The Site Selection Playbook For An Impossible Market

Vacancy is at a 20-year low. Construction is cratering. Costs are up on every front. Here's how the operators who are still growing are doing it.

In 2025, 8,100 retail stores closed, up 12% from the year before. Yet vacancy rates are sitting at 4–5%, the tightest in decades, and new construction is 63% below 2019 levels. If you're doing retail site selection right now, you're hunting for space in a market that has almost none, while costs are rising on every front.

So I brought together three people who are living this problem every day and asked them the questions I keep hearing from operators but haven't heard good answers to, until now.

• Dale Goss - SVP Real Estate @ Raising Cane’s
• Ryan Redus - VP Real Estate @ Hopdoddy Burger Bar
• Andrew Neelon Andrew Neelon - CEO @ FOURWALL

What emerged from the conversation was less a set of tactics and more a coherent philosophy one that challenges some of the most deeply held assumptions in retail real estate.

Question 1: How do you win in a constrained market?

Retail vacancy is sitting at its lowest in years ~ 4-5%. How do you win in this environment?

Ryan: Speed is a differentiator in a landlord market. Ryan's team can issue an LOI within two to three days. That's not purely hustle, it is because they've already done the work. They've mapped their target sub-markets before any broker calls. When a deal surfaces, the only question is fit.

Andrew: Your approval process is your biggest speed problem. In a landlord's market, a slow internal process is indistinguishable from a no. When Andrew was running site selection for Bonobos inside Walmart's ecosystem, deal approvals required multiple committees. He collapsed that down to a guardrail system: if a site hits defined criteria, it fast-tracks.

Andrew: Perfect sequencing is a luxury a constrained market no longer offers: You can’t hold out for location three before you take location six on your list. Retailers going into class A space especially need to be flexible and opportunistic when a new space presents an opportunity.

Dale: If your AUV is high enough, you don't need to find space. You can create it. Raising Cane's stopped looking for available pads and started scraping shuttered restaurants. The unit economics work at their volume where they wouldn't for most QSR brands. This switches their competition for those sites from other QSR’s to things like banks.

Dale: Patience can be a competitive advantage: Being founder-owned means no quarterly earnings call and no PE clock. If the right site isn't there in a target trade area, Raising Cane's waits. They won't close a deal just to close a deal. Most brands under growth pressure can't say the same,  and landlords know it.

Lyden’s Takeaway: Fast speed comes from slow preparation. In contrast to Mike Tyson’s quote “Everyone has a plan until they get punched in the mouth” - these operators can move fast because they have a plan - making decisions faster while adapting opportunistically.

Question 2: How Do You Navigate Rising Costs?

The real cost of a bad site has effectively doubled. Look at how the market has changed since 2021:

• Build-out costs at $155/psf nationally (+5%)
• 12.7m square feet under construction (-23%)
• National asking rent is $25.29/psf (+22.9%)
• Class A rent gaps are widening (42%)

So how do you navigate growth goals with rising costs?

Andrew: Fix your existing store P&Ls before underwriting your next one. Andrew exports P&Ls line by line to find what he calls “paper cuts”, small margin leaks that add up at scale. The goal isn't just fixing existing stores. It's building the conviction to underwrite a healthier pro forma on the next site than your current fleet average would suggest.

Ryan & Dale: Sizing down the box can outperform the larger format. HopDoddy’s smallest new restaurant (2,800 sq ft vs. their standard 3,500–4,000) is currently #3 in the chain. Dale watches channel sales carefully (mobile orders, pickups, doordash, drive through) - getting the footprint right on the front end is cheaper than fixing.

Dale: Build re-evaluation gates into the deal, not just an approval at committee. Raising Cane's has formal check-ins at three moments: real estate committee, contract signing, and the end of the inspection period. At each one they reserve the right to walk. Committee approval is a conditional green light, not a commitment.

Ryan: Score risk in three separate buckets on every deal. Ryan quantifies risk at the site level, the center level, and the trade area level independently. This stops a strong trade area from masking a weak site, or a great landlord from hiding a bad market.

Andrew: Before you build a forecast, ask what you need to believe first. Given your typical OPEX, staffing, and buildout, what does year-one revenue need to be for this deal to pencil? If that number is obviously impossible, or laughably easy, you don't need a forecast. You need a decision.

Lyden’s Takeaway: Know your numbers cold, your markets deeply, and build risk removal into the deal itself. The panel’s advice lines up with what I have been seeing from winning 2025 retailers like Dutch Bro’s, Ross, and Chili’s winning with smaller footprint-cost effective stores.

Question 3: What Do You Do When The Data Lies?

Site selection is multi-variate. It is rare to find the “perfect” site. With all the new data available like mobile data, credit card data, psychographics, demographics, how do you make a decision when the data conflicts?

Dale: Use data to select trade areas. Use humans to select sites. All of Raising Cane's analytics work — models, segmentation, correlation databases — is about understanding trade area demand. The actual site decision, which corner, which building, which access point, is a human activity. Data narrows the field, humans close the deal.

Andrew: Run the football field - forecast the site four different ways and find where the ranges agree. Andrew borrows a financial valuation method, three or four distinct methodologies, mapped visually to find where their ranges overlap. Consensus across methods is conviction. An outlier on one is a data question worth digging into. This reminds me of Jack Thompson who built Tango Analytics and ran three different models for every site: predictive, proxy, and person.

Dale: When your two forecast models diverge, the investigation begins. Raising Cane's runs Kalibrate alongside an in-house model. When they disagree, it triggers a drill-down into Unique Trade Area population, regional competition, and geodemographic segmentation. Standard models built on census data and co-tenancy won't catch sociological nuance. That's where segmentation earns its keep.

Dale: Mythbusting sacred cows, Average daily traffic is one of the worst predictors of sales. Raising Cane's ran a systematic correlation of every variable they could think of against actual sales. ADT — the metric old-school real estate that has been treated as gospel for decades landed near the bottom. A lot of people driving by their store does not mean a high-volume store. It might be a predictor of success for your business, point is: Don't inherit assumptions. Test them.

Ryan: A predictive model is a seat at the table, not the decision-maker. Ryan's model with Kalibrate runs within about 20 R-squared, useful, but far from a silver bullet. The CEO, COO, and Ryan personally visit every site before signing. The model earns a site a closer look. Humans make the call.

Ryan: Keep evaluating the deal through the entire lease process, not just at LOI. Over a four to six month lease negotiation, the co-tenant mix can change, a competitor can announce nearby, or a construction risk can surface. Ryan treats every lease turn as an ongoing evaluation. The deal can still be killed at any point.

Andrew: Run a pre-mortem before the deal closes. Instead of asking "do we think this works?", ask "if this store fails in year two, what would we look back at right now and say was obviously the warning sign?" That reframe surfaces real risks that optimism tends to suppress in committee.

Andrew: The most underweighted variable in any model is GM quality — and you can't model it. Andrew has watched stores tank 20% in two months when a GM leaves, then recover even higher when a strong one comes in. No forecasting model accounts for this. The implication: never underwrite a site assuming a B-team will run it, and be honest about your bench depth before committing to aggressive growth.

Lyden’s Take: It is important to keep in perspective the deluge of data is there simply to get you in the range of the right neighborhood. But humans getting boots on the ground is ultimately the final decision maker.

Question 4: How Are You Using AI… Really?

Andrew: When a 30 page broker tour book lands in his inbox, he dumps it into Claude and prompts it to build an interactive web app. It plots a driving route, rent vs. square footage, etc. The insight here is to use AI to transform other people’s work into a useful format for you.

Ryan: The layup here is lease portfolio management. Tracking expiration dates, option windows, etc. across the portfolio. This is an opportunity for an agent to push you notifications when you are managing a larger portfolio so you don’t miss key dates. Similarly, Dale mentioned with 850+ active leases being able to load those up and query it “what is our standard position on this clause” is massively time saving.

Dale: The Raising Canes team built a second forecasting model using ChatGPT and now runs alongside their Kalibrate model. It is now a legitimate second opinion. I’ll be curious if we see more of these home-brew models or if it only works when you have the leverage of 850 data points.

Lyden’s Take: I was genuinely surprised that we are seeing real valuable use cases in retail site selection. My sense was AI had failed in the past to understand Tobler’s Law - that things are related spatially when they are in closer proximity to each other - but use cases like lease and legal management do not have the spatial component. And perhaps AI can use other things to approximate the map.

Summation

Dale said it best: "We use data to select trade areas. But we use humans to select sites."

The models, the segmentation, the correlations exist to get you to the right neighborhood. But the operator who wins in 2026 is the one who shows up, knows their numbers cold, and makes the call. The data gives you conviction. The boots on the ground close the deal.

Onwards.