ai technology
13 min read

Instructions Engineering for Real Estate: How to Write Better Property Search Criteria

The quality of your AI-filtered property search depends entirely on the quality of your instructions. A well-written instructions eliminates 95% of irrelevant listings. A poorly written one either lets too many through or filters too aggressively.

This guide teaches you how to write effective instructions for AI property filtering.

The Anatomy of a Good Instructions

Every effective property search instructions has four components:

1. Hard Criteria (Must-Have)

These are absolute deal-breakers. If a listing fails any of these, it's rejected:

Reject if ANY of these are true:
- Price per m² above €9,000
- Less than 2 bedrooms
- Ground floor without private outdoor space
- DPE rating F or G

2. Calculations

Derived metrics that require computation:

Calculate:
- Price per m² = price ÷ surface area
- Monthly cash flow = (bedrooms × €550) - PMT(3.5%/12, 300, price × 1.08) - charges - (rent × 15%)
- Renovation estimate = €1,200/m² if listing mentions 'to renovate'

3. Soft Criteria (Preferences)

Criteria that improve a listing's attractiveness but aren't deal-breakers:

Strongly prefer:
- South or west facing
- Above 3rd floor
- Recently renovated
- Within 5 minutes of metro

4. Exceptions

Overrides that catch exceptional value:

Also forward ANY listing where:
- Price per m² is 20%+ below neighborhood average
- Cash flow exceeds €300/month regardless of other criteria

Template: First-Time Buyer

Forward apartments that meet ALL of:
- Located in [neighborhoods]
- Price under €[max]
- At least [X] rooms / [Y] bedrooms
- Not ground floor
- Has elevator if above 3rd floor
- Price per m² below €[amount]

Prefer:
- South/west facing
- Balcony or terrace
- Quiet street
- Good DPE (A-D)

Also forward if price per m² is 15%+ below average for the area.

Template: Investor (Cash Flow)

Forward if monthly cash flow is positive:
- Revenue: (total rooms - 1) × €[rent per room]
- Loan: PMT([rate]%/12, [years]×12, price × [multiplier])
- Expenses: charges + (revenue × 15%)
- Cash flow = revenue - |loan| - expenses

Also forward if gross yield > [X]% (annual rent ÷ price × 100).

Reject:
- DPE F or G (unless price/m² < €[threshold])
- Buildings with 'travaux importants' planned

Template: Lifestyle Buyer

Forward if ALL criteria match:
- [Neighborhoods or areas]
- [Min rooms] rooms, [min surface]m²+
- Natural light: south/southwest facing, or above 4th floor
- Quiet: no boulevards, no main roads, ideally mention of 'calme'
- Outdoor space: balcony ≥ 5m² or terrace or garden
- Good building condition: no mention of 'travaux' or major works

Bonus (forward even if one soft criterion is missed):
- Recently renovated
- Period building with character features
- Views

Common Instructions Mistakes

Too Vague

❌ "Find me a nice apartment in Paris" ✅ "Forward 3-room apartments in Paris 10th-12th, price/m² under €8,500, not ground floor, quiet street"

Too Strict

❌ Listing 15 hard criteria = almost nothing passes ✅ Use 4-6 hard criteria + soft preferences + exceptions

Missing Calculations

❌ "Good investment property" ✅ "Monthly cash flow positive assuming [specific formula with rates and assumptions]"

No Exceptions

❌ Rigid criteria only ✅ Add "Also forward if price/m² is 20% below average" — catches unexpected value

Iterating Your Instructions

Your first instructions won't be perfect. Plan to iterate:

  1. Week 1: Run your initial instructions. Review matches AND rejections.
  2. Adjust: Too many matches? Add criteria. Too few? Relax criteria or add exceptions.
  3. Week 2: Run adjusted instructions. Review again.
  4. Stabilize: After 2-3 iterations, your instructions should be delivering the right balance.

The goal: 3-8 matched listings per day in an active market.


Write better instructions, find better properties. Try FilterWise free — AI-powered filtering that understands natural language criteria.

Ready to filter smarter?

Start using AI to filter real estate listings — free, no credit card required.