The Location Screen as a Kill-Shot, Not Just a Checklist
Many in CRE tech are racing to speed up underwriting. They’re solving the wrong problem.
AI-powered models. Auto-populated rent rolls. Instant cap rate sensitivities. The whole industry is sprinting to compress the time from offering memo to the investment committee.
Underwriting speed is not the bottleneck for the deals that actually matter. Selection is.
If you underwrite a deal that never should have crossed your desk, you’ve just burned time and tokens.
The leverage is upstream. It’s in the filter that decides what gets the spreadsheet treatment in the first place. And the first filter, the one that has to fire before any number gets touched, is location. Get it wrong, and no underwriting model in the world saves you. Get it right, and a mediocre deal still makes you money.
The Augusta lesson
We had a screening rule we’d used for years. New Starbucks going in nearby. Capital follows capital, and Starbucks runs serious site selection work before signing a lease. So when we saw a fresh Starbucks in a small retail cluster near the asset, we treated it as confirmation. The Starbucks didn’t make us buy the deal. It made us more confident in a deal we were already underwriting.
That’s the trap. Confirmation feels like analysis when you’re already leaning toward yes.
The Starbucks sat on the right side of the tracks. Our property sat on the wrong side. Same zip code, different world.
We drove the area. No abandoned homes. No dead retail strips. No visible gang activity. The telltale signs weren’t there during business hours on a weekday.
We didn’t know the market well enough to read what wasn’t there. The rule that had worked elsewhere validated a decision it should have killed.
Insider knowledge is an edge you can’t fake
If you didn’t grow up in a market, you don’t know where the lines are. You don’t know which streets locals avoid after dark. You don’t know which side of which road carries the perception that quietly kills lease-ups.
Our default is to buy in markets we know. When we step outside that circle, we need a screen that catches what local intuition would catch for free. And we need to be honest that no algorithm replaces a local. What an algorithm can do is organize the questions you’d otherwise forget to ask.
A paid 45-minute call with a local broker, property manager, and former residents, run through a structured questionnaire, is worth more than ten hours of remote screening.
Many buyers screen for green flags. The discipline is screening for veto reds.
A green-flag screen asks what’s good about a location and adds points. A veto-red screen asks what would kill this deal regardless of price, and looks for any single one of them.
The Augusta Starbucks was a green flag. The wrong side of the tracks was a veto red we never checked for.
Institutional credit memos have always carried deal-killer lists. The discipline isn’t novel at the top of the market. What’s novel: same rigor available before the underwriting team gets handed the deal, in markets where local intuition isn’t there to do the work for free.
So what actually makes a great location?
Eight pillars. Each one earns its place because it either kills the deal when it’s wrong, or it shows up in the data when properties default, vacate, or struggle to lease.
1. Supply pipeline and barriers to entry
The single biggest predictor of near-term performance, and the pillar that earns the top spot. Demand recovers within a cycle. Oversupply takes years to absorb and burns through every operator who underwrote the prior rent roll.
Fannie Mae’s benchmark is that a market is in balance when 2 to 4 percent of inventory is underway. Markets above 10 percent get crushed. Austin, Phoenix, Nashville, Charlotte, Denver, and Tampa all underperformed national rent growth in 2025 and 2026 for exactly this reason.
The screen pulls active permits, units under construction as a percentage of inventory, and absorption trends. Then it asks the harder question. What stops the next wave?
Geographic barriers. Zoning friction. Regulatory complexity. Coastal markets and dense infill submarkets get protected. Suburban greenfield with friendly entitlements gets buried in supply every cycle.
The flip side, which many screens miss. Is this submarket about to be upzoned? Comprehensive plan revisions, transit-oriented development overlays, density pushes from the city council. The wrong side of the tracks today is sometimes the right side in seven years if the political class has decided it should be. Or vice versa. A great location can be ruined by a council that floods it with entitlements after you close.
City planning records and council minutes are public. AI scrapes them in minutes.
2. Crime, diagnosed correctly
The first thing most screens look at and the most misunderstood. The question is not how much crime. It’s where is the crime?
Property-level crime is fixable. You evict the cause, screen the next batch harder, light the parking lot, fence the perimeter, change the locks, and within two years you have a different building. Painful, expensive, fixable.
Area-level crime is something else. You can run the cleanest building on the block, and your prospects will still drive past, see what surrounds you, and lease somewhere else. The asset becomes a hostage to its surroundings.
The screen separates those two situations. Same heat map, completely different futures. One you can fix with capital and time. The other owns you.
3. Perception versus reality
Rents are set by perception, not police reports. A neighborhood with falling crime and a bad reputation will underperform a neighborhood with rising crime and a good reputation for two to three years. The lag is real, and it’s measurable.
What you can pull from afar:
- Google reviews of comp properties within a mile. Sentiment-tagged.
- Local Facebook groups, especially the neighborhood-specific ones.
- Reddit threads on the city’s main subreddit.
- TikTok and Instagram geo-tagged content.
- Local news coverage of the submarket over the last 24 months.
This is where AI earns its keep on signals humans can’t audit at scale. One analyst can’t read 800 reviews of fifteen comp properties. A model can, in an hour, and surface the patterns that show up in renewal rates eighteen months later.
4. Absence of essential services
The pillar that many don’t include and that quietly drives more lease-up failures than crime.
The camera car catches what’s there. It misses what isn’t. No grocery store within a mile. No daycare. No urgent care. No bank branch. No pharmacy. Each absence is a friction point in a prospect’s daily life. Stack three or four and you’ve described a location people leave when their lease ends.
Map every property within a one-mile radius. Score the absences. The wrong side of the tracks in Augusta probably had a different absence profile than the right side, and the map would have shown it without a single phone call.
Cheapest pillar to run. Most likely to surface what daytime drives miss. If you can only run one, run this one.
5. Demographic trajectory
A neighborhood’s median income today tells you almost nothing. The direction of travel tells you everything. Is the prime renter cohort, ages 25 to 44, growing or shrinking? Is household formation positive? Is the population moving in or out?
That’s the trend layer. It compounds over decades.
Underneath the trend sits the cycle. Absorption-to-delivery ratio over the next 24 to 36 months. A submarket can have a positive ten-year demographic trajectory and still get crushed in a three-year window because absorption hasn’t caught up to supply.
Confusing the two is how good operators buy at the top of a local cycle in a structurally fine market. The trend rewards them eventually. The cycle wipes them out before eventually arrives.
For student housing, this gets specific. Enrollment trend at the anchor university is the trend. Institution-specific exposure within the demographic cliff is the cycle. Pull both. They produce different answers.
6. Employment base and diversification
Strong job growth is a green flag. The deeper question is whether the employment is diversified or concentrated. One employer, one industry, one military base. Concentration risk that doesn’t show up until it shows up.
Markets like Kansas City, Cincinnati, and St. Louis posted above-average rent growth through the recent supply wave because their economies are diversified, and supply was constrained. Markets riding one industry get hammered when that industry coughs.
Map proximity. Drive time matters more than zip code.
7. Walkability and access
A 2014 study by Gary Pivo at the University of Arizona analyzed roughly 37,000 Fannie Mae multifamily mortgages and found that walkability reduced default risk in a nonlinear way. The directional finding holds. Very low walkability correlates with very high default risk. The threshold matters more than the score itself.
Take the directional finding. Don’t anchor to the precise percentages. Loan vintages have shifted, work-from-home rewrote some of the math, and Walk Score below 80 covers most suburban garden multifamily anyway.
Access matters separately. How easy is it to physically reach the property? Highway access, transit if relevant, sidewalks, lighting. We do a Street View walkthrough on every screen because the camera car catches things the data misses.
One honest caveat. Most Street View imagery is captured during daylight hours on weekdays. The neighborhood that drove well at 11 AM on a Tuesday is not the neighborhood your prospects experience at 9 PM on a Saturday. Where night-time imagery exists, use it; where it doesn’t, route a question to a local.
8. Institutional anchors and capital signals
Hospitals, universities, government employment centers, and large stable corporations. These anchor tenants stabilize a submarket through cycles, generate consistent renter demand, and rarely move on short notice.
Capital signals layer on top. Where institutional retail goes, household income and rent growth follow. Starbucks, Whole Foods, Trader Joe’s, Chipotle, Target. These tenants run real site selection work before signing leases (and they close locations too, so presence is a signal, not a guarantee). Their decisions are free market research.
Augusta proved you have to know which side of the tracks they landed on. Capital can sit one block away from a problem area without crossing into it. The signal is useful only if you check it against the perception pillar and the absence map.
Where AI actually helps
Back to the original point. Some CRE tech is overweighting one part of the problem. Faster underwriting on a bad deal still produces a bad deal, just sooner. The leverage isn’t only in the model. It’s also in the gate before the model.
That’s where AI earns its keep on the front end. Running every address through the same eight-pillar screen, in every market, the same way every time. Pulling supply pipelines, crime data, demographic trends, perception signals from review sites and social media, absence maps, and capital signals in five to ten minutes. Flagging veto reds. Producing a verdict before anyone touches a spreadsheet.
For our analysts, the value is consistency. Same rigor on a market we know cold and one we’re seeing for the first time. The screen doesn’t replace local knowledge where we have it. It substitutes for local knowledge where we don’t, and forces a check against the green-flag bias that fooled us in Augusta.
The screen isn’t only for buyers. It’s a disposition tool too. If your screen would now veto-red a property you already own, that’s a sell signal. Markets shift. Anchor employers leave. Reputations crater. Supply pipelines fill up. The location that screened green in 2019 might screen red in 2026 and you’d never know unless you ran it. Annually is the right cadence.
If the screen kills enough bad deals at the gate, the underwriting team gets to spend its time on deals that actually matter. That’s the productivity gain no faster spreadsheet model can match. The math depends on your pipeline, your buy box, your markets, and how disciplined you’ve been historically. We’ve found it material. Your mileage will vary.