PULSCAR ANALYSIS OF NHTSA PUBLIC DATA — FUEL PUMP FAILURE, RECALLS AND REPAIR COST Published September 2026. License: CC BY 4.0 (credit "PulsCar analysis of NHTSA public data, September 2026"). ------------------------------------------------------------------------------- 1. SOURCE FILES ------------------------------------------------------------------------------- NHTSA public ODI flat files, downloaded 17 September 2026 from https://static.nhtsa.gov/odi/ffdd/ (field definitions: cmpl/CMPL.txt, rcl/RCL.txt, ftsb/FLAT_TSBS.txt), loaded into a local SQLite database. complaints 777,040 rows, received 1 January 2015 - 8 September 2026 recalls one row per make/model/year per campaign (campaigns counted DISTINCT) tsbs manufacturer communications Owner complaints are self-reported, unverified and have no denominator. Nothing here is a failure rate, and no "most/least reliable" claim can be made from it. ------------------------------------------------------------------------------- 2. THE COMPLAINT SET (1,656 complaints) ------------------------------------------------------------------------------- Query: a) NHTSA's own component tag contains FUEL PUMP, OR the tag contains FUEL SYSTEM and the narrative contains "FUEL PUMP" -> 1,754 b) AND the narrative contains a failure cue: STALL, WILL/WOULD NOT START, NO START, SPUTTER, HESITAT, LOSS OF POWER, LOST POWER, SHUT OFF/DOWN, DIED/DIES, FAIL, REPLACE, CRANK -> 1,656 Headline figures on those 1,656: towed 639 38.6% (baseline: 39,739 of 777,040 = 5.1%) crash reported 6 0.4% (baseline 4.4%) fire 18 1.1% (baseline 1.8%) stall or loss of motive power described 784 47.3% no-start / failed-to-restart described 340 20.5% mention a recall or a campaign number 899 54.3% say the remedy part was not available 157 9.5% say their VIN was not included in the recall 271 16.4% name a high-pressure pump (HPFP / "high pressure") 304 18.4% name a fuel pump control, driver or power module 202 12.2% name a relay or a fuse 115 6.9% mention fuel pressure or PSI at all 31 1.9% Failure mileage: 1,480 report a usable figure (drop 0 and >= 400,000). Median 82,698; 25th percentile 50,000; 75th percentile 121,000; 60.7% under 100,000. The 304 high-pressure-pump complaints have a median of 101,000 miles and a towed rate of 51.3%. Complaint volume by year received: 2015: 61, 2016: 73, 2017: 42, 2018: 87, 2019: 57, 2020: 77, 2021: 104, 2022: 154, 2023: 226, 2024: 266, 2025: 334, 2026 (to 8 Sep): 175. Honda and Acura complaints in this set by year received: 2021: 10, 2022: 8, 2023: 17, 2024: 73, 2025: 54, 2026: 24 — the jump follows recall 23V858 (December 2023), which is a reporting effect as much as a failure effect: recall letters teach owners that NHTSA exists. By make and model: fuel-pump-complaints-by-make.csv (makes with >= 8 complaints) and fuel-pump-complaints-by-model.csv (models with >= 8), each with the towed share, median failure mileage, top model years, and how many of that model's complaints name a high-pressure pump. ------------------------------------------------------------------------------- 3. WHAT OWNERS SAY THEY PAID (199 figures, 183 complaints) ------------------------------------------------------------------------------- Extracted from a wider pool than section 2: all 1,451 complaints anywhere in the file whose narrative contains "FUEL PUMP" and a dollar sign. Regex \$\s?([0-9][0-9,]{2,6})(\.\d{1,2})? , and an amount was kept only when, within 70 characters on either side: - the window names the part: FUEL PUMP | PUMP ASSEMBLY | FUEL PUMP MODULE - the window contains a cost verb: REPLAC | NEW FUEL PUMP | INSTALL | REPAIR | FIX | QUOT | ESTIMAT | COST | CHARGE | PAID | PAY | BILL - the window does not name a different repair: TRANSMISSION, ENGINE REPLAC, NEW ENGINE, CATALYTIC, RADIATOR, TIRES, BATTERY, TOW, THROTTLE BODY, WIN MODULE, FUSE BLOCK, MAIN SEAL, FUEL LINE, SPARK PLUG, BELT, INJECTOR, GAS TANK, FUEL TANK, AIR FILTER, SENDING UNIT, ALTERNATOR, STARTER, COMPRESSOR, TUNE UP, or a second amount ("and $...") - the window is not about the car's value, a loan, a lease, a deductible, a rental, a refund, a settlement or an MSRP - the amount is between $150 and $4,000 - the vehicle is not a motorcycle or scooter (model matching ^(F|R|K|S|G)\d{3,4} or HARLEY, GOLD WING, CAN-AM, KAWASAKI, YAMAHA, SUZUKI, DUCATI, TRIUMPH, INDIAN, VICTORY, MOTORCYCLE) Result: 199 amounts from 183 complaints. under $500 30 15.1% $500-$799 49 24.6% $800-$1,199 59 29.6% $1,200-$1,799 35 17.6% $1,800 and up 26 13.1% median $900 | 25th percentile $600 | 75th percentile $1,366 | range $175-$3,800 median odometer of the cars behind those figures: 77,000 miles SAMPLE READING. We drew 30 of the kept amounts at random and read each in full. 25 of 30 were a genuine fuel-pump repair or quote figure. The five that were not: two motorcycles (which is why the motorcycle filter above exists and was added after this reading), one where the shop found gas lines rather than the pump, one where the amount was for a Totally Integrated Power Module, and one where the figure belonged to an earlier unrelated repair. Observed false-positive rate 5/30 = 17% before the motorcycle filter, 3/30 = 10% after it. An earlier, looser version of the query (120-character window, no parts exclusion list) was read first and rejected: it was pulling in throttle bodies, fuse blocks, rear main seals and multi-part totals. LIMITATIONS: these are owner-reported numbers, not invoices we have seen; they mix quotes with amounts paid, parts-only with parts-and-labor, dealer with independent; they span 2015-2026 and are not inflation-adjusted; and people who write to a federal agency are not a random sample of people who buy fuel pumps. Treat the distribution as a reality check on a quote, not as a national average. Published as fuel-pump-repair-quotes.csv (complaint id, make, model, model year, failure mileage, amount). No free text, no VIN, no dealer name, no city. ------------------------------------------------------------------------------- 4. 110 COMPLAINTS READ IN FULL — WHAT IT TURNED OUT TO BE ------------------------------------------------------------------------------- From the 1,614 complaints in the set whose narrative contains an outcome cue (DEALER, SHOP, MECHANIC, TECHNICIAN, REPLACED, DIAGNOS, TOLD ME, FOUND, RECALL) we drew 110 at random (Python random.seed(2026)) and read every one, assigning a single primary outcome: in-tank fuel pump or pump module named 22 high-pressure (direct-injection or diesel) pump named 21 a recall is the story — part unavailable, VIN excluded, repair delayed, or the owner paid before the campaign 19 no diagnosis stated, or the shop never found one 17 an electrical part, not a pump — fuel pump control/driver/power module, relay, fuse, melted fuse box, TIPM 14 another fuel component — tank, filler neck, lines, injectors, EVAP canister, leaks 12 off target — steering, throttle body, a motorcycle, a gas-station nozzle: not a vehicle fuel pump problem at all 5 Derived figures used in the article: - 14 of 110 (13%) ended at an electrical part rather than a pump. - 21 versus 22: high-pressure and in-tank pumps appear in nearly equal numbers, which is why "fuel pump" is an ambiguous quote on a direct-injection car. - 19 of 110 are recall-process complaints rather than diagnosis complaints. - Off-target rate of the set as read: 5/110 = 4.5%. ------------------------------------------------------------------------------- 5. RECALLS (164 campaigns) ------------------------------------------------------------------------------- recalls WHERE component LIKE '%FUEL PUMP%', campaigns counted DISTINCT: 164 campaigns, 19,727,625 vehicles or units potentially affected 137 campaigns and 18,603,212 units reported since 1 January 2015 11 campaigns (1,912,636 units) have a software remedy; most of the rest replace or inspect the pump 13 campaigns (5,110,575 units) have a defect description that names the pump IMPELLER. That count UNDERSTATES the impeller family, because Toyota's two campaigns summarize the same Denso defect without using the word in the flat file; their Part 573 reports do use it (see below). The campaigns read in full for this article: 20V-012 (Toyota, 13 January 2020, 1,830,752 vehicles) and 20V-682 (Toyota, 4 November 2020, 1,517,721 in the Part 573 report / 1,525,742 in the flat file). Both: "this recall applies to certain vehicles with specific fuel pumps supplied by Denso, containing impellers produced during specific periods." The defect description in 20V-012: impellers "manufactured with lower density" which, if also of a lower-surface-strength type or "exposed to production solvent drying for longer periods of time", crack, absorb fuel, deform, and interfere with the pump body until the pump stops. Toyota's estimated percentage of vehicles actually containing the defect: "NR" — not reported, with a note that it cannot be estimated. The remedy section instructs owners who already paid for this repair to seek reimbursement under Toyota's General Reimbursement Plan. Documents: RCLRPT-20V012-2152.PDF, RCLRPT-20V682-3280.PDF. 20E-026 (24 April 2020, 2,156,057 units) and 20E-085 (17 November 2020, 2,483,801 units), filed by DENSO International America as equipment recalls for the pumps themselves. Equipment campaigns overlap the vehicle campaigns above; the two sets must not be added together. 23V-858 (Honda, 18 December 2023, 2,490,460 vehicles, 2013-2023 Honda and Acura models). Part 573: "The fuel pump impeller was improperly moulded, resulting in low density impellers... Over time, the low-density impeller can deform and interfere with the fuel pump body, rendering the fuel pump inoperative." Estimated percentage with the defect: 1%. Document: RCLRPT-23V858-1287.PDF. 25V-455 (Ford, 7 July 2025, 850,318 vehicles, 2021-2023 Bronco, Explorer, Aviator, Super Duty, F-150, Mustang, Navigator, Expedition): "The low-pressure fuel pump may fail, which can result in an engine stall while driving." The remedy splits by model: replace the low-pressure pump on Super Duty trucks, update the powertrain control module software on everything else. Document: RCLRPT-25V455-7716.pdf. All four Part 573 reports were downloaded from static.nhtsa.gov/odi/rcl/ and read. Caveat: "potentially affected" is a population, not a count of failures; the same vehicle can appear in more than one campaign; and equipment (E) campaigns and vehicle (V) campaigns describe overlapping populations. ------------------------------------------------------------------------------- 6. WHAT WE DID NOT DO ------------------------------------------------------------------------------- - No normalization by vehicles on the road. No reliability ranking. - No price re-verification: the repair-price ranges in the article are prior published estimates and are labelled as such. Consumer estimator sites block automated access and we do not work around that. The only September 2026 price evidence on the page is the 199 owner-reported figures in section 3. - No attempt to distinguish OEM from aftermarket pumps in the complaint text. ------------------------------------------------------------------------------ CASE SENSITIVITY — SECTION ADDED 19 SEPTEMBER 2026 ------------------------------------------------------------------------------ NHTSA stores complaint narratives in ALL CAPS for records filed through roughly 2020 and in mixed case for records filed from 2021 on. A case-SENSITIVE search of the narrative field loses about 92% of the complaints filed after 2020, so this is the single largest way an analysis of this database can go wrong while still looking finished. Every query behind this file is case-insensitive, in two layers: * the complaint set is built with an FTS5 MATCH full-text search, which folds case before matching, so it cannot miss a mixed-case record; and * where the narrative is then tested with patterns, the narrative is upper-cased and whitespace-normalized first, so the patterns compare upper against upper. No query in this work uses GLOB or any case-sensitive comparison. This was not taken on trust. On 19 September 2026 every published folder was re-checked against the same database. Re-run here: step (a) returns 1,754 complaints case-insensitively — exactly the figure published above — against 834 the same query would return case-sensitively. The terms are written in capitals in section 2 as a notation for readability; that is the write-up's convention, not the query's.