The Data Layer Nobody Talks About in Site Selection

Most location decisions I review go wrong in the same place: not in the final analysis, but in what got filtered out before anyone started asking the right questions.

The industry has developed an almost religious reverence for data over the past decade. Dashboards, datasets, geospatial overlays — the tools are genuinely impressive. But more data is not the same thing as better decisions. And nowhere is that gap more expensive than in corporate site selection, where a missed variable early in the process can cost a company tens of millions of dollars and six to twelve months they cannot get back.

I have been working on both sides of these transactions for fifteen years. I have sat in the room with companies searching for sites and in the room with economic development organizations trying to close deals. That vantage point has taught me something the data-first consultants rarely admit: the data tells you what, but experience tells you why. Both matter. And when you have only one, you are flying partially blind.

Before vs. After: What Actually Changed

Ten years ago, site selection consultants were largely working off spreadsheets, published incentive guides, and whatever labor data a state ED office handed over in a presentation. The process relied heavily on relationships and reputation. That was not entirely bad — relationships still close deals — but it meant a lot of consequential analysis was happening on incomplete information.

Today, tools like Lightcast give us granular workforce data at the ZIP code level: labor shed radius, wage benchmarks by occupation code, commute patterns, competing employer density. CoStar and Crexi surface real estate comps, vacancy rates, and industrial inventory that would have taken weeks of broker calls to approximate. GIS platforms like ArcGIS Pro let us layer environmental constraints, utility infrastructure, transportation access, and parcel attributes onto a single map and identify fatal flaws in thirty minutes that would have survived all the way to a site visit in the old process.

The difference is speed and precision. The risk is that speed and precision can create a false sense of certainty.

The Multi-Layer Approach That Actually Works

When we run a site evaluation at Hyphen, we work through three distinct layers — and the order matters.

Fatal Flaw Screening: The Unglamorous Work That Saves Projects

The concept I come back to more than any other is the fatal flaw. In every site search, there is at least one factor that, if ignored, will blow up the deal at the worst possible moment — typically during due diligence, after the client has already announced to their board that they found a location.

Fatal flaw screening means using data early to surface the deal-killers before anyone falls in love with a site. I have pulled a site from the shortlist because geospatial analysis showed a tributary creek running through what the surface topography made look like buildable acreage. I have killed a finalist location because the substation serving it had an eighteen-month lead time on transformer capacity that no broker disclosure mentioned and that did not appear in any marketing materials.

Neither of those findings came from a gut feeling. They came from structured, layered analysis run before the proposal phase, not after it. That is the difference between data analytics done well and data analytics used as a retrospective justification for decisions already made.

Raw Data Is Not Actionable Intelligence

Here is the version of this conversation I have had more times than I can count: a company hands me a stack of deliverables from a prior engagement — beautiful charts, detailed tables, a seventy-page report — and asks why they still do not know what to do.

The answer, almost always, is synthesis. The data is all there. What is missing is someone who has seen enough deals close and enough deals fall apart to know which numbers actually drive decisions and which ones look important but do not.

Lightcast will tell you the current labor supply for a given occupation code in a given market. It will not tell you that three of the four largest employers in that market have announced expansions in the next eighteen months that will absorb the available supply before your client breaks ground. You need to know the market to read the data correctly. The model does not know the local context. You do.

This is where the dual-sided experience I carry matters practically. Having worked inside an economic development organization, I know which incentive packages are genuinely negotiable and which numbers are anchored to legislative caps that no relationship can move. Having worked on the corporate side, I know which site characteristics a company's operations team will accept a workaround on and which ones will cause them to walk. That knowledge does not live in a dataset. It lives in pattern recognition built over years of watching deals succeed and fail.

A Note on AI in Site Selection Analytics

I would be doing this topic a disservice if I did not address it directly. AI is changing the front end of our work — we use private-instance tools to accelerate RFP and RFI analysis, reduce first-draft time on research deliverables, and surface initial patterns across large geospatial datasets faster than manual review allows.

What AI cannot do, at least not yet, is replace the interpretive judgment that comes from understanding the stakeholder dynamics on both sides of a deal. It cannot tell you that the incentive package an EDO is offering is structured the way it is because of an internal political constraint, not a financial one. It cannot read the tone of a site visit and understand that the infrastructure upgrade being promised is contingent on a capital budget approval that has not happened yet.

AI is a tool for working faster and with more information. It is not a substitute for knowing what questions to ask or understanding why the answers matter. The consultants who will use it best are the ones who already know the territory.

What This Means for Your Next Location Decision

If you are planning a significant expansion or relocation — manufacturing, logistics, data center, anything with a meaningful capital commitment — the question worth asking your advisory team is not “how much data do you use?” but “how do you decide what to trust and what to investigate further?”

The answer to that question will tell you more about the quality of the analysis you are about to receive than any software stack or dataset list will.

The data is widely available now. The judgment about what to do with it is not.