PULSCAR ANALYSIS — P0456, THE SMALLEST EVAP LEAK: WHAT 133 OWNER COMPLAINTS AND 77 NON-STELLANTIS BULLETINS SAY IT ACTUALLY WAS, AND HOW OFTEN IT WAS THE CAP Pulscar analysis of NHTSA public data, September 2026. License: CC BY 4.0. Attribution: "Pulscar analysis of NHTSA public data, September 2026." Published alongside https://www.pulscar.io/blog/p0456-code ------------------------------------------------------------------------------ 0. WHAT THIS FILE ADDS, AND WHAT IT DELIBERATELY DOES NOT REPEAT ------------------------------------------------------------------------------ Three Pulscar pages already publish numbers on the EVAP leak codes. None is recomputed here; where this file uses one of their figures it cites it. * The Check Engine Code Index counts the manufacturer bulletins naming P0456 (269 distinct bulletins, 11 manufacturer groups, at least 194 software updates, 0 tied to a coverage or emissions campaign) and publishes its own method, its per-bulletin CSV and the complaint figures used below. https://www.pulscar.io/blog/check-engine-code-index * The P0455 guide classifies the 69 bulletins naming the LARGE leak code into 31 distinct problems outside Stellantis. https://www.pulscar.io/blog/p0455-code * The P0442 guide does the same for the SMALL leak code: 36 bulletins, 30 problems, and the regulatory test conditions. https://www.pulscar.io/blog/p0442-code This file answers what neither of those answers for the SMALLEST threshold: what 133 owner complaints naming P0456 turned out to be when read end to end, how often the fuel cap was the answer, and what the 77 non-Stellantis bulletins name as the problem once they are grouped. ------------------------------------------------------------------------------ 1. SOURCE FILES AND DATE RANGE ------------------------------------------------------------------------------ COMPLAINTS NHTSA ODI complaint flat files COMPLAINTS_RECEIVED_2015-2019, _2020-2024, _2025-2026 https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt (field layout) 777,040 distinct vehicle complaints (product type V), de-duplicated by ODI report number, received 1 January 2015 to 8 September 2026. TSBs NHTSA manufacturer communications flat files https://static.nhtsa.gov/odi/ffdd/tsbs/TSBS.txt (field layout) plus the published Check Engine Code Index bulletin list /research/check-engine-code-index/bulletins.csv RECALLS NHTSA Part 573 flat file and the public recall API https://api.nhtsa.gov/recalls/campaignNumber?campaignNumber=18V735000 REGULATION 13 CCR 1968.2 (California OBD II), read in full for sections (d)(3.2.1), (e)(4.2.1), (e)(4.2.2), (e)(4.3) and (g)(4.5.4). Data pulled 14 September 2026; this analysis run 19 September 2026. No third-party data was used. ------------------------------------------------------------------------------ 2. THE COMPLAINT SET, AND THE FALSE-POSITIVE CHECK ------------------------------------------------------------------------------ Query: the narrative field (DESCR) matched against the whole-word regular expression \bP0456\b, CASE-INSENSITIVE. Case matters: NHTSA narratives are stored in ALL CAPS through roughly 2020 and in mixed case afterwards, so a case-sensitive query silently loses most of the recent file. SELECT * FROM complaints WHERE descr LIKE '%P0456%'; -- 136 rows then kept only rows matching \bP0456\b -- 133 rows The three rejected rows were read. All three are run-together strings in which the code is not the code being reported: "PP0456" (a 2014 Ford), "P045600" (a 2020 Kia) and "P04567A" (a 2022 Hyundai). Excluding them keeps the set identical to the one behind the Check Engine Code Index, which uses the same whole-word rule and independently reports 133. FINAL SET: 133 complaints. This is the denominator for every percentage below unless another denominator is named. FALSE-POSITIVE READING. The brief for this work requires reading at least 30 random matches before publishing a count. We read all 133 instead of a sample. Result: 0 of 133 are a mis-parse. But 26 of 133 (19.5%) are "off target": the code is genuine and incidental inside a complaint whose real subject is a transmission, an oil pump, an unintended acceleration event, a valve cover, a CVT, an AWD warning or a used-car dispute. Those 26 are kept in the denominator and shown as their own category rather than quietly dropped, because dropping them would inflate every other share. Self-reported, no denominator, not a failure rate. People file federal complaints about expensive, unresolved or frightening problems. A P0456 that stopped coming back after somebody tightened the cap will essentially never produce one. Nothing in this file measures how often any repair works. ------------------------------------------------------------------------------ 3. THE CLASSIFICATION (p0456-complaints-classified.csv) ------------------------------------------------------------------------------ All 133 narratives were read end to end by a human reader and each was assigned exactly ONE primary outcome — the thing the story turned out to be about, as the owner tells it. Categories were written after the reading, not before. purge, vent or EVAP valve ................ 38 (28.6%) off target, code incidental .............. 26 (19.5%) no diagnosis stated ...................... 18 (13.5%) tank, filler tube, vent tube or hose ..... 18 (13.5%) no cause found, repair repeated .......... 11 (8.3%) cap, filler neck or capless filler ........ 9 (6.8%) software, calibration or campaign ......... 7 (5.3%) fuel pump or pump module seal ............. 6 (4.5%) ---- 133 Definitions used while reading: * "purge, vent or EVAP valve" — the narrative names a canister purge valve, vent or vent control valve, canister close valve, N80 valve, ESIM or the Ford canister-purge-valve recall family as the cause or the repair. * "tank, filler tube, vent tube or hose" — a cracked or corroded tank, filler tube, vent tube, rollover valve, tank check valve or vapor hose. * "cap, filler neck or capless filler" — the cap or the capless filler assembly is the named subject, whether or not replacing it worked. * "fuel pump or pump module seal" — the leak was located at the fuel pump, pump module or its seal. * "software, calibration or campaign" — a reprogram, a calibration or a recall is the event or the remedy. * "no cause found, repair repeated" — parts were replaced or the shop could not find anything, and the code returned; no settled cause. * "off target" — see section 2. * "no diagnosis stated" — the narrative reports the code and nothing else. HOW OFTEN WAS IT THE CAP — the number this page exists to answer. * 12 of 133 complaints (9.0%) mention the fuel cap at all. * In exactly ONE of those 12 does a cap replacement appear as the answer: a 2012 Honda Fit whose dealer wrote that the car "has not have the updated gas cap" and quoted $160 to replace it. The complaint does not confirm the outcome. * In the other 11, the cap was checked, tightened or replaced and the code stayed. Three owners report replacing the cap more than once: a 2015 Toyota Sienna owner is on their fourth cap with a clean smoke test, a 2007 Volkswagen Passat owner replaced it in 2014, 2015 and again, and a 2012 Jeep Patriot owner replaced it and kept the code. READ THE BIAS BEFORE THE NUMBER. This is not evidence that the cap is rarely the cause. It is evidence that the cap is rarely the cause OF A P0456 THAT REACHES A FEDERAL COMPLAINT FORM, which is a different and much narrower claim. A cap that fixes the problem ends the story before anybody files anything. The page states it that way. Other flags recorded per complaint, by whole-word regex on the narrative and checked against the reading: also_names_p1450 .......... 21 of 133 (15.8%) also_names_p0455 ........... 8 of 133 (6.0%) also_names_p0442 ........... 1 of 133 (0.8%) mentions_stall_or_no_start . 40 of 133 (30.1%) mentions_fuel_smell ........ 17 of 133 (12.8%) The P1450 and P0455 counts match the companion-code figures the Check Engine Code Index publishes for P0456 (P1450 21, P0455 8), which is a useful consistency check: the two sets were built independently and agree. STRUCTURED OUTCOME FIELDS. Of the 133, the crash flag is Y in 1 (0.75%), fire in 0, towed in 1 (0.75%). Baselines across all 777,040 complaints in the same window: crash 4.37%, fire 1.82%, towed 5.11%. This is a low-consequence code in the federal record — which sits oddly beside the 40 complaints that mention a stall or a no-start, almost all of them Ford canister-purge-valve cars where the stall, not the leak, is what the recall is about. MILEAGE. Rule: keep MILES where 100 <= miles < 400,000 — the same convention the Check Engine Code Index uses. 65 of 133 complaints state a usable mileage. Median 79,000 miles; middle half 57,226 to 109,500. The Index publishes the same median from the same rule. GEOGRAPHY, AND WHY WE DID NOT DRAW A CONCLUSION FROM IT. California supplies 21 of the 133 complaints (15.8%), the largest single state, ahead of Florida 13 and Texas 11. Hyundai and Kia publish a family of bulletins blaming salt particulate inside the natural vacuum leak detection sensor in "the California coastal areas that are subject to marine layer cloud cover." It is tempting to connect the two. We did not, and the page does not: California is also the largest car-owning state in the file, we have no per-state vehicle denominator, and only 2 of the 133 complaints are Hyundai. A geographic cause needs a geographic denominator, and we do not have one. ------------------------------------------------------------------------------ 4. THE BULLETINS (p0456-bulletins-outside-stellantis.csv) ------------------------------------------------------------------------------ We did NOT recount the bulletins. We took the 269 P0456 rows straight from the published Check Engine Code Index bulletin list, which applies the exclusions documented in its own method (heavy-truck and motorcycle makers, helpline analysis lists, non-US-market documents, Honda dealer messages that repeat a numbered bulletin, entries whose ID has no digits), and normalizes revisions so one bulletin counts once. 269 bulletins naming P0456 192 of them (71.4%) Stellantis 77 everyone else: Ford 19, Hyundai 17, Kia 14, Nissan 9, Mazda 6, VW Group 6, Honda 2, Subaru 2, JLR 1, Toyota 1 THE SOFTWARE SPLIT, which the Index publishes only as a single figure. Of the 194 bulletins its software-detection rule flags, 178 are Stellantis and 16 are everyone else. So the software share is 72% of the whole set but 21% (16 of 77) outside Stellantis. Stellantis publishes large numbers of flash bulletins that list dozens of codes at once; treating that as evidence about what causes P0456 on a Mazda or a Kia would be a mistake, and the page says so. We then read the 77 non-Stellantis documents and grouped repeats — a Nissan and an Infiniti version of the same bulletin, a revision, the same text reissued for more model years — into 44 DISTINCT PROBLEMS: Purge, vent or NVLD valves ....................... 10 Software or calibration ........................... 8 Cap, filler neck or capless filler ................ 6 No leak found, or remedy not stated ............... 6 Lines, hoses and wiring ........................... 5 A procedure to find the leak, no single part ...... 5 Canister and its mounting ......................... 3 Fuel pump and tank hardware ....................... 1 --- 44 Every row of the CSV carries the bulletin IDs exactly as the Index lists them, the models and the latest communication date, and a short quotation from the document text so the grouping can be checked. Judgement calls, stated so they can be disagreed with: * The Hyundai NVLD salt family (7 bulletins) and the Kia air-filter-and-NVLD family (3 bulletins) are counted as two problems, not ten, because each family repeats one finding across model lines. Counting documents instead would make "NVLD" look ten times more common than it is. * Ford's three C1198 wiring documents are one problem; so are Ford's two "software level of the PCM" documents and Mazda's four cap documents. * Six problems are documents that name no repair in their summary text. They are shown as their own category rather than assigned a cause we cannot see. ------------------------------------------------------------------------------ 5. THE REGULATION, READ RATHER THAN SUMMARIZED ------------------------------------------------------------------------------ 13 CCR 1968.2, California's OBD II regulation, read 19 September 2026 at https://www.law.cornell.edu/regulations/california/13-CCR-1968.2 (e)(4.2.2)(C): the OBD II system shall detect a malfunction when "The complete evaporative system contains a leak or leaks that cumulatively are greater than or equal to a leak caused by a 0.020 inch diameter orifice." (B) sets the same requirement at 0.040 inch. The word doing the work is CUMULATIVELY: the threshold is the sum of every leak in the system, not one hole. (e)(4.2.1): "an orifice shall be defined as an O'Keefe Controls Co. precision metal 'Type B' orifice with NPT connections with a diameter of the specified dimension (e.g., part number B-20-SS for a stainless steel 0.020 inch diameter orifice)." The threshold is a purchasable part, not a metaphor. 0.020 inch is 0.508 mm. (d)(3.2.1)(B): the minimum acceptable in-use monitor performance ratio is "0.260 for monitors designed to detect malfunctions identified in section (e)(4.2.2)(C) (i.e., 0.020 inch leak detection)" and "0.520 for monitors designed to detect malfunctions identified in sections (e)(4.2.2)(A) and (B) (i.e., evaporative system purge flow and 0.040 inch leak detection)." The smallest-leak monitor is required to run only half as often as the large-leak monitor. That is a regulatory floor, not a measurement of any one car, but it is the published reason a P0456 takes longer to come back, longer to clear, and longer to set readiness before an inspection. (g)(4.5.4)(F): "an OBD II system with individual evaporative system diagnostics for 0.040 inch and 0.020 inch leaks shall separately report 0.040 inch and 0.020 inch test results." This is why a scan tool that shows Mode 6 can show you the 0.020-inch result and its limit as a number. (e)(4.2.3): on vehicles with fuel tank capacity greater than 25.0 gallons a manufacturer may request relief from the requirement. We did not attempt to determine which vehicles received it. These are California requirements. Federal OBD rules incorporate California's OBD II regulation for most light-duty vehicles, but the page attributes the 0.020-inch and 0.040-inch thresholds to 13 CCR 1968.2 rather than claiming a federal wording we did not read. ------------------------------------------------------------------------------ 6. RECALLS AND INVESTIGATIONS OPENED AND READ ------------------------------------------------------------------------------ NHTSA 18V-735 / Ford 18S32, report received 22 October 2018 — 1,282,596 2012-2018 Focus with the 2.0L GDI or GTDI engine; "The Canister Purge Valve (CPV) may malfunction, possibly causing excessive vacuum in fuel vapor management system," consequence "an engine stall while driving, without warning and without the ability to restart the vehicle." Remedy: reprogram the PCM, check for fault codes, replace the CPV as needed, and inspect and replace the carbon canister, fuel tank and fuel delivery module as necessary. NHTSA 19V-515 / Ford 19S22, 5 July 2019 — 57,488 vehicles "previously recalled under 18V-735" whose PCM "did not have the Powertrain Control Module (PCM) updated as intended." NHTSA 24V-110 / VW 20UF, Audi 20YF, 14 February 2024 — 261,257 2015-2020 Audi A3 and Volkswagen Golf, GTI, Golf Sportwagen and Jetta GLI; "Due to a problem with a suction jet pump seal inside the fuel tank, fuel may flow into the evaporative emissions (EVAP) system and possibly leak out of the charcoal canister." Remedy: replace the suction pump. NHTSA RQ23-007, opened 12 December 2023, closed 1 July 2024 — a recall query into the remedy effectiveness and scope of VW recall 16V-647, which produced 24V-110. Read from the recall flat file and the public recall API on 19 September 2026. WHAT WE LOOKED FOR AND DID NOT FIND * No Part 573 safety recall has the string P0456 in its defect, consequence, remedy or notes text. The recalls above are reached through the component and the narrative, not the code, which is exactly why owners in our set kept finding them by accident. * The Check Engine Code Index flags ZERO of the 269 P0456 bulletins as a warranty extension or emissions campaign, the only code in the index with no coverage document at all among the bulletin set. Hyundai and Kia service campaigns (T4C, T4N, T5A, SA152, SA395, SA397, SA421, SA447, SA498) appear in the bulletin text, but none of them is a warranty extension. * We could not re-price this repair in this pass. The consumer estimator sites that publish these ranges refuse automated access, and we do not work around that. Every dollar range on the page is carried forward from the earlier review of our cost guides and is labelled as an estimate. ------------------------------------------------------------------------------ 7. FILES ------------------------------------------------------------------------------ p0456-complaints-classified.csv One row per complaint, all 133. Columns: seq, odi_report, make, model, model_year, received (YYYYMMDD), state, miles_reported (blank where under 100 or absent), primary_outcome, owner_mentions_the_cap, cap_reported_as_the_answer, also_names_p0455, also_names_p0442, also_names_p1450, mentions_stall_or_no_start, mentions_fuel_smell, crash, fire, towed. No personal data: no VIN, no city, no dealer name, no narrative text. p0456-bulletins-outside-stellantis.csv One row per distinct problem, 44 rows covering the 77 non-Stellantis bulletins. Columns: category, distinct_problem, manufacturer_group, bulletin_ids_as_indexed, models_as_indexed, latest_communication_date, what_the_document_says (short quotation). p0456-by-make-and-model.csv make, model, complaints_naming_p0456, complaints_for_this_make, share_of_133_pct. p0456-model-year-and-report-year.csv year, complaints_on_this_model_year, complaints_received_this_calendar_year. Reuse freely under CC BY 4.0 with the attribution line at the top of this file. Corrections: support@pulscar.io.