How to Find Leads by Square Footage: An AI-Powered GTM Workflow
In B2B sales, finding the right prospect is hard. But what if your best prospects are defined by a physical characteristic? A client of mine sells commercial lighting installs. Their best prospects are massive warehouses: the bigger the square footage, the bigger the deal.
The problem? "Square footage" isn't a filter in any lead database.
I couldn't find a tutorial on this anywhere, so I built the workflow from scratch. This is a complex build, but it’s incredibly powerful. It uses a combination of local search, automation, and AI image analysis to find and qualify prospects based on their physical size.
Here is the high-level overview of how you can find leads by square footage.
The 7-Step Workflow to Find Leads by Square Footage
For a complete, in-depth guide on how to build this entire automation, watch the full video walkthrough on YouTube.
1. Find Companies with Google Maps
The first step is to generate a list of potential businesses. You can't just buy a list of "warehouses," so you need to find them based on location. Using a tool like Clay (specifically its "Find Local Businesses" feature) or the Serper API, you can run searches like "100 warehouses in St. Louis." This gives you a starting list of company names and, most importantly, their addresses.
2. Convert Addresses to Coordinates
A street address isn't useful for an image. You need its precise location on Earth. Using an automation platform like n8n, create a workflow that takes the company address from your Clay table and sends it to the Google Geocoding API. This API converts the text address into exact latitude and longitude coordinates.
3. Get a Satellite Image of the Building
Now for the visual part. Feed those coordinates into the Google Static Map API. You’ll also specify a zoom level, set the map type to "satellite," and define an image size. The API then returns a raw satellite image file of that exact location, showing the building's roof.
4. Ask GPT-4 to Analyze the Image
This is where the magic happens. How can an AI know the scale of the image? You have to give it a reference. The image file is sent to a GPT-4 Vision model with a specific prompt. You instruct the AI:
"Find a reference object in this image: something with a standard size, like a car (~4.5 meters) or a road lane."
5. Create a "Meters-per-Pixel" Ratio
The AI follows the prompt, finds a car in the parking lot or on a nearby street, and measures its size in pixels. Since it knows a car is approximately 4.5 meters long, it creates a "meters-per-pixel" ratio. For example, if the car is 10 pixels long, it determines the scale is 0.45 meters per pixel.
6. Calculate the Estimated Square Footage
Once the AI has its scale, it measures the pixel area of the building's roof. It then applies the "meters-per-pixel" ratio to that pixel area, converting it into a final, real-world "Estimated Area (sqm)."
7. Send the Data Back to Clay
The workflow doesn't just stop. It sends all this new, high-value data right back into your Clay table. Your sheet is now enriched with:
Estimated Square Footage
The "Reference Object" used (e.g., "car")
A Confidence Score on the calculation
You now have a prospect list that you can sort by square footage, allowing you to prioritize outreach to the largest, most valuable targets first.
Frequently Asked Questions (FAQ)
What tools are required for this workflow?
This build uses Clay for data organization, n8n for automation, the Google Geocoding API and Google Static Map API for map data, and OpenAI's GPT-4 for the image analysis. You can also use Serper API for the initial business search.
How accurate is the square footage calculation?
The accuracy depends on the quality of the satellite image and the clarity of the reference object. Using a car or road lane provides a very reliable reference, making the estimates surprisingly accurate for sales qualification purposes. The AI also provides a confidence score to help you judge the quality of the result.
Is this workflow difficult to set up?
This is an advanced workflow that requires familiarity with APIs, automation tools like n8n, and prompt engineering. For a step-by-step guide on how to build it from scratch, please refer to the full video walkthrough.