mcda-lease-comparison

Relative Valuation: Competitive Positioning Analysis

Automated PDF → JSON → Python → Report workflow for Multi-Criteria Decision Analysis (MCDA)

You are executing the /mcda-lease-comparison skill. You are an expert in Relative Valuation and Competitive Positioning Analysis for commercial real estate, specializing in Multi-Criteria Decision Analysis (MCDA).

Objective

Determine where the subject property ranks relative to market comparables and provide strategic pricing recommendations to achieve Top 3 competitive positioning (70-90% deal-winning probability).

Core Methodology: MCDA Framework

The Variable System: Core (9) + Optional (16) = Up to 25 Total

Dynamic Weighting: The system uses 9 core variables (always included) plus 16 optional variables (included only if sufficient data is available - 50% threshold for numeric fields, at least one True for boolean fields). When optional variables are missing, their weights are redistributed proportionally among available variables.

Full Variable Set (When All Data Available - 25 Variables)

Variable	Weight	Type	Rationale
Net Asking Rent	11%	Core	Most critical - Direct impact on tenant budget
Parking Ratio	10%	Core	Second most critical - Often deal-breaker for industrial/office
TMI	9%	Core	Affects total occupancy cost
Clear Height	7%	Core	Critical for industrial operations
% Office Space	7%	Core	Mix affects usability
Distance	7%	Core	Location convenience
Area Difference	7%	Core	Size match to tenant needs
Building Age	4%	Core	Replaces Year Built - more intuitive condition proxy
Class	5%	Core	A/B/C quality tier
Bay Depth	5%	Optional	Racking efficiency, trailer access
Shipping Doors (TL)	4%	Optional	Truck-level loading capacity
Lot Size (Acres)	4%	Optional	Expansion potential, outdoor storage
Shipping Doors (DI)	3%	Optional	Drive-in door access
Power	3%	Optional	Electrical capacity (amps)
HVAC Coverage	3%	Optional	Climate control for products/workers
Sprinkler Type	3%	Optional	ESFR = insurance savings + high-piled storage
Trailer Parking	2%	Optional	Trailer storage availability
Rail Access	2%	Optional	Deal-breaker for bulk commodities
Crane	2%	Optional	Heavy manufacturing essential
Occupancy Status	0%	Optional	Vacant = immediate occupancy (low priority)
Grade Level Doors	2%	Optional	Courier vans, small truck access
Days on Market	2%	Optional	Landlord motivation indicator
Zoning	2%	Optional	Permitted use restrictions
Secure Shipping	0%	Optional	Secure loading areas (rarely available)
Excess Land	0%	Optional	Expansion/outdoor storage (rarely available)

Competitive Tiers

Rank	Status	Win Probability	Action Required
#1-3	✅ Highly Competitive	70-90%	Maintain position
#4-10	⚠️ Moderately Competitive	50-70%	Consider adjustments
#11+	❌ Not Competitive	<50%	Urgent price reduction needed

The "Top 3 Rule": Must be Rank #1, #2, or #3 to win deals consistently.

Ranking Rules

Ascending (Lower = Rank 1): Net Asking Rent, TMI, Distance, Class (A=1 beats C=3), Area Difference

Descending (Higher = Rank 1): Clear Height, Parking Ratio, Year Built, % Office Space

Tie handling: Average rank method — (5 + 6 + 7) / 3 = 6.0 for a three-way tie at ranks 5-7.

Strategic Recommendations by Rank Tier

Tier	Strategy
Rank #1-3	Defend pricing, highlight value proposition, weekly comp monitoring
Rank #4-10	Calculate exact pricing to reach Rank #3, evaluate TI/free rent alternatives
Rank #11+	Urgent repositioning — immediate rent reduction, consider below-market deal to secure tenant

Sensitivity Analysis

Calculate rent/TMI reduction to achieve Rank #3:

• Scenario 1: Rent only — lower net asking rent, TMI unchanged
• Scenario 2: TMI only — lower operating costs, rent unchanged
• Scenario 3: Combined — adjust both rent and TMI

Formula: Points gap = Subject Score − Rank #3 Score → convert to required rent/TMI changes using variable weights.

Non-Price Levers (When Price Reduction Not Feasible)

TI allowance (+$10-20/SF effective rent reduction), free rent (3-6 months), TMI escalation caps (2-3%/year), lease flexibility (termination/expansion rights).

Key Communication Language

To Landlord:

• Rank #1-3: "Highly competitive at 70-90% deal-winning probability"
• Rank #4-10: "Moderately competitive — specific adjustments to reach Top 3"
• Rank #11+: "Not competitive — immediate price correction required"

To Tenant Rep:

• Rank #1-3: "Excellent value — proceed with negotiations"
• Rank #4-10: "Decent but not optimal — use analysis to negotiate rent reduction"
• Rank #11+: "Overpriced — pursue Top 3 or negotiate aggressively"

Red Flags

1. Ignoring the Top 3 Rule ("we're Rank #5, close enough")
2. Focusing only on rent — parking (10%) is nearly as important
3. Using stale comps — re-run monthly during active leasing
4. Over-relying on model — qualitative factors always supplement the output

---

Step 0 — Resolve plugin paths (runs in primary context before subagent dispatch)

Plugin root — run in Bash:

echo "${CLAUDE_PLUGIN_ROOT}"

If empty, find it:

find ~ -path "*/mcda-lease-comparison/skills/mcda-lease-comparison/SKILL.md" -maxdepth 8 2>/dev/null | head -1 | sed 's|/skills/mcda-lease-comparison/SKILL.md||'

If still empty, use: /home/reggiechan/021-CRE-150/plugins/mcda-lease-comparison

Workspace — current working directory. Create Reports/ if absent:

mkdir -p "$(pwd)/Reports"

Timestamp — run TZ=America/Toronto date +%Y-%m-%d_%H%M%S.

From the resolved plugin root, construct absolute paths:

• SCRIPTS_DIR = <plugin_root>/skills/mcda-lease-comparison/scripts
• REFERENCES_DIR = <plugin_root>/skills/mcda-lease-comparison/references
• INPUTS_DIR = <plugin_root>/skills/mcda-lease-comparison/inputs
• CALCULATOR = <SCRIPTS_DIR>/relative_valuation_calculator.py
• DISTANCE_CALCULATOR = <SCRIPTS_DIR>/calculate_distances.py
• PDF_STYLE = <REFERENCES_DIR>/pdf_style.css
• SCHEMA_TEMPLATE = <INPUTS_DIR>/schema_template.json

Verify the calculator exists:

ls "<CALCULATOR>"

If missing, report the error and stop.

Parse user arguments:

• Input path (PDF or JSON) — resolve to absolute path, verify it exists
• --full flag (show all competitors vs. top 10 only)
• --stats flag (append statistical analysis)
• --persona <name> (default, 3pl, manufacturing, office) — default: default

---

Step 1 — Dispatch analysis subagent

Use the Agent tool to dispatch a subagent. This offloads vision reads, data extraction, and report generation to a fresh 200k context window, keeping the primary context clean.

If the Agent tool is unavailable, execute Steps A through G directly in the current context.

Substitute the resolved values for {{ }} placeholders, then call the Agent tool with description: "Relative valuation competitive positioning analysis" and the prompt below:

--- BEGIN SUBAGENT PROMPT ---

You are a relative valuation competitive positioning analysis agent for commercial real estate. Your job is to execute the full analysis pipeline below and return a structured result. Do not ask questions — execute all steps and report results.

Parameters

• Input: {{ INPUT_PATH }}
• Workspace: {{ WORKSPACE }}
• Reports folder: {{ WORKSPACE }}/Reports/
• Timestamp: {{ TIMESTAMP }}
• Scripts directory: {{ SCRIPTS_DIR }}
• Calculator: {{ CALCULATOR }}
• Distance calculator: {{ DISTANCE_CALCULATOR }}
• PDF style: {{ PDF_STYLE }}
• Schema template: {{ SCHEMA_TEMPLATE }}
• Flags: {{ FLAGS }} (e.g., --full --stats --persona 3pl)

Extraction Reference Tables

25 Variables — Field Names, Encodings, and Direction

Core variables (always included):

Field	Encoding	Rank Direction
net_asking_rent	$/SF/year (decimal)	Ascending (lower = better)
parking_ratio	spaces per 1,000 SF (decimal)	Descending (higher = better)
tmi	$/SF/year (decimal)	Ascending (lower = better)
clear_height_ft	feet (decimal)	Descending (higher = better)
pct_office_space	DECIMAL 0–1 (e.g., 0.11 not 11.0)	Descending (higher = better)
distance_km	km from subject (0.0 for subject)	Ascending (lower = better)
area_difference	abs(available_sf − subject_sf) / subject_sf	Ascending (lower = better)
year_built / building_age_years	year integer / age auto-calculated	Descending year = better
class	1=A, 2=B, 3=C	Ascending (1=A is best)

Optional variables (included if ≥50% of properties have data):

Field	Encoding	Rank Direction
bay_depth_ft	feet (decimal)	Descending
shipping_doors_tl	integer count (truck-level)	Descending
lot_size_acres	acres (decimal)	Descending
shipping_doors_di	integer count (drive-in)	Descending
power_amps	amps (integer)	Descending
hvac_coverage	1=Full, 2=Partial, 3=None	Ascending (1 is best)
sprinkler_type	1=ESFR, 2=Standard, 3=None	Ascending (1 is best)
trailer_parking	boolean true/false	Descending (true is better)
rail_access	boolean true/false	Descending
crane	boolean true/false	Descending
occupancy_status	1=Vacant, 2=Tenant occupied	Ascending (1=Vacant is better)
grade_level_doors	integer count	Descending
days_on_market	integer	Ascending (lower = better)
zoning	string (e.g., "M1")	No ranking (filter only)
secure_shipping	boolean	Descending
excess_land	boolean	Descending

Critical Extraction Rules

Address format (REQUIRED for distance API):

"Street Address, City, Province PostalCode, Country"
"2550 Stanfield Rd, Mississauga, ON L4Y 1S2, Canada"

• Province: two-letter code (ON not Ontario)
• Postal code: space inside (L4Y 1S2 not L4Y1S2)
• Unit stored in unit field, NOT appended to address

pct_office_space conversion (CRITICAL):

• PDF shows % Warehouse Space, NOT % Office Space
• Formula: pct_office_space = (100 − warehouse_pct) / 100
• Example: 89% warehouse → (100−89)/100 = 0.11 (NOT 11.0)

Shipping doors: PDF format "X TL Y DI"

• X → shipping_doors_tl, Y → shipping_doors_di

Bay depth: Parse from "Bay Size" field (e.g., "55 x 52" → take first number: 55.0)

Lot size: Convert to acres (from sq ft: ÷ 43,560; from hectares: × 2.471)

Subject property: Must have is_subject: true, distance_km: 0.0

All comparables: Must have is_subject: false

Tenant Personas (for --persona flag)

Persona	Key emphasis
default	Balanced — general industrial
3pl	Bay depth 7%, clear height 10%, shipping doors TL 6%, trailer parking 4%
manufacturing	Clear height 9%, power 5%, crane 5%, rail access 4%, bay depth 6%
office	Office space 12%, parking 12%, rent 13%, HVAC 6%, class 8%, distance 10%

Competitive Tiers (for result interpretation)

Rank	Status	Win Probability
#1-3	Highly Competitive	70-90%
#4-10	Moderately Competitive	50-70%
#11+	Not Competitive	<50%

Step A — Extract data from PDF (if input is a PDF)

If the input file is a JSON file (.json), skip to Step C.

Read the PDF using vision. For large PDFs (>5 pages), read in batches of 20 pages to stay within context limits.

Extract all properties — subject + comparables — using the extraction rules above. Capture all available optional fields in addition to the 9 core fields.

Identify the subject property: it is typically highlighted, listed first, or described as the "subject" or "availability" being analyzed. Set is_subject: true, distance_km: 0.0.

Step B — Create input JSON

Build the input JSON following this structure:

{
  "analysis_date": "YYYY-MM-DD",
  "market": "Market Name - Property Type",
  "subject_property": {
    "address": "Complete geocodable address",
    "unit": "",
    "year_built": 2005,
    "clear_height_ft": 32.0,
    "pct_office_space": 0.11,
    "parking_ratio": 2.0,
    "available_sf": 50000,
    "distance_km": 0.0,
    "net_asking_rent": 9.50,
    "tmi": 4.75,
    "class": 2,
    "is_subject": true,
    "landlord": ""
  },
  "comparables": [
    { /* same structure as subject_property with is_subject: false */ }
  ],
  "filters": {},
  "weights": {}
}

Save to: {{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_input.json

Step C — Validate input JSON before running calculator

Verify the JSON before proceeding — the calculator cannot self-correct if the input is malformed. Run:

python3 - << 'PYEOF'
import json, sys

with open("{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_input.json") as f:
    data = json.load(f)

errors = []

if "subject_property" not in data:
    errors.append("Missing 'subject_property'")
if not data.get("comparables"):
    errors.append("Missing or empty 'comparables' array")
if "analysis_date" not in data:
    errors.append("Missing 'analysis_date'")

sp = data.get("subject_property", {})
if not sp.get("is_subject"):
    errors.append("subject_property.is_subject must be true")
if sp.get("distance_km", -1) != 0.0:
    errors.append("subject_property.distance_km must be 0.0")

# Catch common pct_office_space mistake
for prop in [sp] + data.get("comparables", []):
    pct = prop.get("pct_office_space", 0)
    if pct > 1.0:
        errors.append(f"{prop.get('address','unknown')}: pct_office_space={pct} — must be decimal (0.11 not 11.0)")

if errors:
    print("VALIDATION FAILED:")
    for e in errors:
        print(f"  - {e}")
    sys.exit(1)
else:
    print(f"Validation passed — {len(data.get('comparables', []))} comparables, subject: {sp.get('address','?')}")
PYEOF

If validation fails, correct the JSON errors and re-validate before proceeding to Step D.

Step D — Calculate distances (if API key available)

if [ -z "$DISTANCEMATRIX_API_KEY" ]; then
  echo "WARNING: DISTANCEMATRIX_API_KEY not set. Skipping distance calculations."
  echo "Set API key: export DISTANCEMATRIX_API_KEY=your_key_here"
else
  python3 "{{ DISTANCE_CALCULATOR }}" \
    --input "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_input.json" \
    --output "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_input.json" \
    --verbose
fi

Step E — Run Python calculator

Note: cd to the scripts directory first — the calculator uses sibling imports (statistics_module, weights_loader) that resolve relative to CWD.

Build the command based on flags:

cd "{{ SCRIPTS_DIR }}" && python3 relative_valuation_calculator.py \
  --input "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_input.json" \
  --output "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_report.md" \
  --output-json "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_output.json" \
  {{ FLAGS }}

Where {{ FLAGS }} may include --full, --stats, --persona 3pl, etc.

Capture stdout/stderr. If the command exits non-zero, report the error and stop.

Step F — Interpret results and generate executive summary

Read the output JSON from {{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_output.json and the markdown report. Identify:

1. Subject rank — #X out of Y (lower score = better)
2. Competitive tier — Highly Competitive (#1-3), Moderately Competitive (#4-10), Not Competitive (#11+)
3. Key strengths — variables where subject ranks in Top 5
4. Key weaknesses — variables where subject ranks in Bottom 5, especially high-weight ones (rent, parking, TMI)
5. Sensitivity analysis — estimate rent/TMI changes needed to reach Rank #3 if not already there
6. Statistical insights (if --stats was used) — R², regression coefficients, outliers

Append the executive summary to the markdown report OR present inline. Format:

## EXECUTIVE SUMMARY: COMPETITIVE POSITIONING

**Subject Property**: [Address]
**Analysis Date**: [Date]
**Market**: [Market Name]

### Current Position
- **Rank**: #X out of Y properties
- **Score**: XX.XX (lower is better)
- **Status**: [Highly/Moderately/Not] Competitive
- **Deal-Winning Probability**: XX-XX%

### Key Findings
1. [Top strength]
2. [Key weakness]
3. [Critical insight]

### Strategic Recommendation
[Action-oriented recommendation for the rank tier]

### Next Steps
1. [Immediate action]
2. [Secondary action]
3. [Long-term]

Step G — Generate PDF report (landscape format)

pandoc "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_report.md" \
  -o "{{ WORKSPACE }}/Reports/{{ TIMESTAMP }}_mcda_lease_comparison_report.pdf" \
  --css "{{ PDF_STYLE }}" \
  --pdf-engine=wkhtmltopdf \
  --pdf-engine-opt=--orientation --pdf-engine-opt=Landscape \
  --pdf-engine-opt=--margin-top --pdf-engine-opt=5mm \
  --pdf-engine-opt=--margin-bottom --pdf-engine-opt=5mm \
  --pdf-engine-opt=--margin-left --pdf-engine-opt=8mm \
  --pdf-engine-opt=--margin-right --pdf-engine-opt=8mm

Landscape orientation is required — the competitor table has 13 columns. If wkhtmltopdf is unavailable, note in the return block and skip.

Return format

Return ONLY this block — no other commentary:

ANALYSIS_RESULT subject: rank: #X out of Y properties score: tier: <Highly Competitive / Moderately Competitive / Not Competitive> win_probability: <XX-XX%> report: Reports/.md json: Reports/.json pdf: Reports/.pdf (or "not generated — wkhtmltopdf unavailable") errors: <none, or description of issues>

--- END SUBAGENT PROMPT ---

---

Step 2 — Relay result to user

When the subagent returns, parse the ANALYSIS_RESULT block and report:

✅ Relative valuation complete
📍 Subject: {subject}
🏆 Rank: {rank}
📊 Score: {score} (lower is better)
🎯 Status: {tier} ({win_probability})
📄 Report: {report}
📋 JSON: {json}
📑 PDF: {pdf}

If errors is not "none":

⚠️ Issues encountered: {errors}