PULSCAR ANALYSIS OF NHTSA PUBLIC DATA — TRANSMISSION FLUID LEAKS PulsCar analysis of NHTSA public data, September 2026. License: CC BY 4.0. Files in this folder may be reused with attribution to PulsCar (pulscar.io). -------------------------------------------------------------------------- 1. SOURCE FILES -------------------------------------------------------------------------- NHTSA public flat files downloaded from static.nhtsa.gov and loaded into a local SQLite database: Complaints FLAT_CMPL.txt 777,040 owner complaints (ODINO), received 1 January 2015 - 8 September 2026. Field definitions: https://static.nhtsa.gov/odi/ffdd/cmpl/CMPL.txt Recalls FLAT_RCL.txt one row per make/model/model year per campaign; campaigns counted DISTINCT on campaign number. Field definitions: https://static.nhtsa.gov/odi/ffdd/rcl/RCL.txt Communications FLAT_TSBS.txt manufacturer communications; counted DISTINCT on the NHTSA document id. Field definitions: https://static.nhtsa.gov/odi/ffdd/tsbs/TSBS.txt A complaint is an owner's voluntary account. It is not a verified repair, not a failure rate, and it has no denominator. A make with more complaints may simply have sold more vehicles. Nothing in this folder ranks reliability, and none of it is a per-vehicle risk. This file is also strongly selected in one specific way that matters here: people write to a federal safety agency after something frightening or expensive happened. Routine leak repairs — a pan gasket replaced on a Saturday for $250 — almost never appear. Everything below therefore describes the bad end of the distribution and should never be read as "what usually happens when a transmission leaks." -------------------------------------------------------------------------- 2. THE COMPLAINT QUERY (3,391 -> 2,441 -> 770 -> 747) -------------------------------------------------------------------------- Step 1 - FTS5 search for (transmission OR transaxle OR transmissions) AND (leak OR leaks OR leaking OR leaked OR puddle OR puddles OR dripping OR drip OR drips OR seep OR seeping OR seepage) -> 3,391 complaints. Step 2 - mere co-occurrence within 100 characters kept 2,441. Reading 24 of those at random found roughly a quarter were about something else entirely: brake fluid from rusted lines, a fuel leak near the transmission, an air conditioning leak, a leaking roof. Co-occurrence was abandoned. Step 3 - explicit grammatical constructions. The narrative had to contain one of these patterns (case-insensitive), where T = transmission | transaxle | trans | tranny | gearbox, L = leak* | drip* | seep* | puddl* | pour* | weep*, F = fluid | oil | atf: T [F] [is|was|has been|started|began|kept|keeps|had been] L L [up to 3 words] (from|out of|at|in|on|under|near|around) [the|my|his|her] T T F L F L [up to 3 words] T T (pan|case|housing|seal|gasket|cooler line|cooler|line) [up to 2 words] L L [up to 2 words] T (pan|seal|gasket|cooler|line) -> 770 complaints. Step 4 - negation and wrong-fluid filters applied to each match window (70 characters either side). A match was rejected if the window contained "no leak(s)", "no sign/evidence of a leak", "not leaking", or "could not find a leak", or if the fluid named immediately before the leak verb was gas, gasoline, fuel, coolant, antifreeze, brake fluid or power steering fluid. A complaint was kept if ANY match survived. This removed 12 rows for negation and 11 for the wrong fluid -> 747 complaints. Sample reading of the final set: 30 random complaints read in full. 26 of 30 described a leak of transmission or transaxle fluid. 4 of 30 were false positives — engine oil dripping from or past the transmission (two), a rear-differential leak on an all-wheel-drive vehicle, and one where the leak could not be attributed to any named component. Measured false-positive rate: 13.3% (4/30). Every count below carries it. -------------------------------------------------------------------------- 3. MILEAGE -------------------------------------------------------------------------- Owner-entered failure mileage, with values <= 0 or >= 400,000 dropped. 485 of 747 complaints (64.9%) carry a usable mileage. Whole set: median 82,000 miles, first quartile 50,000, third quartile 115,381. 19.8% at or under 36,000 miles; 32.4% at or under 60,000 — the two mileages at which most basic and powertrain warranties expire. -------------------------------------------------------------------------- 4. CONTEXT SUBSETS (leak-complaints-by-context.csv) -------------------------------------------------------------------------- Regexes over the narrative. Subsets overlap and do not sum to the total. cooler / lines cooler line | cooler hose | transmission cooler | cooler fitting | cooler tube | radiator cooler | aux* cooler coolant crossover coolant|antifreeze|radiator [<=3 words] into the trans | strawberry | milkshake | cross-contamin* | coolant in the trans | contaminat* (the) transmission seal seal | seals | main seal | pump seal | axle seal | shaft seal pan transmission pan | pan gasket | trans pan cracked case crack* [<=3 words] (case|housing|transmission) | transmission (case|housing) (was) crack failed/replaced transmission [<=3 words] (failed|failure|replac*| rebuil*) | new transmission | replace the transmission | rebuild the transmission slipping slipp* | jerk* | shudder* | would(n't) shift | hard/harsh shift | delayed engagement | flar* lost power lost power | stalled | shut down | shut off | would(n't) move | limp mode | limp home warning shown warning light | check engine | service now | insufficient transmission fluid | warning message | wrench light no warning no warning | without warning | there were no warning recall mentioned recall* | campaign number coverage denied not covered | denied | refus* | vin was not included | not included in the recall | out of warranty Baseline row: the towed, fire and crash shares of all 777,040 complaints (5.11% towed, 1.82% fire, 4.37% crash). Counting note: the "failed/replaced" subset counts complaints in which the same narrative describes both a leak and a transmission failure or replacement. It does NOT establish that the leak caused the failure, and it cannot: many of these owners describe a failure first and a leak found afterward. It measures how often the two appear together in a file written by people whose day went badly, nothing more. -------------------------------------------------------------------------- 5. MAKE AND MODEL (leak-complaints-by-make-and-model.csv) -------------------------------------------------------------------------- Rows with fewer than 8 complaints are omitted. "level" is make or model. These are report counts. They are not failure rates and not a ranking. The coolant-crossover subset is the one concentration worth naming: 31 of its 32 complaints are Nissan, and all of those are Pathfinder, Xterra, Frontier or Titan. That is a known radiator-internal-cooler design, and its concentration in the file is the reason the article names it. -------------------------------------------------------------------------- 6. RECALLS (transmission-fluid-leak-recalls.csv) -------------------------------------------------------------------------- Query: recall rows whose defect text contains "transmission fluid", "transmission oil" or "transaxle fluid" AND contains "leak" or "seep". Counted DISTINCT on campaign: 31 campaigns, 506,451 vehicles potentially affected in total. fire_named_in_consequence = the consequence text contains "fire": 15 of the 31 campaigns, covering 423,850 of the 506,451 vehicles. 20 campaigns name a crash risk (the two overlap). These are the manufacturer's and agency's own words for why the leak matters. The count includes non-passenger vehicles (a Freightliner truck campaign, a Honda scooter campaign); they are visible in the makes column and were left in rather than silently filtered. -------------------------------------------------------------------------- 7. WHAT OWNERS SAID THEY PAID (owner-reported-amounts.csv) -------------------------------------------------------------------------- Query: within the 747-complaint set, the first dollar figure between $100 and $15,000 whose 400-character window also names a transmission, transaxle or "trans". One figure per complaint. Result: 83 figures. Published columns: odino, make, model, model year, failure miles and amount. The matched narrative is not republished; the ODI number is given so any row can be looked up in NHTSA's own file. Distribution: median $4,000; first quartile $1,600; third quartile $6,000. 13.3% under $600, 9.6% $600-$1,500, 14.5% $1,500-$3,000, 37.3% $3,000-$6,000, 25.3% $6,000 and above. Sample reading: 20 random rows read in full. 16 attached the figure to transmission work. 4 did not — an air-conditioning repair, an unrelated $100 item, a bundled fuel-tank repair, and one that could not be tied to any named repair. Measured precision 80% (16/20). This is a distribution of what frightened people wrote down. It is not a price for fixing a leak, and the article says so where it is used. The selection described in section 1 applies to it twice over: a dollar figure appears in a complaint most often when the figure was shocking. -------------------------------------------------------------------------- 8. MANUFACTURER COMMUNICATIONS -------------------------------------------------------------------------- Query: FTS5 on the bulletin summary for ("transmission fluid" OR "transaxle fluid" OR "transmission oil") AND (leak OR leaking OR leaks OR seep OR seepage), counted DISTINCT on the NHTSA document id: 250 documents. No CSV is published for this query; the four bulletins cited in the article are linked individually to their own NHTSA PDF. -------------------------------------------------------------------------- 9. LIMITATIONS, IN ONE PLACE -------------------------------------------------------------------------- - Complaints are self-reported and unverified. - There is no denominator. Nothing here is a failure rate. - The file is selected toward expensive and frightening outcomes; routine leak repairs are almost absent from it. - Keyword sets carry the false-positive rates measured in sections 2 and 7. - 35.1% of the set has no mileage; medians describe the rest. - Subsets overlap and do not sum. - Co-occurrence of a leak and a failure in one narrative is not causation. - Recall "potentially affected" counts are population estimates, not failures. Questions or a correction: support@pulscar.io