CO or Backdraft Discovered During Inspection: Response Procedure

Purpose

Sometimes the technician's own combustion analyzer or draft test discovers a carbon monoxide or backdraft problem on a routine visit where no one reported an alarm or a symptom. This procedure guarantees the appliance gets shut down and the home protected off the numbers on the meter, the same way every time, rather than off a judgment call made under time pressure with a customer standing nearby. Skip it and a tech either shuts down a borderline reading too cautiously and burns trust, or waves off a reading that should have stopped the appliance because the visible cause looked minor.

Scope

Covers the response when CO or backdraft evidence surfaces from the technician's own testing during a scheduled sweep or inspection: a combustion analyzer reading, a draft test under worst-case depressurization, or a physical cue such as sooty deposits or a yellow flame on a gas appliance. Does not cover a customer-reported alarm event already in progress when the technician arrives, owned by the Why Did the CO Alarm Trip decision tree, and does not cover general CO detector code requirements or homeowner education, owned by the Carbon Monoxide Detectors reference. This procedure owns what the technician does in the minutes after their own meter reveals the finding, including the re-fire check after a correction.

Roles and responsibilities

The procedure

1. Stop the combustion test and shut the appliance down the moment the reading crosses the stated ceiling. CO above 100 ppm air-free signals a combustion problem to investigate before clearing the appliance; CO above 400 ppm air-free is the ANSI Z21 ceiling and means shut the appliance down and tag it immediately. Acceptance: the analyzer reading is recorded with a timestamp, and the appliance is shut down, fuel valve or main switch off per appliance type, the moment 400 ppm air-free is confirmed. Wrong looks like letting it run a bit longer to see if the reading settles. Stop rule: no appliance stays firing past a confirmed 400 ppm air-free reading, and any reading between 100 and 400 gets investigated before the appliance is cleared rather than dismissed because it did not hit the hard ceiling. Hazard: standing at the flue opening with the probe in place during a spillage event puts the technician directly in the gas stream; keep a personal CO monitor on and step back from the opening once the reading is captured rather than lingering to watch the number settle.

2. Check your own exposure, and evacuate or ventilate any occupied space that reads elevated. Read the personal monitor first; if it alarms or reads above outdoor ambient in the room you are standing in, leave that room before continuing. If step 1's finding also shows elevated CO in an occupied room rather than only at the appliance or flue, get occupants out and open windows or doors to ventilate. Acceptance: the personal monitor and any occupied-space reading both return to at or near outdoor ambient, confirmed by a recheck, before anyone continues working or re-occupies that room. Wrong looks like pushing through a monitor alarm because the job is not finished, or shutting the appliance down while the room itself stays unvented. Stop rule: a technician's own monitor alarming means that technician leaves immediately, and occupants do not re-enter a room that tested elevated until the recheck confirms it, not the first reading alone.

3. If backdraft rather than a high steady-state reading is the finding, confirm it under worst-case depressurization before concluding anything. Run the draft test with a manometer or smoke pencil at the appliance opening, appliance hot, with the home in worst-case configuration: exhaust fans on high, the dryer running, and a competing fireplace lit if one is present. Acceptance: a stated draft reading or clear smoke-pencil direction recorded under that worst-case configuration, not a calm-house baseline. Wrong looks like testing once in a quiet house and calling backdraft ruled out. Stop rule: a test not run under worst-case depressurization does not count as ruling backdraft out; retest under that configuration before clearing the appliance either way. Hazard: deliberately inducing worst-case depressurization at a hot appliance opening recreates the exact spillage-exposure condition step 1 warns about; keep the personal CO monitor on and do not linger with your face over the opening once the reading or smoke direction is captured, and if the personal monitor alarms during this test, step back per step 2's rule before finishing the read.

4. Rule out or document non-chimney sources before closing the visit as chimney-only. Walk the home and note every other fuel-burning appliance, its vent type, and its status. If you lack the equipment or license to test a furnace heat exchanger or a water heater's draft, say so in writing rather than letting a chimney-clear finding read as a whole-house clear. Acceptance: a written list of every fuel-burning appliance observed with a status of tested-clear, tested-flagged, or not tested-refer. Wrong looks like "the chimney's fine" language with no scope note on what else was or was not checked. Stop rule: a report that clears the chimney without naming what else in the home was tested is not complete; add the scope note before the visit closes.

5. If the visible cause is corrected and you want to re-fire the appliance, the re-fire gets its own numeric recheck, not a visual sign-off. A cleared blockage, a repositioned damper, or an identified competing exhaust source looking obviously fixed is not the same as confirmed fixed. Re-run the combustion analyzer and the same worst-case draft test used in step 3 after the correction, before telling the customer the appliance is clear to use. Acceptance: a second CO reading below 100 ppm air-free and a draft reading that holds under the same worst-case configuration. Wrong looks like reinstating use because the visible cause is gone. Stop rule: no appliance is cleared for use on a corrected-cause assumption alone; a recheck that does not hold under worst-case configuration keeps the appliance locked out regardless of how obvious the fix looked, and gets referred for further diagnosis rather than cleared on the strength of the obvious fix. Hazard: this re-fire test recreates the same combustion and spillage exposure already confirmed dangerous once at this appliance; the same personal-monitor and step-back protections from steps 1 and 2 apply in full here, not just on the first pass, and a reading climbing back toward the 400 ppm ceiling during this recheck stops the test immediately rather than waiting to see if it settles.

6. Confirm CO detector coverage in the home before you leave, regardless of outcome. Verify presence and function on every level and near sleeping areas per the detector reference this procedure does not re-derive; check the manufacture date against the stamped service life. Acceptance: each level checked, test button pressed, result and any gap noted. Wrong looks like skipping this because the chimney work feels like the real fix. Stop rule: a home with a confirmed combustion or draft finding and no working CO alarm nearby gets that gap documented in writing before you leave, even if the customer says they will handle it later.

7. File the complete record and reseal anything opened for testing. Write up every reading with a timestamp, the appliance status, the non-chimney appliance list, and the CO detector status. If a test port was drilled in the vent pipe for the analyzer probe, plug it with the correct sized plug and confirm no gap remains before leaving. Acceptance: the customer has a written copy the same visit, and any test port is visibly plugged and seated. Wrong looks like a verbal "it's fine now" with the readings only in the technician's head, or a test port left open. Stop rule: no visit closes without the numeric readings on paper, and an unsealed test port is a permanent leak the next spillage event traces straight back to this visit.

The record this produces

A combustion-safety file: every timestamped CO and draft reading, the appliance's lockout or cleared status, the non-chimney appliance list with each one's test disposition, and the CO detector check. This is the record that answers "what did the technician actually measure" if the appliance trips a customer's own alarm weeks later, and it is what a referred trade reads before they arrive on a source this visit could not test.

One run of this procedure, filled in

Annual Level 1 sweep and combustion check on a gas fireplace insert, no reported symptoms on the work order. Outdoor ambient CO reads zero.

• Step 1: the analyzer reads 460 ppm air-free on the initial test, above the 400 ppm ceiling. Appliance shut down and tagged immediately; the technician steps back from the opening once the reading is captured.
• Step 2: the personal monitor does not alarm and the room itself reads at outdoor ambient, so the finding is contained to the flue and appliance rather than the occupied space; no evacuation needed.
• Step 3: draft test under worst-case configuration, bath fan and dryer both running, shows clearly reversed flow at the appliance opening. Backdraft confirmed.
• Step 4: the home also has an atmospheric water heater in the same closet. The technician is not equipped for a combustion test on it, notes it as not tested-refer, and tells the customer directly rather than implying the whole house is cleared.
• Camera scan finds a partial bird's nest blocking roughly half the flue's cross-section, a plausible cause for both the CO reading and the backdraft. It is cleared.
• Step 5: after clearing the blockage, the appliance is re-fired for the recheck. Fail, stop rule taken. The second CO reading comes back at 150 ppm air-free, still above the 100 ppm investigate threshold even though the obvious blockage is gone. The appliance stays locked out and is referred for further diagnosis rather than cleared on the strength of the visible fix.
• Step 6: one CO alarm in the home is found nine years past its stamped service life; documented as a gap and a replacement recommended.
• Step 7: the report lists every reading, the appliance's continued lockout status, the water heater referral, and the alarm gap. A small test port drilled for the analyzer probe is plugged and checked before the crew leaves.

The step 5 failure is the part worth keeping. The nest looked like an obvious, complete explanation, and a visual sign-off would have cleared the appliance for use with a real combustion problem still present, which is exactly the gap the numeric recheck is built to catch.

When the response does not go as written

No combustion analyzer is on the truck, or the technician is not combustion-certified for that fuel type. Physical cues still apply: a yellow flame instead of blue, soot on the appliance body, or rust staining around a vent connection are enough to lock the appliance out and refer it for combustion testing, even with no numeric reading in hand.

The customer downplays the finding, saying the appliance has always run a little rough. A prior pattern does not change the numeric thresholds in steps 1 and 5; lockout and recheck apply the same regardless of how long the condition is believed to have existed.

A solo technician has no one to walk the rest of the home for other appliances. Complete the walk yourself before closing the visit rather than skipping step 4; a chimney-only report with an untested water heater in the same house is incomplete either way.

References

• ANSI Z21, combustion product limits referenced in the shop's annual inspection standard.
• See related: Why Did the CO Alarm Trip, Real Spillage vs Sensor vs Nuisance Decision Tree, for post-alarm troubleshooting once an alarm has already sounded.
• See related: Carbon Monoxide Detectors and Chimney Safety, for detector code requirements and customer education this procedure does not re-derive.
• NFPA 54, National Fuel Gas Code, current edition, combustion appliance venting requirements.
• CSIA (Chimney Safety Institute of America), spillage diagnosis and worst-case depressurization testing guidance.