# Comps Adjustments

> Support comparable selection, the adjustment grid, and reconciliation reasoning — organizing appraiser-supplied market data and stress-testing the logic. Use when the user says "work on comps", "build the grid", or supplies MLS/sale data for comparables.

- Skill: `bedardandy/comps-adjustments` (Agent Skill)
- Install (CLI): `npx skillmds@latest add bedardandy/comps-adjustments`
- Raw SKILL.md: https://api.skillmd.com/api/skills/bedardandy/comps-adjustments/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: bedardandy (https://skillmd.com/u/bedardandy)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/bedardandy/comps-adjustments

---


# Comps & Adjustments

**Boundary:** the appraiser selects the comps and concludes the adjustments. This skill
organizes the data, checks the logic, and drafts the support narrative. It never invents sale
data and never proposes a value conclusion on its own.

## Steps

1. **Define the search frame with the appraiser**: neighborhood boundaries, competitive market
   area, date range, property-type filters. Record the definition and the search criteria in
   `comps.md` (UAD 3.6 asks for more explicit market-area rationale than legacy forms).
2. **Ingest candidate sales** the appraiser supplies (MLS sheets, data exports, dictated
   "comp note" entries from the site visit). Normalize each into a consistent record: address,
   sale date/price, concessions, financing, DOM, GLA, beds/baths, lot, condition/quality (C/Q),
   site influences (waterfront, road class, view), verification source.
3. **Screen and rank**: similarity to the subject on the elements of comparison; flag
   non-arm's-length signals (family transfer, estate, short DOM + big discount, concession-heavy).
   Present the ranked slate — appraiser picks.
4. **Build the adjustment grid** with the appraiser's adjustments. For each adjustment, capture
   *support*: paired-sale references, cost-basis reasoning, market-extraction notes, or sensitivity
   analysis. UAD 3.6 and reviewers increasingly expect stated support — "appraiser's experience"
   alone gets flagged.
5. **Stress-test** (report, don't fix silently):
   - Net/gross adjustment magnitudes; any single adjustment dominating value
   - Adjusted range spread; does the reconciled value fall inside the adjusted range?
   - Consistency: same feature adjusted the same $/unit across comps; subject's C/Q used
     consistently with the site-visit worksheet
   - Time/market-conditions adjustment vs. the market-analysis trend (if the 1004MC-style data
     says increasing market, a $0 time adjustment on a 9-month-old comp needs explanation)
   - Bracketing: is the subject bracketed on GLA, condition, and sale price? If not, draft the
     explanation ask
   - Fair-housing screen: no references to protected classes or demographic composition in any
     neighborhood/market language
   - Guideline-vs-rule discipline: comp age, distance, and net/gross adjustment thresholds are
     *guidelines* requiring commentary, not bright-line rules — never reject or accept a sale by
     citing a threshold as if it were a rule; state the substantive comparability reason
6. **Draft reconciliation support**: given the appraiser's stated weighting and conclusion,
   draft the narrative explaining why (proximity of comp X, least adjusted, most similar…).
   State-module check: apply local factors (e.g., Maine: waterfront frontage type, seasonal-road
   access, leased vs. owned land, current-use enrollment on comps' land).

## Output

End with: the grid, the stress-test findings (each marked resolved / needs-appraiser-input),
drafted support narrative, and next skill (`report-assembly`).