SF construction costs by neighborhood — what projects actually run

Cost benchmarks built from 12,000+ recent SF DBI permit filings · Updated April 2026

San Francisco's Department of Building Inspection publishes the project value declared on every permit. Aggregating that data across 12,000+ recent filings yields cost benchmarks for typical scopes by neighborhood — useful for setting client expectations, sanity-checking a budget, and figuring out which neighborhoods are running which kinds of jobs.

This guide walks through the methodology and surfaces the most useful benchmarks. The full interactive matrix is built into Permit Pipeline's lead drawer.

Why neighborhood-level cost data matters

"What does a kitchen remodel cost in San Francisco?" is a question with a useless answer because SF is not one market. A kitchen remodel in Pacific Heights runs different numbers than the same scope in the Excelsior. Construction costs are driven by labor pool, material logistics, neighborhood code complexity (historic districts, soft-story retrofit overlays), and what the surrounding market will support.

The DBI permit dataset has both the project value and the neighborhood (neighborhoods_analysis_boundaries) on every row. Bucket by (neighborhood, scope category), compute the median and 75th percentile, and you have a real benchmark.

Methodology

We pull issued and in-review permits from DataSF's i98e-djp9 dataset over a rolling 180-day window. Drop rows with revised_cost equal to zero or missing — about 1% of recent rows, mostly cancelled or denied. Drop outliers above $200M (data-entry errors). Classify each remaining permit into a scope category using a mix of permit_type_definition + work-description keyword matching. Bucket by (neighborhood, scope), compute percentiles.

The resulting matrix has 36 neighborhoods × 21 scope categories. Some cells are dense (Financial District × commercial TI = hundreds of permits, tight benchmark). Others are thin (Seacliff × new construction = fewer than 10 permits, fall back to district or citywide aggregate). Permit Pipeline's lead drawer applies a sample-size threshold and falls back to coarser geos when needed, so the displayed comp is always honest about its source.

The most useful benchmarks — selected scopes

Bathroom remodels (Type 3 & 8 alterations with bath scope)

SF bathroom remodels show the widest geographic spread of any common scope. Median project value across the city sits in the high $30Ks; the 75th percentile sits closer to $80K. Pacific Heights and Marina commonly run double the citywide median — large bathrooms, premium materials, structural complexity in older buildings. Outer Sunset, Excelsior, and Bayview Hunters Point run closer to or below the median.

Kitchen remodels

Kitchens follow a similar pattern to bathrooms but at higher absolute values. Citywide median sits around $50K, with 75th percentile around $130K. Pacific Heights, Russian Hill, and Marina commonly run 2–3× the median for full kitchen remodels with structural changes; outer-district kitchens tend toward straightforward replacements.

Major alterations — whole-home or whole-floor renovations

Type 3 permits ("additions, alterations, or repairs") with project value above $500K are mostly whole-home renovations or whole-floor commercial TIs. Median sits around $1.1M; 75th percentile around $2.5M. Financial District / South Beach, SoMa, and Mission Bay run the highest medians for commercial TIs (full-floor scopes regularly exceed $5M). Russian Hill, Pacific Heights, and Noe Valley dominate the residential end.

New construction (Types 1 & 2)

Most new construction in SF is wood-frame mid-rise residential. Median project value across all SF new-construction permits sits around $1.4M for single-family or small multi-unit; the 75th percentile climbs above $5M when commercial mid-rise is included. Mission Bay, SoMa, and Financial District / South Beach skew the high end. Outer Richmond, Sunset, and West of Twin Peaks dominate single-family new builds in the $800K–$2M range.

Commercial tenant improvements

Commercial TI scopes are concentrated in Financial District / South Beach, SoMa, and Mission Bay — the dense commercial cores. Median TI project value in these areas runs around $750K; 75th percentile clears $2.5M. Outside the commercial cores, TIs are rarer and tend smaller.

Soft-story retrofits

SF's mandatory soft-story program affects multi-unit wood-frame buildings built before 1978 with parking on the ground floor. Retrofit project values cluster between $80K and $250K depending on building size. The Mission, Inner Richmond, Inner Sunset, and the Western Addition have the highest retrofit volumes citywide.

How to use the data without misleading yourself

A few principles:

• Use median, not mean. A handful of $20M permits will skew the mean wildly. Median is robust.
• Watch sample size. A neighborhood × scope cell with fewer than 10 permits is not a benchmark; it's noise. Fall back to district (11 supervisor districts) or citywide.
• Project value is owner-declared. Owners sometimes under-state value to keep fees down. The 75th percentile is more reliable than the median for the upper end of a scope.
• Trust the recent data. SF construction costs moved meaningfully in 2024–2025; older benchmarks misrepresent current pricing.

Where the comp data is wrong (and where it's right)

The comp matrix is built from owner-declared project value, not invoices, contracts, or completed-cost data. That has implications:

• It's right about relative magnitude. A 90th-percentile commercial TI is, in fact, much larger than a median residential remodel. The ranking is real.
• It's right about geographic spread. The cost gap between Pacific Heights and the Excelsior is real and consistent across scope categories.
• It systematically undershoots final costs. Project value is filed at permit time, before change orders, before bidding clears, sometimes before the design is final. Most jobs come in higher.
• It's coarse on scope. "Bathroom remodel" lumps a guest powder room with a primary-bath gut. The percentile spread within a category captures some of that, but if you're scoping precisely, look at the work description.

Why the public dataset is more useful than industry estimating tools

Industry cost estimating products (RSMeans, Craftsman, etc.) are excellent for theoretical bottoms-up estimates — but they don't reflect what owners actually file in SF. The public DBI dataset reflects the actual market: real owners, real properties, real declared values, real distribution. For setting client expectations, the empirical median from the permit feed is harder to argue with than a hypothetical estimator number.

Pairing the comp data with permit type sharpens it further: a Type 3 permit at $500K in Hayes Valley is signaling a different scope than a Type 8 OTC at $500K in the same neighborhood.