5 min read

What is the risk score and how is it calculated?

Augrented's risk score combines open HPD violations, DOB complaints, litigation history, and rent-roll trends. Here's the exact weighting and what moves it.

Updated Jun 18, 2026

What is the risk score and how is it calculated?

Category: Reports & Data
Read time: 7 minutes

The risk score is the first thing you see when you look up a building — a percentage estimate of how likely the building is to have serious or dangerous HPD violations filed in the next 12 months.

It's not a guess. It's the output of a machine learning model trained on 7 years of city inspection data across 169,000+ buildings in NYC and San Francisco.

The number: what it tells you

The score is expressed as a percentage. Lower is safer.

Score range What it means
0–15% Low predicted risk. The building has a strong track record relative to similar properties.
15–35% Moderate risk. Some signals warrant attention — check the specific factors driving the score.
35–50% Elevated risk. The model sees clear patterns of concern in violation history or landlord track record.
50%+ High risk. This building's profile matches properties that historically had frequent serious violations.

The score is also mapped to a risk band displayed as a letter grade on the full premium report.

The score is a probability estimate for the next 12 months — not a permanent label. A building's risk changes as new violations are filed, old ones are resolved, and the landlord's portfolio activity evolves.

What the model considers

The model combines signals from five categories, weighted by how predictive each has proven historically:

1. Violation history (highest weight)

Current and historical HPD violations — the strongest single predictor of future filings:

  • Number and severity of open violations — Open Class C (immediately hazardous) violations carry the most weight
  • Ratio of open to closed — Landlords who resolve violations quickly score better
  • Recurrence patterns — Same violation type filed repeatedly (e.g., heat complaints every winter) signals a systemic issue
  • Violation density — Violations per unit, not just per building, since a 200-unit building with 5 violations is different from a 10-unit building with 5

2. Complaint history

Patterns in complaints filed with HPD, 311, and DOB:

  • Complaint frequency — How often tenants at this building file complaints
  • Complaint types — Serious complaints (gas leaks, no heat) weighted higher than routine ones
  • Seasonal patterns — Winter heat complaints that never get resolved are a stronger signal than a one-time issue

3. Landlord portfolio record

The landlord's track record across all their properties — not just this one:

  • Portfolio-wide violation rate — High violation counts across many buildings indicate a management pattern, not bad luck at one address
  • Cross-building comparisons — If this building is an outlier in a clean portfolio, the problem is likely specific to the building. If it's average for a bad portfolio, the landlord is the issue.
  • Entity registration history — Frequent ownership changes or shell entities can be signals

4. Building characteristics

Structural factors that correlate with maintenance risk:

  • Age — Pre-war buildings have different maintenance profiles than newer construction
  • Building class — Certain classifications correlate with maintenance patterns
  • Size and units — Larger buildings have more systems that can fail

5. Historical trajectory

How the building has changed over time — not just where it is now:

  • 7-year trend — A building that's been getting worse for 5 years is a different risk than one that just had a bad year
  • Recent acceleration — Spikes in the last 12 months carry more weight than old history
  • Period of ownership — Did the current owner inherit a bad building or create one?

Neighborhood percentile

Next to the risk score, the report shows a neighborhood percentile — how this building compares to similar buildings in the same neighborhood:

  • 90th percentile = safer than 90% of comparable nearby buildings
  • 50th percentile = average for the area
  • 10th percentile = among the most at-risk in the neighborhood

The percentile matters more than absolute violation counts. Five open violations in a low-violation neighborhood is a red flag. The same number in a high-violation area may be typical.

Top risk factors (SHAP analysis)

The premium report shows the three factors that most influenced this building's specific score, ranked by weight. This is called SHAP analysis — it doesn't just tell you what the model thinks, but why.

For example, the top factors for one building might be: 1. Landlord portfolio violations — The owner's other properties have high open violation counts 2. Recurring heat complaints — Winter heat complaints have been filed every year for the last 4 years 3. Building age — Pre-war construction with older plumbing and heating systems

For a different building, the top factors might be: 1. No recent complaints — No signal of current tenant issues 2. High closure rate — All historical violations were resolved within reasonable timeframes 3. Clean portfolio — The landlord's other properties also score well

Predicted violation outcome

The model also estimates the probabilities of three outcomes in the next 12 months:

  • No serious violations — The building maintains or improves its current condition
  • Some violations (1–9) — Minor to moderate new filings are likely
  • Many violations (10+) — A significant deterioration is predicted

These probabilities reflect the model's confidence and are displayed alongside the main risk score in premium reports.

What the risk score does not measure

Important to be clear about what this score is not:

  • Not a rent stabilization indicator — Those signals are in a separate section of the report
  • Not a legal determination — The score is a statistical prediction, not building code enforcement
  • Not a guarantee — A low score doesn't guarantee a perfect tenancy; a high score doesn't mean the building is uninhabitable
  • Not neighborhood-specific — The risk score itself is trained on citywide data and normalized against comparable buildings

How often the score updates

The risk score is recalculated whenever new data comes in. HPD violations sync daily, DOB permits weekly, and other datasets on their respective schedules. A building's score can change as new violations are filed or old ones are resolved.

The Renter plan includes building watch alerts that notify you when a building you're watching gets a new violation filing — so you're not relying on a stale score.

Check any NYC or SF address — the free report shows the risk score and neighborhood percentile immediately.