BRENNEN

Industries / Construction and property / Damage restoration

Rebuilding

Rebuilding is the repair or reconstruction that follows the cleanup: putting back walls, floors, siding and roofs so that they match what was there and meet today's codes and, after a disaster, rebuilding whole streets.

4

Milestones achieved, from 1990 to 2025.

1

To watch for: announced, not yet confirmed outside.

2

Open deliverables. Nobody has shipped them.

The road

AchievedWatch forOpen

Achieved

  1. 1990

    A rule for matching

    A repair often puts new material beside old: new boards in a floor, new panels on one wall of siding. The NAIC's model regulation on unfair property claim settlement practices, adopted in 1990 and amended in 1991, says that when replaced items "do not match in quality, color or size, the insurer shall replace all items in the area so as to conform to a reasonably uniform appearance", inside and outside the home, at no cost to the policyholder beyond the deductible. Each state decides whether and how to adopt it.

    LessonThe rule names a goal, a reasonably uniform appearance, but gives no measure of when two materials stop matching.

    content.naic.org

    Credit

    National Association of Insurance Commissioners

  2. 1993

    Code costs become an option

    Before Hurricane Andrew in 1992 most homeowners policies excluded the extra cost of meeting current building codes, and many owners found that rebuilding to the current code cost more than their limits, by the Florida Office of Insurance Regulation's account. A law of November 1993 (Chapter 93-410, creating section 627.7011) made Florida insurers offer "law and ordinance" cover for those costs at 25% of the dwelling limit; amendments in 2005 added a 50% option, and a policy now includes the 25% cover unless the owner refuses it in writing. In 2006 the regulator found Florida was the only state to require the offer, and that insurers were not tracking these payments accurately, partly because contractors were not required to separate code costs from other repair costs.

    LessonCover for code costs is hard to price or check when estimates do not show those costs on lines of their own.

    floir.govleg.state.fl.us

    Credit

    Florida Legislature, on the recommendation of its Study Commission on Property Insurance and Reinsurance

  3. 2015

    An insurer repairs 66,000 homes

    After the Canterbury earthquakes of 2010 and 2011, New Zealand's Earthquake Commission (EQC) ran home repairs itself through the Canterbury Home Repair Programme, with Fletcher Construction as project manager, to hold down repair prices, make sure claim money went into repairs and keep the region's houses in good condition. By 30 June 2015 it had practically completed 66,252 repairs. The Auditor-General found that the average repair cost rose 14.2% from February 2011 to June 2014, against 30.9% for new houses in Canterbury, but that about 8% to 10% of repairs needed more work after they were finished, some did not meet the Building Code, and project management had cost about NZ$340 million.

    LessonRunning repairs at scale held prices down, but about one repair in ten had to be revisited.

    oag.parliament.nzoag.parliament.nz

    Credit

    Earthquake Commission; Fletcher Construction; findings by New Zealand's Office of the Auditor-General

  4. 2025

    Rebuilding counted lot by lot

    The Urban Institute, a US research group, collected property-level permit records from local governments for four wildfires. By February 2025, more than six years after the 2018 Camp Fire destroyed nearly 14,000 homes in Butte County, California, building permits had been issued for 29% of properties and certificates of occupancy, the sign-off that a house can be lived in, for 23%. Three years after Colorado's Marshall Fire, 75% of the affected houses had permits; 18 months after the 2023 Maui fires, 14% of 1,898 destroyed housing units did.

    LessonRebuilding went fastest in a metropolitan area with streamlined permits and slowest in a remote town that had lost nearly everything.

    urban.org

    Credit

    Urban Institute

Watch for

  1. Permits three times faster

    California's governor said on 5 January 2026 that, a year after the January 2025 fires in Altadena and the Pacific Palisades, local governments had issued 2,617 rebuilding permits from 6,191 applications, nearly three times as fast as comparable house permits in the five years before, with reviews averaging under 30 days; a year after the Camp Fire, 385 permits had been issued. CalMatters put the permits at roughly one for every five of the nearly 13,000 homes lost, and a builder it quoted warned that a permit is not a finished house.

    Achieved whenan outside count of certificates of occupancy, such as the Urban Institute's lot-by-lot method, shows Los Angeles homes being finished faster than homes after earlier fires at the same point.

    gov.ca.govcalmatters.org

    Claimed by

    Office of Governor Gavin Newsom, January 2026

Open

  1. 01

    When old siding stops matching

    A measured test of how far aged vinyl siding drifts in color from new stock sold today under the same product name and color, read with a spectrophotometer, together with a viewing test that finds the color difference at which people can see the mismatch from the street. Set the results beside matching rules that ask for a "reasonably uniform appearance".

    The brief
    Credit · unclaimed
  2. 02

    Code costs inside repairs

    An open set of interior repair estimates from water and fire losses in which every line is tagged as ordinary repair or code upgrade (such as added smoke alarms, arc-fault breakers, insulation to the current energy code or larger bedroom windows), and a model that predicts the code share of a repair from the house's age, location and scope, scored on estimates it has not seen.

    The brief
    Credit · unclaimed

Open 01 · the brief

When old siding stops matching

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

A measured test of how far aged vinyl siding drifts in color from new stock sold today under the same product name and color, read with a spectrophotometer, together with a viewing test that finds the color difference at which people can see the mismatch from the street. Set the results beside matching rules that ask for a "reasonably uniform appearance".

Why it is still open

The NAIC's matching rule gives no measure of a match. ASTM D6864, a specification for the color retention of plastic siding (replaced by D7856 in 2018), judges whether a product's color change over years of service is acceptable: it compares a panel with its own color when new, not with today's production of the same color beside it. Matching services work one claim at a time: SidingMatch said in 2008 that it reads each submitted sample with a spectrophotometer under controlled daylight, and ITEL compares samples with a database of more than 60,000 siding products from more than 130 makers. Neither publishes its measurements. Searches found no published set of weathered siding measured against new stock and no tested threshold for when a difference shows.

content.naic.orgastm.orgrandrmagonline.comwww2.itelinc.com

Built with what exists

A handheld spectrophotometer such as the X-Rite Ci64, or a low-cost colorimeter such as the Nix Spectro 2; the CIEDE2000 color-difference formula in the open-source colour-science Python package; panels removed during re-siding and remodeling of undamaged houses, with the year each was installed and the wall it faced; new panels bought from suppliers; a viewing panel of volunteers.

  1. First step

    Ask one siding contractor to set aside ten removed panels of known install year from this month's re-siding jobs, buy new panels of the same names, and measure both.

  2. Done looks like

    At least 50 aged samples covering at least 10 named colors from at least three makers, each with its install year, facing and CIEDE2000 difference from new stock; a viewing test of 30 people giving, for each color, the difference at which half of them see the mismatch at 3 m and at 10 m; all data under an open license.

Open 02 · the brief

Code costs inside repairs

Credit · unclaimed

The first to publish is credited on this line, and the deliverable moves from Open to Achieved.

What to build

An open set of interior repair estimates from water and fire losses in which every line is tagged as ordinary repair or code upgrade (such as added smoke alarms, arc-fault breakers, insulation to the current energy code or larger bedroom windows), and a model that predicts the code share of a repair from the house's age, location and scope, scored on estimates it has not seen.

Why it is still open

Florida's regulator found in 2006 that insurers were not tracking law-and-ordinance payments accurately and that contractors were not required to separate code costs from other repair costs. IRMI, an insurance publisher, wrote in February 2026 that some building owners use a rule of thumb of 1% to 2.5% for each year of a building's age, and that no rule of thumb can tell what amount of cover is enough. OneClick Code sells reports that give, address by address, a yes or no for each line item local code makes enforceable, but no labeled test set is published; this research found no public set of repair estimates with code items tagged.

floir.govirmi.comoneclickcode.com

Built with what exists

Closed repair estimates in which contractors listed code items separately, shared with consent and with personal details removed; the International Residential Code editions of the International Code Council (ICC) and each state's adoption dates; an extraction model such as Muse Ember to read estimate PDFs into rows (see the Note "The jobs agents repeat"); Label Studio for tagging; Python and scikit-learn for the model.

  1. First step

    Ask two restoration contractors for 20 closed water-loss estimates in which code items were listed separately, remove personal details, and tag each line.

  2. Done looks like

    At least 300 estimates from at least three states, each line tagged by two people who agree on at least 95% of lines; a model whose median error on 100 held-out estimates is published in percentage points of the estimate total; data and code under an open license.