Reverse-Engineering 6,600 Property Sales to Build a Lead Scoring System
We're not going to pretend we have proven conversion numbers or a huge client list. We don't. What we do have is 6,600+ closed residential sales from two Georgia counties, broken down and analyzed to figure out what actually matters when someone sells.
This is how we built our lead scoring model from the raw, messy data.
Why Most Lead Generation Fails
We've covered why most lead generation fails in a previous post. The short version: companies optimize for volume because it's easier to sell big numbers than quality filtering.
They come up with a basic schema, hit go, and never look back. Minimal effort for change so long as they scale it enough for profitability. They drown you in quantity and never worry about quality.
We wanted to build something different. Not more leads, better leads. To do that, we needed to understand what actually predicts a sale.
What We Analyzed
Every residential sale in Richmond and Columbia counties over the past few years. Every closed transaction.
We broke down:
Property characteristics: equity, ownership duration, improvements
Owner profiles: local vs. out-of-area, single property vs. multiple
Market context: sale price vs. assessment, time on market
Transaction patterns: seasonal trends, cash vs. financed
There are a lot of impressive sounding correlations, and I could probably woo someone with fancy graphs and charts, but those won’t close deals. We were looking for patterns that repeat in success.
We didn't look at web scrapes, social media signals, or predictive AI models. Just public records and MLS data for closed transactions.
Vacant Properties Are the Strongest Signal. 5% of all sales in our dataset involved vacant properties.
That doesn't sound like much until you compare it to how rare vacant properties actually are in the housing stock.
But here's what matters: certain agents specialize in vacant properties.
We found an agent who specialized in vacant properties, with some closing 5x the market average. She had other characteristics that shone in her sales that were intriguing, it showed her niche and how she excelled there.
She's not taking whatever walks through the door. She's built a model around vacant property acquisition. She knows how to find motivated sellers, structure offers, and close deals other agents skip.
That’s the kind of work we aim to support. Driven agents who know their market. We want to help them go a step further.
The Scoring Model
We built a 100-point system based on what showed up in recently closed sales. We’re analyzing long term data as well, but markets change, so this study was only 2024-2025 for relevancy.
A 75-point lead isn't guaranteed to sell. It means properties with those combinations showed up repeatedly in our closed sales. That's the difference between data and guessing. This may sound simple, but it was quite an undertaking. We’ll have to manually check the data for changing market conditions as we continue to ensure the validity of this model.
Local Market Context
National lead platforms use the same scoring everywhere. But our Georgia counties have a median sale price of $220k against a median assessed value of $95k. This suggests either rapid appreciation, lagging assessments, or both. Either way, it’s not a distressed market where sellers are desperate.
We see:
Properties selling at well above their assessed value.
High cash transaction rates
Significant out-of-area ownership
Sales concentrated in specific property types
Generic national scoring doesn't account for this. Our model was built specifically for markets like this one.
What We Don't Do
We don't predict the future. Anyone claiming they can is lying.
We don't guarantee conversion rates. Market conditions shift, agent skill varies, and seller motivation isn't always visible in public data.
We don't automate everything. Human review matters when data conflicts or edge cases appear.
What we do: identify properties that share characteristics with deals that already closed, score them by signal strength, and match them to agents who specialize in those transaction types.
We don’t believe in a one-size-fits-all methodology for our country’s real estate market. We’re focusing our efforts on the CSRA first. Here's what we have:
A scoring model built from actual closed sales
A few hundred high-scoring leads (60+ points) ready to go
Target agents identified who specialize in the property types we score highest
A hypothesis that these leads convert better than generic lists
What we're doing now
Calling agents and testing whether our leads interest them.
If the agent I mentioned looks at our vacant property list and says "send them," we've validated the approach. If she says "already have those" or "not interested," the model needs work. We’ll continue to refine it.
Analysis is worthless if the end user doesn't want the output.
Why This Is Different
Most lead companies optimize for volume because volume sells. Big numbers look good.
We optimize for signal. We'd rather deliver 100 leads with 7-8 conversions than 1,000 leads with 5 conversions.
The difference:
We started with what sold, not what sounds good
We built market-specific scoring, not generic formulas
We match lead types to specialist agents, not broadcast lists
We validate before we scale
This is the same discipline we used in intelligence work. Understand the environment. Test assumptions. Verify before committing. Don't mistake activity for results.
What's Next
We're in controlled testing. Calling agents who match our lead profiles, seeing if our analysis translates to interest.
If you specialize in:
Vacant property acquisitions
Out-of-area or inherited properties
Distressed or motivated sellers
Markets with cash transactions
We might have what you need.
If you're frustrated with generic lead lists and low conversion, we can show you a different approach.
Reach out to us at contact@crowseyellc.com or connect on LinkedIn. We'll look at your market and specialization, see if our data matches what you actually close.
—
Chris, Founder
Crow's Eye LLC