Pulscar P0441 analysis — method
Pulscar analysis of NHTSA public data, September 2026.
Built 19 September 2026. Licence: CC BY 4.0. Credit "Pulscar" and link to
https://www.pulscar.io/blog/p0441-code

SOURCES
- NHTSA ODI complaint flat files COMPLAINTS_RECEIVED_2015-2019 / 2020-2024 / 2025-2026.
  Field dictionary: https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt
  777,040 distinct vehicle complaints (ODINO, product type V) received 1 January 2015 to
  8 September 2026. Every figure below comes from this one file unless stated.
- NHTSA manufacturer communication (TSB) flat files TSBS_RECEIVED_2015-2019 / 2020-2024 /
  2025-2026. Field dictionary: https://static.nhtsa.gov/odi/ffdd/tsbs/TSBS.txt
- NHTSA recall flat files and the public recall API
  (https://api.nhtsa.gov/recalls/campaignNumber?campaignNumber=...).
- NHTSA investigation flat file FLAT_INV. Layout: https://static.nhtsa.gov/odi/ffdd/inv/INV.txt

CASE SENSITIVITY
Complaint narratives are recorded in ALL CAPITALS through 2020 and in mixed case after that.
Every text query below is case-insensitive (SQL LIKE on a lower-cased column, or a Python
regular expression with re.I). A capitals-only query would drop most of the post-2020 file.

QUERY 1 — complaints naming the code
  descr LIKE '%P0441%' (case-insensitive)  ->  29 complaints.
The Pulscar Check Engine Code Index publishes 28 for the same code because it matches the code
as a whole word (\bP0441\b); the extra row here writes the code run together with a scan-tool
suffix. The difference is one report and changes nothing. Allowing the typo "P441" adds 1 more.
All 29 were read end to end, and each was given one label under a written rule:
  "subject"  = the failure described is the evaporative / purge / fuel-vapor system, or the
               owner names a purge part as the diagnosis or the repair;
  "code list attached to another failure" = P0441 appears in a list of stored codes belonging
               to a different described failure (a seized engine, an ABS module, a fuel pump).
Result: 16 subject, 13 code list. Reported outcomes inside the 16: 3 report a repair the owner
says worked, 3 report EVAP parts replaced with the fault unchanged, 10 report no repair reached.
Published per-report in p0441-29-complaints-read.csv. One reader, one pass, no second coder.

QUERY 2 — the component set (the main analysis on the page)
  descr LIKE '%purge%' (case-insensitive)  ->  1,593 complaints.
Chosen because P0441 is a code most owners never see: only 14 of these 1,593 quote it.
FALSE-POSITIVE READING: 30 complaints drawn at random (Python random.sample, seed 20260919)
were read in full before any count was published. All 30 describe an evaporative emission purge
component (purge valve, canister purge valve, purge solenoid, purge pump, purge line). 0 of 30
false positives, so the query was not tightened. The word "purge" does appear in other automotive
senses (purging a brake or air-conditioning line); none surfaced in the sample, and any that
remain are a small minority of a 1,593-row file.
Exclusions: none beyond the file's own scope (product type V, passenger vehicles).
Mileage figures drop rows with a reported odometer of 0 or 400,000 and above: 886 of 1,593
complaints state a usable mileage. Median 77,710 miles; quartiles 54,000 and 108,000.

PHRASE GROUPS (regular expressions, all case-insensitive, applied to the whole narrative)
  stall or loss of motive power:
    \bstall|lost (motive )?power|loss of (motive )?power|shut(s|ting)? (off|down)|cut(s|ting)? off
  refuelling mentioned:
    \brefuel|refill|fill(ing|ed)? (up|the tank)|gas station|after (getting|pumping) gas|fuel(l)?ing
  hard start or no start:
    (hard|difficult|trouble|unable|won't|will not|fail(s|ed)? to)[^.]{0,60}start
  fuel tank deformed or collapsed:
    tank[^.]{0,40}(deform|collaps|crush|suck|implod|warp|dent)
    |(deform|collaps|crush|implod|warp)[^.]{0,30}(fuel )?tank
  a recall or campaign named:  \brecall
  fault returned or repeated:
    (again|second time|third time|multiple times|repeatedly|keeps? (failing|happening|coming back)|came back)
  shop could not find or duplicate it:
    could not (find|duplicate|replicate|determine)|cannot (find|duplicate|replicate)
    |unable to (find|duplicate|replicate|diagnose)|no (fault|problem) found
  parts named: purge (valve|solenoid|control valve|pump)|canister purge|\bCPV\b ;
    (charcoal|carbon|vapor|evap(orative)?) canister ; fuel tank|gas tank ;
    fuel pump|fuel delivery module ; gas cap|fuel cap|filler cap ;
    (pcm|ecm|ecu|computer|module)[^.]{0,40}(reprogram|update|software|flash)|software (update|reprogram)|reflash
A phrase group counts a complaint once however many times the words appear. Groups overlap:
one complaint can be counted in several. Full results in
purge-complaints-what-owners-describe.csv.

BASELINES
Across all 777,040 complaints: towed 5.11%, crash 4.37%, fire 1.82%.
Across the 1,593 purge complaints: towed 3.20%, crash 0.82%, fire 0.31%, 11 injuries, 1 death.

QUERY 3 — owner-reported money
All 1,593 narratives were split into sentences; every dollar amount between $20 and $30,000 in a
sentence that also names a purge, canister, fuel tank, fuel pump or EVAP part was extracted:
71 sentences in 60 complaints. Each was then READ and given one category, and amounts that were
diagnostic fees only, vehicle values, or bills for unrelated work were dropped by reading rather
than by rule:
  A "purge valve or solenoid job" — 26 figures, median $500, range $147 to $1,043.
  B "tank, pump or canister after a stuck-open valve" — 21 figures, median $1,500, range
    $331 to $2,700; the 11 Ford Focus rows inside B have a median of $1,960.
These are owner reports of what they were quoted or paid, not invoices we have seen, and 47
figures is not a price survey. Published per-row in purge-owner-reported-amounts.csv.

QUERY 4 — manufacturer communications naming P0441
  tsbs.summary LIKE '%P0441%'  ->  2,021 make/model/year rows, 225 distinct NHTSA document IDs,
  from 7 manufacturer groups: Stellantis 157, VW Group 30, Honda/Acura 25, Kia 7, Hyundai 3,
  Mazda 2, Volvo 1. 136 of the 225 describe a software or calibration change, matched with
  re-?program | re-?flash | software (update|upgrade|logic) | calibration (update|change) |
  logic improvement | ECU upgrade | ECM upgrade | improved software.
  Published in p0441-bulletins-by-group.csv.
THIS IS A DIFFERENT NUMBER FROM THE PULSCAR CHECK ENGINE CODE INDEX, WHICH PUBLISHES 82, AND
BOTH ARE CORRECT. The index counts distinct bulletins after four exclusions this raw count does
not apply: helpline "analysis lists" (Volkswagen's VIN4APIN documents), Honda and Acura dealer
messages and parts-collection requests, documents whose ID contains no digits, and revision
normalization that folds Rev A / Rev B / _1 / _2 into one bulletin. The article cites the index
figure of 82 for bulletin counts and uses 225 only where the point is the raw document trail.
Index method: https://www.pulscar.io/research/check-engine-code-index/method.txt
ZERO RESULT, PUBLISHED AS A FINDING: Ford, Lincoln, Mercury, Chevrolet, GMC, Buick, Cadillac,
Toyota, Lexus, Nissan, Infiniti, Subaru, BMW, Mini and Mercedes-Benz have no document in the
file whose summary names P0441.

QUERY 5 — recalls
  recalls where defect, consequence, remedy or component contains "purge" (case-insensitive)
  -> 6 distinct campaigns: 18V-735 (1,282,596), 15V-793 (411,205), 26V-369 (257,868),
  19V-515 (57,488), 11V-134 (52,000), 17V-424 (193). Populations are the "potential number of
  units affected" recorded in the campaign; campaigns overlap, so they are not added except
  where the article says so.

QUERY 6 — investigations
  investigations where subject or summary contains "purge" -> RQ23-006 (Ford Focus, opened
  15 September 2023, closed 20 December 2024), PE12-016, PE13-016, RQ06-002, RQ23-007 and one
  1998 air-brake file. Resume texts read from the flat file.

FORD FOCUS TIMELINE
507 of the 1,593 complaints are Ford Focus. 506 are model years 2012-2018, the range of recall
18V-735. 385 of the 507 (75.9%) were received on or after 19 August 2019, the date the 18V-735
recall repair began. 32 pair a recall number with a word for the fault returning.

LIMITATIONS
- Complaints are self-reported, unverified, and are not a failure rate. Counting them measures
  what reaches a federal form, which is driven by recall publicity as much as by failures:
  43.4% of this set already names a recall, and the Ford Focus alone is 31.8% of it.
- Raw counts across makes reflect very different sales volumes. Nothing here ranks reliability.
- Phrase counts measure vocabulary, not diagnosis. "Purge valve named" means an owner typed it.
- Model year is a blunt age proxy; the warranty clock runs from the in-service date.
- The 29-report hand coding is one reader's single pass on a small set.