Why Does Building a Market Survey Eat 40% of a Tenant-Rep Deal?

The 2026 CRE broker productivity research puts a number on the most familiar chore in tenant rep: a market survey takes 4 to 8 hours of manual work, and by these estimates surveys can consume up to 40 percent of the deal process for a tenant-rep broker. Comp reports run about 8 hours for a ten-agent firm doing it by hand. The pattern holds across deliverables because they all draw from the same well.

The hours are not in the thinking. They are in the fetching. A survey means opening CoStar, LoopNet, and Crexi, finding the candidate buildings, copying availabilities and asking rents into a spreadsheet or a template, chasing the pieces those platforms miss with phone calls to listing brokers, then laying it all into a branded document. None of that is analysis. It is data entry against three sources that do not talk to each other.

Then the client asks to add two submarkets, or drop the ones over budget, or refresh it because a month passed. Because the data was never stored in a reusable structure, the revision does not edit a small part of the work. It reruns most of it. The broader time studies show the shape of the problem: the 2026 DNA of CRE Broker Report (Buildout and TheBrokerList) found brokers spend 46 percent of their time on administrative and manual tasks, and ranks missed follow-ups the number one deal-management challenge. The survey is where a large share of that administrative half lives.

Split any survey into three buckets and the waste becomes obvious.

Gather is finding and pulling the raw facts: which buildings, what is available, what is the asking rent, who owns it, what is the loan situation. This is the biggest bucket and the one with the least client value, because the client is paying for your read on the market, not for your ability to retype a LoopNet listing.

Format is turning the raw facts into a clean, branded, client-ready document. It matters for how the work lands, and interactive surveys hold attention far better than a static PDF (one deliverable study found 4:30 average time-on-page for interactive versus 1:15 for PDF). But formatting is a template problem, and template problems are solved once, not per deal.

Judgment is the part clients actually hire you for: which buildings belong on the list, how to frame them for this specific tenant, what the effective rent really is once you net out concessions, which landlord will actually deal. This bucket should be most of your survey time and usually is the smallest, because gather and format crowd it out.

The goal is not to eliminate the survey. It is to move your hours out of gather and format and into judgment, where the value and the fee actually come from.

Partly, and the gap between "partly" and "fully" is exactly where brokers get burned. The tooling is real. ScoutSpace auto-extracts property details into interactive branded surveys, and one broker reported a survey dropping from 4 to 5 hours to about 15 minutes (a customer testimonial on LinkedIn, so treat it as a vendor-favorable data point). Vendors in the category advertise time savings as high as 87 percent, which is a marketing figure, not an audited one.

Here is the catch the demos skip. AI-populated survey data, pulled from CoStar or LoopNet, still needs a broker's verification, and that verification is where the saved time goes to die if you are not careful. The trust data is blunt: 66 percent of CRE pros use AI weekly or daily, but only 5 percent trust it enough to inform an actual deal decision, and 17 percent use it only with heavy verification (First American Data & Analytics and DealGround, 255 professionals, April 2026). A survey that looks like four hours of work but was assembled in fifteen minutes still carries three questions per building: is that block still available, is the rent net of concessions or a headline number, does the ownership flag attach to the real entity. If you hand that survey to a client without checking, one fabricated free-rent figure costs more credibility than the document ever saved in hours. We wrote the mechanics of this up separately as the verification tax.

So AI is genuinely good at one half of the job, format, and unreliable at the half that matters most, judgment and load-bearing facts. Using it well means pointing it at the format problem and keeping your eyes on the facts.

Structure your submarket data once, so the survey becomes assembly instead of archaeology. Five steps, in order.

1. Audit one real survey into gather, format, and judgment hours. You cannot fix a time sink you have not measured, and the split tells you how much of your survey time is recoverable.
2. Build a living block list for the submarkets you actually work. Maintain your own inventory of buildings, availabilities, ownership, and lease terms, updated as you learn things, so each new survey starts from stored data instead of a blank scrape. This is the single change that turns a 6-hour survey into a 1-hour one over time.
3. Make every data point carry its receipt: a source and a date beside each value. When a client questions a number, or a revision lands three weeks later, you check a stored citation instead of re-researching from scratch. This is also what makes AI-assisted work verifiable in seconds rather than hours.
4. Use AI for transformation, not recall. Do not ask a model to go find current availabilities; it will produce confident, wrong ones. Hand it your verified data and a fixed template and let it do the formatting and first-draft narrative, which is what it is reliably good at.
5. Verify the three load-bearing fields on every building before it goes to a client: current availability, effective rent net of concessions, and the real ownership entity. Everything else can be approximate. These three cannot, because they are what the client acts on.

Pick your two or three core submarkets and start the block list there, not everywhere. Populate it from your next three surveys instead of a big upfront build, so the system pays for itself as you work rather than as a project you never start. By the third survey out of the same submarket, most of the gather bucket is already sitting in your list, and the survey becomes a judgment exercise with a formatting step, which is what it should have been all along.

Then measure again. Re-audit a survey against the same gather, format, judgment split you started with. If the gather hours have dropped and your judgment hours have held or grown, the system is working and the client-visible quality is going up, because you are spending your time on the read instead of the retype.

The honest counter is that for a broker doing a handful of surveys a year across scattered submarkets, the overhead of building and maintaining a structured block list is not worth it, and a manual pull for each deal is the rational choice. Systematization pays off with volume and with submarket focus; without both, you are maintaining a database to save time you were not losing at scale.

The second counter is that some of the 40 percent is not waste at all. The best brokers gather data by hand precisely because the act of pulling every availability keeps them close to the market, and a fully automated survey can produce a broker who no longer knows their own inventory. That is a real risk, and it is the reason the judgment bucket should grow, not vanish, as gather shrinks. The point is not to stop touching the data. It is to stop retyping it, and to spend the time you win back on the read the client is actually paying for.