How Do I Estimate Developer ROI on a Luxury Teardown Rebuild?

Ranking measured 8 September 2026 by varying one input at a time against the active SF Peninsula listings AddressIntel can price, using the same proforma the site runs on each property page. It describes how much the model moves, not how much the market moves. We publish the order and not multipliers on purpose: the count of clearing lots is small enough that a ratio would move by half on a single lot trading. Cost figures are benchmarks, not measurements.

First, The Uncomfortable Part: Your Assumptions Decide The Answer

Before the method, a warning about how to use it. We run this model against every active SF Peninsula listing we can price. Vary one input at a time across a defensible range, hold the rest at the model’s defaults, and watch how far the count of lots clearing a 15% return moves:

The market is the same market in every row. Only the assumption moved. The resale figure outweighs hard cost by a wide margin, which is the reverse of where most underwriting attention goes, and it is the reason this page no longer tells you what share of the market pencils. Carry and the return bar are both minor, and close enough that which of them comes third depends on whether you rank by the ratio or by the absolute change in the count, so we do not order them against each other.

We could publish that share. We are choosing not to, because it would rest on the one input we cannot yet defend. The model prices a rebuild against size-matched comps drawn from recent sales, and those sales are overwhelmingly existing houses. A newly built home may sell for more per square foot than the older stock beside it, and if it does, every figure the model produces is too low. We have tried twice to measure that premium from our own data and failed both times: matching sold homes to their year built gives a sample of 47 across eleven cities, whose apparent 1.2x premium falls to 1.08x once city and size are held constant, on cells of two to five sales each. Matching new-residence building permits to later sales gives eleven matches, most of which turn out to be land sold with permits attached rather than finished houses, priced against the square footage of the cottage that was about to be demolished.

So the honest position is that we can tell you how the model behaves and we cannot yet tell you where this market sits. A 25–40% return remains the band developers underwrite toward. It is a target, and on this page it is no longer offered as anything else.

The practical consequence: the arithmetic below is the easy part. Any competent developer can run an RLV model in a spreadsheet. What decides whether the answer is worth anything is the resale number you feed it, which is why the rest of this page spends more time on comps than on cost, and why the product is about the search.

The “Residual Land Value” Model

In high-barrier markets like Silicon Valley and the San Mateo Peninsula, developers rarely price a teardown based on the value of the existing structure. Instead, they use the Residual Land Value (RLV) model.

The RLV model works backward. You determine the maximum price you can pay for a piece of land by taking the projected value of the finished spec home and subtracting all costs required to build it, including your profit margin. Once you know your costs and target profit, ROI is simply profit divided by total project cost.

1. Projecting the Finished Home Value (ARV)

The After Repair Value (ARV) or “Finished Home Value” is the anchor of your entire equation. To calculate this in cities like Atherton or Menlo Park, you must pull very recent, highly localized comps for new construction.

Look for homes sold in the last 6 months within a 1-mile radius that feature similar square footage, lot size, and luxury finishes. Two cautions, both learned the hard way on our own model. Match on size as well as location, because smaller houses sell for more per square foot than larger ones and a comp set drawn from the existing small stock will misprice a large rebuild. And check that your comps sit in the same submarket, not merely the same city or zip: 94025 spans Belle Haven and west Menlo Park, whose medians differ by roughly 50%, and 94303 straddles East Palo Alto and Palo Alto. Our own screener priced a Belle Haven parcel off west Menlo Park sales and produced a return that was positive on paper and negative on its own block.

Be equally careful with a headline per-square-foot number for new construction. Figures in the $2,000–$3,500 range circulate for prime Peninsula zip codes, and we are not able to reproduce them from our own sales: new builds we can identify in the eleven Peninsula cities we price come in at a median nearer $1,600 per square foot, on a sample small enough that we would not publish it as a benchmark either. Use the comps, not the aphorism.

2. Estimating Hard & Soft Costs

This is where most novice developers fail. Building a luxury spec home requires factoring in highly variable costs:

• Hard Costs: The physical materials and labor. For a builder pricing a Peninsula spec home, hard costs land between $400 and $700 per square foot. Note that $800 to $1,200+ is a retail quote, what a general contractor charges a homeowner for a one-off custom build, and it is not what a spec builder actually pays. Bay Area operators we have compared notes with report roughly $260–300/sq ft above land on prefab, and about $617/sq ft all-in on a 3,000 sq ft stick build under a fixed-price contract with owner-grade finishes (closer to $500/sq ft with spec-grade finishes). One operator self-performing with their own crew reported Palo Alto and Santa Clara builds under $300/sq ft, below the $400 floor above, but that figure is hard costs only: construction management, soft costs and carry are absorbed into their margin rather than paid out. It is a wage for doing the work, not a cheaper build, and it is not comparable to the all-in numbers beside it. Do not plug it into a model that charges soft costs and carry separately, which this one does. Cost per foot also falls as the house gets bigger, so a flat rate overstates the cost of a large rebuild.
• Soft Costs: Architectural plans, engineering, and the demolition permit. The 10–15% of hard costs usually quoted folds in loan interest and property tax during development. This guide breaks those out as carrying costs below, which leaves soft costs at about 6%. Quoting 10–15% and a separate carry line double-counts the financing.
• Demolition Costs: Depending on the size of the existing structure and whether asbestos abatement is required, simply clearing the lot can cost $30,000 to $60,000.

Your buildable square footage isn’t a free variable, either, it’s capped by zoning. Before you trust an ARV, confirm what you can actually build under local Floor Area Ratio (FAR) and lot-coverage limits.

3. Factoring in Time (The Holding Cost)

Time is the silent killer of developer ROI. Every month your project sits waiting for a permit, you are paying interest on your loan.

Treat the approval window as an assumption you set, not a number you can look up. Peninsula cities do not publish demolition or new-construction processing-time standards, and the permit records we ingest carry an issued date but no application date, so neither we nor anyone else can measure a true application-to-approval duration from the public record. Model it as a range instead, and run the RLV at a fast case and a slow case to see whether the deal still works at the slow one.

The carrying cost itself you can compute, and you should, because it is not a flat monthly figure. It is dominated by the land, which on the Peninsula is the largest line in the deal:

monthly carry ≈ (land × (loan rate + 1.25% property tax) + ½ × hard cost × loan rate) ÷ 12

Half the hard cost, because a construction loan draws down over the build rather than funding on day one. Run over five Peninsula teardown lots that sold for $1,700,000 to $3,150,000 in August 2026, at 8% money and hard costs across our verified $400 to $700/sq ft band, the formula returns $16,500 to $32,200 per month. The $15,000/month benchmark this guide used to quote, and that circulates widely, sits below the bottom of that range and is less than half the top of it. On the worked example further down, where the land solves to $5.35M, carry computes to about $51,250 per month. A flat monthly figure is really a statement about a land price, so anyone quoting you one is quoting a lot cheaper than the one you are looking at.

On duration, this guide models 18 months, and after checking it, 18 stays. Two Peninsula operators gave us timings in August 2026: a builder doing buy-to-relist in 17 months, and an owner-build at 24 months, a year to permit and a year to build, with the delay conceded to be on the owner’s side. The worked example below is a spec builder selling the finished house, so 17 to 18 months is the case that fits it, and 24 months is the slow case rather than the base one. The gap is not a rounding error: six extra months costs about $308,000 on the worked example, and $107,000 to $180,000 on those $1.7M to $3.15M lots. Who builds the house moves the return about as much as what it costs per foot.

What you can pin down is which reviews will attach to the job, since those are published per city, and they are what stretches the schedule. Our demolition permit guide walks the ones that bite hardest on the Peninsula: historic review triggered by building age, heritage tree protection, hillside grading, and the BAAQMD asbestos notification.

A Worked Example: 5,000 sq ft Peninsula Rebuild

Here is the full RLV and ROI math for a representative luxury teardown rebuild, using AddressIntel’s mid-range Peninsula benchmarks:

Read it as one scenario, not as an answer. It assumes $600/sq ft, inside the verified $400–$700 builder band but above the $550 midpoint our own screener computes at, and it assumes a $2,500/sq ft resale. Re-solve it at your own two numbers before you use it. That is not a disclaimer: on the table further up, those are the first and second most powerful inputs in the whole model.

Illustrative AddressIntel mid-range Peninsula benchmarks; actual figures vary by city, neighborhood, and finish level. The land number is the residual: it’s the most you can pay for the lot and still hit a 25% return on cost. The $51,250/mo carry is not a benchmark either, it is this lot’s own: 8% on $5,352,500 of land, plus 1.25% property tax on it, plus 8% on half of the $3,000,000 hard cost, which computes to $51,259/mo and is shown rounded so the column ties. Soft costs sit at 6% rather than the 10–15% commonly quoted because that quote includes the carry, which is its own line here. Note the direction of travel here. The 25% is an input you chose, not an output the market produced, and the whole table is solved backwards from it. That is the correct way to make an offer and a terrible way to forecast a return, which is exactly the distinction this page used to blur.

Automating the Math with AddressIntel

Manually calculating the Residual Land Value for every property that hits the market is impractical. By the time you run the comps, call the city about FAR restrictions, and estimate your holding costs, the property has already been sold off-market to a larger developer.

This is why we built the Teardown Predictor. It scores residential parcels across 15 Peninsula and South Bay cities and models every active listing it can price, running the RLV model with the same flat, comparable assumptions on every lot ($550/sq ft build, the midpoint of the builder band above; $25,000 demolition; a 5% cost of sale; and no carry, because it models a cash buyer) against recorded-sales comps from the subject’s own city, and assigns a Developer ROI score. An off-market parcel carries a teardown score but no ROI until it lists, because the model needs an asking price.

Building on Nantucket instead? The cost structure and historic-district constraints differ, start with the current Nantucket median price and our HDC & Flippability guide.

Frequently Asked Questions