Ten Homes In. Two Worth a Closer Look: A Worked Example of the Appreciation Screen
Here is a problem every buyer with a shortlist knows: ten homes, similar prices, similar photos, similar ZIP codes — and no way to tell which ones deserve your next three weekends. Below is what that problem looks like after the appreciation screen: ten real scored homes from one metro, anonymized, with the diligence queue that falls out of the read.
What is real here, and what is framing
Ten homes, screened
| Home | Type | Price | Within-market tier | Measured edge vs its market | Confidence | Diligence priority |
|---|---|---|---|---|---|---|
| Home A | Single family | $450,000 | Top fifth | +9.0 pp/yr [+7.9, +10.2] | Medium | Front of queue |
| Home B | Single family | $604,800 | Top fifth | +9.0 pp/yr [+7.9, +10.2] | Medium | Front of queue |
| Home C | Single family | $519,000 | Upper-middle | +0.5 pp/yr [+0.1, +1.1] | Medium | Standard diligence |
| Home D | Single family | $612,000 | Upper-middle | +0.5 pp/yr [+0.1, +1.1] | Medium | Standard diligence |
| Home E | Single family | $549,000 | Middle | ≈ tracks its market | Medium | Standard diligence |
| Home F | Single family | $460,082 | Middle | ≈ tracks its market | Medium | Standard diligence |
| Home G | Single family | $595,000 | Lower-middle | −0.7 pp/yr [−1.0, −0.4] | Medium | Back of queue |
| Home H | Single family | $455,000 | Lower-middle | −0.7 pp/yr [−1.0, −0.4] | Medium | Back of queue |
| Home I | Single family | $555,000 | Bottom fifth | −2.6 pp/yr [−3.5, −2.0] | Medium | Back of queue |
| Home J | Condo | $569,500 | Middle | ≈ tracks its market | Low | Data review first |
Real model output (production scoring, July 2026 calibration), one metro, homes anonymized. “Measured edge” is the historical average by which homes scored in that tier out- or under-performed their own market, in percentage points per year, with the range for the group average in brackets. Group history, not a per-home prediction.
On price alone, these ten are nearly interchangeable — a $455k home sits in the bottom tier while a $450k home sits in the top one. The screen separates them on a different axis: how each home is positioned to appreciate within its own market, how much measured history stands behind that position, and how confident the read is.
What falls out: a diligence queue
- Two homes move to the front. Homes A and B sit in the top fifth of their market — the tier whose homes historically out-appreciated their market average by the widest measured margin. They earn the deep look first: full inspection, comp scrutiny, negotiation planning.
- One home needs its data checked before it can compete. Home J's read is flagged low-confidence — outside trained coverage — so its middle tier is not comparable to the others until a human reviews the evidence. The flag is the product working, not failing.
- Seven homes wait. Not rejected — deprioritized. The middle tiers track their market or carry slight measured drag; if the front-runners fall through on diligence, the queue is still ordered.
The progression, using only what the data supports: ten homes screened → two prioritized for deeper diligence → one routed to data review → seven deprioritized with reasons attached. Which of the two front-runners becomes the offer is decided by diligence — condition, negotiation, and the buyer's own criteria — not by the screen.
The boundary, plainly
Run your own shortlist
The screen reads one address at a time, so start with the home you keep coming back to and work down your list — each read returns the within-market position, measured edge, confidence, and drivers behind it. Start with how to evaluate a property for appreciation for the method, or run an address through the property appreciation analysis to see it on a real home.
Frequently asked questions
Is this a real example?
The scores are real: every tier, measured edge, and confidence flag on this page is genuine output from our production scoring pipeline for ten actual homes in the Denver metro, anonymized as Home A through Home J. The shortlist scenario — a buyer comparing them — is illustrative framing. No customer, purchase, or realized outcome is claimed.
Why doesn’t the screen rank Home A above Home B?
Because the honest read stops at the tier. Both sit in the top fifth of their market, and the measured edge is a group figure — the model does not manufacture false precision between two homes it reads as equivalently positioned. Separating them is diligence work: price negotiation, condition, block-level factors, and your own priorities.
What does the low-confidence flag on Home J actually change?
It moves the home to “data review first”: the property sits outside the model’s trained coverage, so its read should not be compared against the others until a human has checked the comparable evidence. A flagged read is a prompt for more diligence — never a hidden penalty, and never silently blended into the ranking.
Run a real address
See how an individual home ranks for appreciation within its own market.
Try the analysis