PULSCAR ANALYSIS OF NHTSA PUBLIC DATA - WHAT OWNERS MEAN BY "SPUTTERING" PulsCar analysis of NHTSA public data, September 2026. License: CC BY 4.0. Files: sputter-timing-vs-system.csv, sputter-complaints-by-make.csv, sputter-top-models.csv, sputter-context-and-outcomes.csv. THE QUESTION This page's whole thesis is that WHEN an engine sputters names the failing system. We tested it against the only large public record of owners using that word: the federal complaint file. The answer is partly yes and partly no, and the article publishes both halves. SOURCE FILES 1. NHTSA ODI complaint flat file (COMPLAINTS_RECEIVED_*.txt), https://static.nhtsa.gov/odi/ffdd/cmpl/ - field dictionary https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt 777,040 complaints received 1 January 2015 - 8 September 2026, one row per ODINO. 2. NHTSA manufacturer communications flat file - 187,827 distinct document ids. 3. NHTSA recall flat file, counted as DISTINCT campaign. 4. NHTSA recalls API for the two campaigns quoted; both returned HTTP 200 on 18 September 2026. Loaded into local SQLite with FTS5 indexes. Indexed 17 September 2026, queried 18 September 2026. LETTER CASE - THE TRAP THAT DELETES HALF THE FILE Complaint narratives are stored in CAPITALS through 2020 and in mixed case from 2021: the share of complaints whose first 200 characters are entirely uppercase is 100.0% for 2015-2020, 42.5% for 2021, then 1.6%, 1.1%, 0.8%, 0.8% and 0.6% for 2022-2026. FTS5 is case-insensitive, SQLite LIKE is case-insensitive, and every Python regular expression here was compiled with re.IGNORECASE. No uppercase-only pattern and no GLOB was used anywhere. An uppercase pattern would have returned a clean-looking answer built almost entirely from pre-2021 complaints. QUERY 1 - THE SPUTTER SET (denominator 4,523) select c.* from complaints c join complaints_fts f on f.rowid = c.rowid where complaints_fts match 'sputter OR sputters OR sputtering OR sputtered OR sputterred OR stutter OR stutters OR stuttering OR stuttered'; Result: 4,523 complaints. This is a WORD query on purpose: the page is about the symptom people describe with that word, so the owner's vocabulary is the right filter. It is deliberately different from the sets behind our other pages - rough idle, jerking on acceleration and hesitation on acceleration each have their own query and their own folder, and no figure is shared between them. SAMPLE READING 30 random complaints (random.seed(13)) were read in context. All 30 describe the owner's own vehicle sputtering, stuttering or breaking up - a 0-in-30 false positive rate for "the word is being used about this car's behavior". A sample of 30 cannot distinguish 0% from roughly 10%. What the reading DID show is the finding below: a large share of these narratives are about a transmission, not an engine. Owners use "sputter" and "stutter" for a shudder, a slip and a harsh shift as readily as for a misfire. QUERY 2 - TIMING (window) x SYSTEM (whole text), sputter-timing-vs-system.csv TIMING is matched inside a +/-100 character window around every occurrence of /sputter\w*|stutter\w*/i, not across the whole complaint, so that a complaint that merely mentions a highway somewhere else does not land in the highway row. SYSTEM is matched across the whole complaint text, because owners usually name the part in a different sentence from the symptom. Both families are NON-EXCLUSIVE: a complaint that says it sputters when accelerating and at idle counts in both timing rows, and one naming both a coil and the transmission counts in both system columns. Rows and columns therefore add to more than 100%. Timing patterns, verbatim: under acceleration acceler\w*|when I (press|hit|floor|step on)|floor(ing|ed)? it| throttle|gas pedal|pulling away|merg\w*|passing|take off| taking off|uphill|up a hill|going up at highway or steady speed highway|freeway|interstate|\b[4-8]\d\s?mph|cruis\w*| steady speed|high speed|on the road at at idle, stopped, at a stop \bidl\w*|at a (stop|red light|traffic light)|stopped|stop sign| in park\b|neutral\b|sitting still when starting or cranking start\w* (it|the (car|engine|vehicle))|crank\w*|turn(ing)? the key| ignition|cold start|won'?t start|wouldn'?t start in rain or wet weather \brain\w*|\bwet\b|humid\w*|car wash|puddle|damp|snow\w*|moist an RPM ceiling (rpm|rpms)[^.]{0,40}(ceiling|limit|wall|cap|will ?not|won'?t| cannot|can'?t|stuck|max) | (will ?not|won'?t|cannot|can'?t) [^.]{0,30}(rev|go (above|over|past)|accelerate past|exceed) [^.]{0,20}(\d|rpm) when climbing or under load under load|\bload\b|towing|uphill|up a hill|incline|hill System patterns, verbatim: fuel pump, filter, injectors or fuel pressure fuel pump|fuel filter|fuel injector|injectors?|fuel pressure|fuel delivery| in-?tank pump|\bp0087\b|\bp0089\b|\bp01(7[1-4])\b spark plugs, coils, wires or a named misfire spark plug|ignition coil|\bcoils?\b|plug wires?|distributor|misfir\w*| \bp03(0[0-9]|1[0-2])\b MAF, oxygen sensor or another sensor mass air ?flow|\bmaf\b|oxygen sensor|\bo2 sensor|throttle position| crank(shaft)? (position )?sensor|cam(shaft)? (position )?sensor|\bsensor\b transmission, CVT or torque converter transmission|\bcvt\b|torque converter|gearbox|shift\w*|\bgears?\b|clutch catalytic converter or exhaust restriction catalytic converter|\bcat(alyst)?\b|exhaust restrict|backpressure|\bp0420\b|\bp0430\b vacuum leak, intake boot, EGR or PCV vacuum leak|intake (boot|hose|manifold)|\begr\b|\bpcv\b|unmetered air|air leak the engine itself engine (was )?(replac|rebuil)|new engine|head gasket|coolant int|valve|piston| timing chain|\bcompression\b a recall, warranty extension or campaign \brecall\b|warranty extension|campaign HEADLINE RESULTS (all published in the CSV) Timing: under acceleration 1,342 (29.7%); at highway or steady speed 835 (18.5%); at idle or stopped 628 (13.9%); when starting or cranking 368 (8.1%); under load or climbing 161 (3.6%); in rain or wet weather 44 (1.0%); an RPM ceiling 13 (0.3%). System across the whole set: transmission/CVT/torque converter 1,408 (31.1%); engine itself 516 (11.4%); spark plugs/coils/misfire 462 (10.2%); fuel pump/filter/injectors/pressure 458 (10.1%); MAF/oxygen/other sensor 252 (5.6%); catalytic converter or exhaust restriction 51 (1.1%); vacuum/intake/EGR/PCV 48 (1.1%). A recall, extension or campaign is named in 1,020 (22.6%). Within "under acceleration", transmission words rise to 39.3% and within "under load or climbing" to 41.0%, while fuel words do NOT rise (8.9% and 8.7% against 10.1% overall) and ignition words rise only under load (13.7%). WHAT THESE NUMBERS ARE AND ARE NOT. They are counts of WORDS in owner narratives. A complaint lands in the transmission column because the owner (or a shop they quote) used a transmission word - which may be the system that failed, the system they suspected, or the system somebody replaced without curing anything. They are not diagnoses, not causes and not a distribution of what actually breaks. The article states this next to every figure and never claims a cause share. The RPM-ceiling row (13 complaints) and the rain row (44) are published as SMALL: no percentage is built on either, and the article says only that the pattern is rare in this file, not that it is rare in the world. Cars with plug wires and a distributor cap - the hardware the wet-weather failure needs - are mostly older than the vehicles this file is dominated by. QUERY 3 - CONTEXT AND OUTCOMES (sputter-context-and-outcomes.csv) Structured NHTSA fields: towed 121 (2.68%), crash 44 (0.97%), fire 51 (1.13%) against whole-file baselines of 5.11%, 4.37% and 1.82%. 35 injuries, no deaths. Mileage, 0 < miles < 400,000, usable in 2,506 of 4,523: median 70,000, quartiles 34,000 and 102,910; 652 (26.0%) at or under 36,000. Owner-recorded speed present in 2,920: median 40 mph. Word flags: a stall or loss of power is mentioned in 1,720 (38.03%); a warning light in 1,439 (31.82%); the standard "a shop could not find it" phrase list in 71 (1.57%) against 0.94% file-wide; a part replaced with the sputter continuing in 59 (1.30%), which is a floor and not a fix rate. MAKES AND MODELS Counts of complaints, never failure rates: this file has no registration or sales denominator, so a nameplate with more vehicles on the road produces more complaints by arithmetic. Ford 1,225 (27.1%), Chevrolet 397, Hyundai 371, Jeep 336, Kia 303. Ford Focus 372 and Fiesta 73 dominate the top of the model table. The per-model CSV carries a "percent naming transmission" column precisely so nobody reads those rows as engine-fault counts. DOCUMENTS QUOTED (read in the NHTSA flat files or the recalls API, 18 September 2026) GM PIP5446 / PIP5446A - "Harsh Shift 1/2 and 2/3 and Sputter Applying Throttle Between 0 and 30%, Rough idle", DTCs P0101, P0171, P0172, P0174, P0175, cause given as cast flashing near the mass air flow sensor. Chrysler/Jeep 11-001-15 - "Sputter Type Noise Emitted From Exhaust ... inspecting and replacing the resonator if required." Mazda SA-024-20 - sputtering, surging or poor driveability with stored DTCs; inspect the MAP sensor connection for corroded terminals. NHTSA recall 11V303000 (Volvo, 1 June 2011, 8,179 vehicles) - fuel pump software; "there may be engine hesitation and possible engine stall even though the fuel gauge indicates up to 1/4 of the fuel is left in the tank." NHTSA recall 18V169000 (Ford, 12 March 2018, 5,361 vehicles) - 2015-2016 Focus and 2013-2014 Fusion manual-transmission clutch fracture; remedy includes "software that detects and prevents prolonged clutch slip." LIMITATIONS Complaints are unverified owner reports, filed by people who were frightened, stranded or out of pocket. They are not a sample of sputtering cars and not a sample of repairs. Every share above describes complained-about sputtering. Nothing in this folder contains a VIN, a city, a dealer name or free text beyond short snippets. CONTACT: support@pulscar.io