How long an ADU takes, as five clocks

There is a 60-day clock in statute and a 162-day median in the data

Both are true. They are not in conflict, because they do not measure the same span — and understanding why is worth more to your schedule than either number on its own.

The statute. Under Gov. Code § 66317(a)(3), as amended by SB 543 (Stats. 2025, ch. 520, effective 1 January 2026), the permitting agency "The permitting agency shall either approve or deny the application to create or serve an accessory dwelling unit within 60 days from the date the permitting agency receives a completed application if there is an existing single-family or multifamily dwelling on the lot." And if it does not: If the local agency has not approved or denied the completed application within 60 days, the application is deemed approved.

If the applicant requests a delay, the 60 days are tolled for the length of the delay.

The observed durations. Los Angeles: median 162 calendar days across 27,188 non-supplemental ADU building permits, with an interquartile range of 79 to 301 days and 19% taking more than a year. Unincorporated Marin: median 174.5 across 472.

San Francisco: median 445 across 1,713 issued permits, IQR 276 to 707, and 62% over a year. All three are HyreADU calculations on the cities’ own published extracts; the method is on the permit-times research page.

Say the caveat plainly, because it is the whole point. These are calendar days from the published application-side date to issuance, on permits that issued.

The statutory 60 days runs from a COMPLETED application, and none of the three extracts marks the completeness determination.

The gap between the two therefore includes every day an applicant spent curing an incomplete set, paying fees, answering plan-check comments, or asking for delay.

A median above 60 days is not, by itself, a finding that a city missed the shot clock.

So what is the gap made of? Everything between hitting submit and the agency holding a set it considers complete: missing documents, unpaid fees, structural sheets that were not signed, energy calculations that were not attached, plan-check comments that took three weeks to answer, and any delay the applicant asked for — which the statute expressly tolls.

None of that is the counter being slow. Some of it certainly is queue time as well; the extracts do not let anyone separate the two, and we will not pretend otherwise.

The inference that is actually useful. Our reading is that if a large part of the gap is time spent curing an incomplete set, then it is time your team controls.

That reframes what you are buying when you hire a designer. The most valuable thing a designer does on an ADU may not be the elevation — it may be submitting a set complete enough that the statutory clock actually starts on the day you file.

This is analysis rather than a finding in the data, and it is labeled as such, but it is consistent with everything below.

And a note on what a wide distribution means for planning. A median of 162 days with a quarter of cases above 301 is not a duration you can put in a contract.

The honest way to use it is as a distribution: plan against the median, hold a reserve against the upper quartile, and do not build a financing structure whose viability depends on landing near the fast end.

The clock does not start when you submit. It starts when you are complete

The completeness rule is the single most consequential sentence in the statute for your schedule, and a 2025 amendment gave it a clock of its own.

What SB 543 added. SB 543 added a completeness clock: a written completeness determination within 15 business days, a list of what is missing, a limited resubmittal, and a deemed-complete consequence if the agency misses the determination.

An appeal of incompleteness or of a denial gets a final written determination within 60 business days (§ 66317(d)). Those are business days; the 60-day decision is not labeled as business days. The cite is Gov. Code § 66317(a)(2).

Read the units carefully. The completeness determination is 15 business days. The appeal route is in business days too. The 60-day decision is not labeled as business days.

Business days and calendar days are different quantities, and a fortnight of public holidays sits differently inside each of them. When you put dates in a schedule, convert deliberately rather than assuming.

What the amendment gives you that you did not have. Before a completeness clock existed, an application could sit in an ambiguous state — filed but not accepted — for as long as it took, and nothing was running.

A written determination inside a fixed window, with a list of what is missing and a deemed-complete consequence if the agency misses it, converts that ambiguity into two dated events you can hold people to.

That is a genuine change in your position, and it is recent enough that a lot of guidance has not caught up with it.

What it does not do. It does not stop the list of missing items being long, and it does not stop the resubmittal cycle being where the weeks go. The remedy for that is upstream: a complete set on day one. There is no procedural device that substitutes for documents that exist.

And cite it correctly, because that is a test of your sources. SB 477 (Stats. 2024, ch. 7), an urgency measure effective 25 March 2024, recodified State ADU Law out of Government Code § 65852.2 into Chapter 13 of Division 1 of Title 7, sections 66310 through 66342.

A guide that still cites § 65852.2 is citing a repealed numbering.

State ADU law now sits at Gov. Code §§ 66310–66342, and the fee section moved again in 2026: The ADU fee rules that used to sit in former § 65852.2, and then at § 66324, were renumbered to § 66311.5 by SB 543 (Stats. 2025, ch. 520, effective 1 January 2026).

A handout, a consultant or a checklist still working from § 65852.2 has not been revised since March 2024 — and the completeness clock described above did not exist then. That is not a gotcha.

It is a fast way to find out how current the advice in front of you is.

Permit type moves the schedule too, and in the direction you would expect

The same extract, cut by the permit type rather than the review route. It is the closest published proxy this site has to "conversion versus new building".

The three published categories. Bldg-Alter/Repair: median 102 days across 9,683 permits, IQR 38–203, 35% decided within 60 days. Bldg-New: median 218 days across 9,145 permits, IQR 130–373, 7% decided within 60 days. Bldg-Addition: median 175 days across 8,360 permits, IQR 94–316, 14% decided within 60 days.

What that ordering is worth to you. The heaviest category is new construction, which is what a detached ADU usually is, and the lightest is alteration and repair, which is where a lot of conversion work lands.

So the form of the unit you choose is also a choice about the queue you join. That is a real planning input, and it is one people make on cost and aesthetics without noticing that it moves the calendar too.

Be careful about what it does not prove. Permit type is a coding convention in a municipal system, not a description of your building.

A given ADU can be coded more than one way, the categories overlap in practice, and the difference in medians is confounded with scope, complexity and the review route that scope attracts.

We publish the cut because the direction is stable and useful; we do not publish it as a causal claim that changing the code on your form would change your date.

And it does not transfer between cities. This is one city’s coding of one city’s permits.

San Francisco’s median of 445 days across 1,713 permits, with 62% over a year, tells you that the between-city variation dwarfs the between-type variation.

Ask your own counter for their current ADU review times and type that into the form.

Two cities are missing from this data, and that is a finding

We could have substituted a proxy for them. It would have been the least honest thing on the page.

City of Sacramento. The published issued-permit layer has Status_Date (the issued/status date) and Application as a permit number (e.g. RES-2600002). No application/filed/submitted date field is published. Duration cannot be computed. Checked 2026-09-05.

City of San José. CKAN resources publish ISSUEDATE and FINALDATE. No application, filed, received or submitted date. Duration cannot be computed. Checked 2026-09-05.

Why this matters beyond these two cities. A duration needs two dates. Where a portal publishes only the issued date, no amount of analysis recovers the missing one, and any figure produced anyway is an assumption dressed as a measurement.

Both of these are large jurisdictions with active ADU programs, so their absence is not a rounding issue in the coverage — it is a real limit on what this site can say about California as a whole.

What you can do about it for your own city. Ask the building department directly for current ADU review times, in weeks, for a complete set, and ask them to distinguish first review from back-check.

That single question produces a better input for the form above than any statewide figure could, because it comes from the people who will actually do the review.

And treat the absence as a signal about other sources. If a guide publishes a confident permit-duration figure for a city whose open data contains no application date, ask where the number came from. Sometimes there is a good answer. Often there is not.

The permit clock ends and the construction clock starts

The state publishes both permits and completions, and the lag between them is its own schedule — one this form’s later phases are trying to describe.

The published lag. Among matched projects in California Department of Housing and Community Development, Housing Element Annual Progress Report, Table A2’s reporting, the median lag from permit to completion is 1 year: 29% report a completion in the permit year, 79% within one year and 94% within two.

That is a HyreADU calculation on the state’s own dataset, and the method is on the permits-versus-completions research page.

Which puts the permit phase in proportion. If the median permit takes months and the median build takes about a year after that, then a project is a multi-year commitment even when nothing goes wrong. The form above sums five phases without overlap precisely so the total is uncomfortable enough to be realistic.

And the part nobody schedules for. 41% of ADU permit units issued between 2018 and 2023 — 46,944 of 114,651 — have no matching completion in the file.

Read that with the caveat attached: These are reported APR rows, not a physical census of construction.

An unmatched permit may be a unit still under construction, an abandoned project, a unit completed without a reported certificate, or a project whose identifier changed between the permit row and the completion row.

The direction is robust; the exact rate is a reporting artefact as much as a construction one.

Our reading of what that means for a timeline. The schedule risk on an ADU is not mainly that each phase runs a bit long.

It is that the project stops somewhere after approval — which is the point at which somebody has to produce the money. A timeline that assumes financing is arranged is modeling the easy case. Arrange the money before the fees, not after.

A note on inspections. Passing inspections are already inside a competent building-phase estimate. Failed ones are not, and they are the reliable source of schedule surprise in the finishes phase.

Ask your builder how many inspections they expect, what the usual re-inspection interval is at your department, and who pays for a re-inspection fee. That last one belongs in the contract, not in a conversation held after the failure.

When the schedule is the reason to wait

Some totals should stop a project rather than start one. These are the ones worth taking seriously.

When the finance has a shorter fuse than the permit. If a bridging facility, a rate lock or an interest-only period expires inside the range the observed distributions describe, the schedule risk has become a financing risk.

Waiting until the money is arranged on a longer horizon is a legitimate decision, and it is cheaper than discovering the mismatch halfway through plan check.

When the only version that works is the fast end of the distribution. A plan that requires you to land near the 25th percentile is not a plan, it is a hope with dates on it. If the project only works at the fast end, it does not work.

When you have not had the planning conversation yet. Starting the design clock before feasibility is the classic expensive sequencing error, and it is the one this page most wants to prevent. The conversation costs a phone call; the drawings do not.

When a household event sits inside the window. If the reason for the ADU is a relative arriving on a date, then the honest question is whether a multi-year process reaches that date at all, and what the fallback is if it does not.

That is a family question rather than an arithmetic one, and it deserves to be asked before the fees are paid rather than after.

What we are, once. HyreADU does not design, permit or build accessory dwelling units. It does not appraise, lend, invest, let property or prepare taxes, it does not rank or refer contractors, and it takes no referral fee from anyone who does any of those things.

That is what allows these tools to conclude that the numbers do not work, that the answer is to keep the garage, or that the honest next step is a question at the planning counter rather than a deposit.

We hold no dataset of ADU bids, quotes or contracts. Every dollar figure on this site is a named third party’s published figure, arithmetic we performed on one and have labeled as ours, or a number you typed in.