to actually form one — assemble the rent trend, the absorption, the pipeline, the cap-rate drift — and commit it to writing where a peer could challenge it. Most lean firms carry that market view in the principal’s head, undocumented and inconsistent. Writing it down for an audience forces the rigor that sloppy internal work never demands. That rigor is the same rigor good underwriting needs, which is why the report and the deal work are not two projects. They are one.
The resistance is understandable: research feels like a cost center, the kind of thing a firm does once it can afford an analyst to spare. The framework below inverts that. It treats the report as a byproduct of market work you should be doing anyway, made publishable by a production method a small team can actually run.
What “publishable” actually requires
“Publishable” is a specific bar, and it is lower than principals fear and higher than the AI-content crowd pretends. It does not require the production values of an institutional research desk. It requires three things: a genuine point of view, numbers a reader can trust, and a scope you can credibly own.
A point of view means the report says something — “infill industrial in this metro is tightening faster than the headline vacancy suggests, and here is why” — rather than restating public statistics. Trustworthy numbers means every figure in the report traces to a source a skeptical reader would accept, and none of them is invented or fudged. Credible scope means you are not claiming to cover a whole metro’s office market against firms with a hundred analysts; you are covering the corner where your own deal flow gives you a real edge.
Miss any of the three and the report backfires. A report with no view is noise. A report with a wrong number is worse than no report — it tells every reader who catches the error that your underwriting is sloppy too. And a report that overreaches its scope invites the exact comparison to the institutional desks that you lose. The framework is built to hit all three on a small budget.
The Market Report Framework
Five moves take you from a market view in a principal’s head to a report you can put your name on:
- Scope to a corner you own — one property type, one submarket, where your deal flow is proprietary data.
- Inventory your data before you buy any — start from what you already touch, then fill gaps with public and licensed sources.
- Draft and structure with AI — use a general assistant to organize, draft prose, and stress-test the argument.
- Source and verify every figure — every number traces to a source; the model never invents one.
- Publish on a cadence — a modest report on a reliable schedule beats a lavish one-off.
Moves 1 and 2 decide whether the report is credible. Moves 3 and 4 decide whether a lean team can actually produce it without drowning. Move 5 decides whether it compounds. Work them in order.
Move 1: Scope to a corner you own
The single most common mistake is scoping too broad. A small firm that publishes “The [Metro] Commercial Real Estate Report” is picking a fight it cannot win — the reader mentally files it next to CBRE’s version and finds yours thinner. The winning move is the opposite: go so narrow that your firm is the natural authority. Infill industrial under 50,000 square feet in three named submarkets. Neighborhood retail along one corridor. Medical office in a specific county. Class-B multifamily in a single MSA’s inner ring.
The test for the right scope is simple: pick the corner where your own transaction activity gives you data no national desk has. If your firm brokers or buys in that niche, you see the deals that never hit CoStar, the rents in leases that never get reported, the buyer behavior that shows up months before it registers in a public index. That proprietary view is the entire reason a reader would choose your report over a national one. Scope to where that edge is real, and the report writes from strength. Choosing the right comparable set for that niche is its own discipline — the same one that governs good valuation work, covered in the comp selection framework for an AI-assisted world.
Move 2: Inventory your data before you buy any
Principals assume publishable research requires an expensive data subscription. Start the other way. Before you price a single license, inventory what your firm already holds: your closed and tracked deals, the comps you have pulled, the rents in leases you have touched, the broker conversations in your CRM, the tours and LOIs that show demand ahead of the data. For a firm operating in its niche, that internal record is often richer on the specific corner than any purchased dataset — it is the proprietary layer that makes the report yours.
Then fill the gaps with public sources before paid ones. Employment and permit data from the BLS and Census, submarket statistics from NAR, NAIOP, and MBA releases, and the macro framing in the annual Deloitte Commercial Real Estate Outlook or the major brokerages’ own published reports cost nothing and are cite-safe. Where you genuinely need paid data — verified comps at scale from CoStar or Crexi, foot-traffic patterns from Placer.ai, rent benchmarking from a service like HelloData — buy narrowly for the one figure you cannot source otherwise, and verify what each tool actually reports against its current documentation, because proptech feature sets change quarterly. The goal is a report grounded first in data you own, topped up with public data, and licensing only the last mile.
Move 3: Draft and structure with AI
This is where a lean team gets its time back — which is to say, where the production hours collapse. A general assistant like ChatGPT, Claude, or Gemini will not gather your market data or form your thesis, but it will do the work that usually makes a small firm abandon the report halfway: turning a pile of notes, spreadsheets, and figures into a clean, structured draft.
Point it at your assembled inputs and it will propose an outline, draft the narrative sections around the numbers you supply, tighten prose, and produce the executive summary a busy reader will actually read. Give it your figures and ask it to describe the trend they show; give it your thesis and ask it to argue the other side so you can find the weak point before a reader does. That stress-test is one of the highest-value uses — a model is a tireless, unsentimental critic of your own argument. The same triage-and-drafting speed that changes deal screening, quantified in the real cost of screening deals by hand, is what makes a quarterly report sustainable instead of a heroic one-off.
What the assistant produces is a first draft and a structure, fast. What it produces is not fact. That distinction is the whole next move.
Move 4: Source and verify every figure
A language model will state a vacancy rate, an absorption number, or a rent-growth figure with complete fluency whether or not it is true. In casual work that is a nuisance. In a research artifact carrying your firm’s name, a single fabricated number is a credibility event — the reader who catches it now doubts every figure, and by extension your underwriting. So the rule is absolute: the model drafts the prose, but every number in the report comes from a source you assembled, and every one traces back to it.
The practical discipline is a source line for each figure. Before a number goes in the report, it links to your deal record, a named public dataset, or a licensed source you can cite. Anything the model produced that you cannot tie to a source does not go in — not softened, not hedged, removed. This is the same grounding discipline that keeps AI safe in underwriting, where a confidently wrong figure prices a bad offer; it is spelled out across the ten rules of AI-assisted underwriting. A report is a slower-moving version of the same risk, which makes the verification step non-negotiable rather than optional polish.
One more guardrail for a firm with no IT department: your deal data and unpublished market view are confidential. Before you put proprietary figures into any AI tool, get in writing where the data is stored, whether your inputs train shared models, and how you delete your history. Major providers state that business-tier and API data is not used for training by default, but the contract is your safeguard, not the marketing copy.
Move 5: Publish on a cadence
A modest report on a reliable schedule beats a lavish report that ships once and dies. The value of published research is cumulative: the second report references the first, the fourth shows a trend the reader now trusts you to track, and by the eighth you are the standing source on your corner. A quarterly rhythm is plenty for most submarkets, and it is achievable precisely because the framework front-loads the hard part. Once the scope, the data inventory, and the AI drafting workflow exist, each subsequent issue is an update, not a rebuild.
Distribution does not need a marketing team. The report goes to your existing broker, lender, and client relationships directly, lives on your site where it can be found, and gives you a reason to reach back out to every contact each quarter with something useful rather than a check-in. That cadence is the same operating rhythm that makes a lean shop punch above its headcount elsewhere; inside a small shop’s AI-augmented deal pipeline shows how the research cadence sits alongside the deal cadence in a single week.
Where AI helps — and where it must not
The division of labor is narrow and firm. AI does the assembly, structuring, and drafting; a human owns the market view and every fact.
| Report task | AI does this well | A human must own this |
|---|---|---|
| Structure | Propose an outline, organize inputs into sections | Decide what the report argues |
| Drafting | Turn figures and notes into clean prose and a summary | Verify the prose reflects the data honestly |
| Analysis support | Describe trends in supplied data, surface patterns | Form and defend the thesis |
| Stress-testing | Argue the counter-case, find weak points | Judge whether the thesis survives |
| Figures | Format and present numbers you provide | Source every number; remove any it invents |
Cross that line and the report stops being research. A model has no proprietary transaction data and no view of your submarket; asked to supply a number it does not have, it will produce a plausible one. Keep it doing the writing and let it never do the knowing, and it turns a report that used to eat a week into one that fits a lean firm’s quarter.
What it costs to stand this up
The honest cost has two tiers. The first is getting your team fluent enough to run this themselves — prompting a general assistant well, feeding it your data safely, and knowing exactly what to verify before anything ships. That is a workshop-scale investment; market rates for that kind of AI-fluency training run roughly $2–15K, and for most firms it is the whole cost, because the framework runs on tools you already have and data you already own.
The second tier only applies if you decide to systematize the production into a repeatable custom workflow — a pipeline that pulls your deal data, drafts the recurring sections, and assembles each quarter’s issue with less hands-on time. That is a build, typically in the tens of thousands to low six figures depending on how many sources it touches and how automated you want it. Most firms should not start there. Prove the report is worth publishing with the workshop-scale version first, on a couple of quarterly issues, then decide whether the volume justifies building the machine. The buy-versus-build call gets easier once you have run the manual version and know where the real time goes.
The report is your underwriting thesis
The reason this work pays twice is that a market report and an underwriting thesis are the same document pointed at two audiences. When you underwrite a deal, you are betting on a market view — where rents go, how fast space absorbs, where cap rates drift. That is exactly the view the report commits to writing. Producing the report forces you to hold that view to a publishable standard; underwriting a deal spends that view. A firm that publishes rigorous research is a firm whose underwriting rests on a documented, stress-tested thesis instead of a principal’s gut, and that is a materially better place to underwrite from.
That is the through-line: research and deal analysis are one workflow, and both get sharper when the market view is written down and defended. For how the full analysis stack fits together — screening, comps, underwriting, and the market view that anchors all of it — the deal-analysis playbook for lean teams sequences the stages. And the broader pattern, where a small firm out-operates larger ones by pointing scarce judgment at judgment work and handing the mechanical work to AI, is the argument of the small-firm manifesto. A published market report is one of the cleanest examples of that pattern in practice.
The first step is not a data subscription or a research hire. It is picking the one corner of your market you already own, and writing down what you already know about it to a standard you would put your name on.
FAQ
What is a market report framework for a small CRE firm?
It is a repeatable method for producing a credible, publishable market report without a research desk. The framework is five moves: scope to a narrow corner of the market where your own deal flow is proprietary data; inventory the data you already hold before buying any; use a general AI assistant to structure and draft; source and verify every figure to a real source; and publish on a reliable cadence. The point is that a 4-to-20-person firm can produce research that positions it as the local authority and sharpens its own underwriting, using tools it already has and data it already owns.
Can a firm without a research team really publish market research?
Yes, if it scopes narrowly. A small firm cannot out-cover a national research desk on a whole metro, but it can own one property type in one submarket where its transaction activity gives it data no national desk has. That proprietary view — the deals, rents, and demand signals a small firm sees first — is exactly what makes its report worth reading over a bigger firm’s. The framework front-loads scope and data inventory so each quarterly issue is an update rather than a rebuild, which is what makes it sustainable for a lean team.
How does AI help produce a market report?
AI handles the production work that usually makes small firms abandon the report: structuring inputs into an outline, drafting the narrative around your figures, writing the executive summary, and stress-testing your thesis by arguing the counter-case. A general assistant like ChatGPT, Claude, or Gemini does this in a fraction of the manual time. What it does not do is gather your market data, form your thesis, or supply any number — those stay human. Used this way, AI turns a report that used to consume a week into one that fits a lean firm’s quarter.
Will AI invent numbers in my market report?
It can, and that is the central risk. A language model will state a vacancy rate or absorption figure with full confidence whether or not it is true. In a report carrying your firm’s name, one fabricated number is a credibility event — a reader who catches it doubts every figure and your underwriting with it. The safeguard is a strict rule: the model drafts prose, but every number comes from a source you assembled and traces back to it. Anything the model produced that you cannot tie to a source gets removed, not softened.
What data do I need to write a market report?
Start with data you already own: your closed and tracked deals, pulled comps, lease rents, CRM broker conversations, and the tours and LOIs that show demand before it hits public indices. For a firm in its niche, that internal record is often richer than any purchased dataset. Fill gaps with free public sources — BLS and Census data, NAR, NAIOP, and MBA releases, the Deloitte CRE Outlook — and only license paid data like CoStar, Crexi, or Placer.ai for the last-mile figures you cannot source otherwise. The report should rest first on data you own.
How much does it cost to produce publishable market research with AI?
Two tiers. Getting your team fluent enough to run the framework themselves — prompting well, feeding data safely, verifying output — is workshop-scale, with market rates for that kind of AI-fluency training running roughly $2–15K, and for most firms that is the entire cost. Systematizing the production into a custom automated workflow is a larger build, typically tens of thousands to low six figures depending on scope. Most firms should prove the report manually on a couple of issues before deciding whether the volume justifies building the machine.
Is it safe to put my deal data into an AI tool for research?
Only under the right contract. Your deal data and unpublished market view are confidential, so before uploading proprietary figures, get three answers in writing: where the data is stored, whether your inputs train shared models, and how you delete and export your history. Major providers state that business-tier and API data is not used for training by default, but for a firm with no IT department the contract is the safeguard, not the marketing copy. Treat vague answers as a reason to keep proprietary data out of that tool.
How is a market report connected to deal underwriting?
They rest on the same thing: a view of where rents, absorption, and cap rates are heading. Underwriting a deal spends that view; publishing a report forces you to hold it to a defensible standard. A firm that publishes rigorous research underwrites from a documented, stress-tested market thesis instead of a principal’s gut, which is a better position to price deals from. That is why the report pays twice — it is a business-development asset and the underwriting thesis for your deal flow at the same time.
How often should a small firm publish a market report?
Quarterly is enough for most submarkets and achievable because the framework front-loads the hard work. The value is cumulative: each issue references the last, and by the time you have published several, you are the standing source on your corner of the market. A modest report on a reliable schedule beats a lavish one that ships once. Distribution needs no marketing team — send it directly to your broker, lender, and client relationships, which also gives you a quarterly reason to reach back out with something useful.
Key takeaways
- A published market report does double duty for a lean CRE firm: it positions you as the local authority and forces the market view your underwriting should already rest on.
- Scope narrowly to a corner you own — one property type, one submarket where your deal flow is proprietary data — so a reader chooses your report over a national desk’s.
- Inventory the data you already hold before licensing any; ground the report in your own deals, top up with free public sources, and buy paid data only for the last mile.
- AI structures, drafts, and stress-tests the report fast, but a human forms the thesis and sources every figure. A model will invent a number with full confidence — verify all of them to source and remove any you cannot.
- Standing this up is workshop-scale for most firms (market rates roughly $2–15K for AI-fluency training); systematizing it into a custom workflow is a larger build to weigh only after the manual version proves out.
Want to know whether your firm has the data and the workflow to publish research that actually sharpens your deal flow? A short assessment maps that against your submarket, your deal volume, and where AI belongs without introducing risk. Book your free AI-readiness assessment →
Arthur Wandzel
