Section 1 — The headline number is the least important number
Every report opens with a big bold AVM value. Resist the temptation to read just that number and stop. The headline is a point estimate; the actual market range is the headline ± the confidence interval.
Look for two things:
- Point estimate — the single number ("$487,500").
- Confidence band — the range ("$447k–$527k, 80% confidence").
A $40,000 band on a $487,500 home means the model is saying "we're 80% sure the real value is somewhere in this $80,000 window". That window is your negotiating range, not the headline number.
A narrow band (±2%) means the comparables look very similar to the target and the model has high confidence. A wide band (±10%) means the comp set is heterogeneous, the photos are missing, or the public-records data is stale — and you should be more cautious.
Section 2 — Comparable sales (do the comps actually compare?)
This is the most important section in any valuation report and the section most reports bury. Here's what to verify:
- Distance. Comps should be within ~0.5 miles in dense urban areas, ~1 mile in suburbs, ~2 miles rural. If comps are 3 miles away, the local market signal is weak.
- Recency. Comps should be within the last 6 months, ideally 3 months. Markets move; a 12-month-old comp in an appreciating market understates value.
- Similarity. Beds, baths, square footage, lot size, year built, condition. A 4/3 single-family house should not be comp'd to a 2/1 condo.
- Adjustment magnitude. Each comp should have line-item adjustments (+/- $X for sqft, +/- $Y for bath, +/- $Z for condition). If total adjustments exceed 15% of sale price, the comp isn't actually similar — discount it.
- Sale type. Was the comp an arm's-length market sale, or a foreclosure / family transfer / cash investor flip? Anything but arm's-length should be flagged or excluded.
In a Twellie report, all eight comps are shown on a map, with the adjustment table beneath each. You can click through to the underlying listings and verify the data yourself.
Section 3 — Photo condition grades
This is the section free AVMs (Zestimate, Redfin Estimate) miss entirely on off-market homes — they don't have current photos.
Modern AI reports grade each visible room on a 1–5 scale (or A–F):
- A / 5 (excellent) — recently renovated, market-ready
- B / 4 (good) — well-maintained, minor cosmetic refresh needed
- C / 3 (fair) — visible wear, deferred maintenance
- D / 2 (poor) — major systems aging, dated finishes throughout
- F / 1 (very poor) — habitability concerns, rehab required
Each grade carries a value adjustment. A "fair" condition on the kitchen typically costs $15–20k off; "poor" overall condition can take $40–80k off the headline AVM.
Two pitfalls:
- Staged photos hide condition. A staged living room can mask a decade-old roof or a buckling floor. Modern AI vision flags staged photos but a human eye still beats it for subtle defects.
- No photos = no adjustment. If the listing has 3 photos, the grade is unreliable. Discount it accordingly.
Section 4 — True cost of ownership
The mortgage payment is the cheapest part of owning a home. A real report shows:
- Principal & interest — the headline mortgage payment
- Property tax — varies wildly by state (0.4% in HI to 2.5% in NJ)
- Insurance — homeowners + flood (if FEMA SFHA) + earthquake (CA)
- HOA / condo fees — monthly + special assessment risk
- Maintenance — rule of thumb 1–2% of home value per year
- Utilities — heating climate matters
A $400k home in a 2.2% property-tax state with a $250 HOA is a ~$3,400/month true cost on a 20% down conventional loan, not the ~$2,300 the mortgage calculator shows. Always compare houses on true monthly cost, not just price.
The report should also show 5-year equity projection — how much you actually build vs how much you pay in tax + interest + insurance that you don't get back. On a 5-year hold, you typically build $30–60k of equity on a $400k home — substantially less than people expect because most of the early mortgage payment is interest.
Section 5 — Risk profile
This section flags physical risks that don't show up in the headline number:
- FEMA flood zone. Special Flood Hazard Area (SFHA) homes require flood insurance (~$700–$2,000/year extra) and are harder to resell.
- Climate risk. Wildfire (Western US), hurricane (Gulf / SE), tornado (Midwest), sea-level rise (coastal). 30-year horizon.
- Environmental. EPA Superfund proximity, lead paint risk (pre-1978 homes), radon zones, prior agricultural land.
- Earthquake. California, Pacific Northwest, parts of the Midwest (New Madrid).
A C-grade flood risk doesn't kill a deal but it means you carry extra insurance and your resale pool is narrower. An A-grade wildfire risk in a high-fire-risk Cal county is increasingly an insurability problem (some insurers won't write at all).
Section 6 — Recommended offer
The synthesis of everything above. A good recommended-offer section shows three numbers:
- Strong offer — the price likely to win without overpaying (typically AVM mid – 1–3%)
- Aggressive offer — opening price for a soft market (AVM mid – 5%)
- Walk-away — the price beyond which the math no longer works (typically AVM upper-band)
Plus the negotiation context:
- Days on market — 60+ days = motivated seller, lean aggressive
- Price cuts — 1+ cuts = lean aggressive
- List price vs AVM — listed above AVM = list-price expectation is high; listed at or below = seller has correct expectations
What to do if the report and the listing disagree
The listing says $499,900. The AVM says $467k mid-band. Now what?
- Listing 5–10% above AVM mid: normal listing pricing strategy. Offer at AVM mid; counter to AVM upper.
- Listing 10–20% above AVM mid: seller has misaligned expectations. Walk away or wait for price cuts.
- Listing at or below AVM mid: the seller is motivated or the property has hidden issues. Ask why, then strong offer.
- Listing 5%+ below AVM upper-band: likely multi-offer scenario. Lean aggressive, escalate if needed.
Read the methodology, then trust the report
A report you don't understand is worse than no report. Spend ten minutes on the methodology page (/methodology) — the math shouldn't be a black box. Then the headline number, the band, and the comps are tools you actually wield, not numbers you guess at.