PulsCar analysis of NHTSA public data, September 2026
Study: "Car clicking when starting" — owner complaints, starter recalls and no-crank service bulletins
License: CC BY 4.0. Credit "PulsCar analysis of NHTSA public data, September 2026" with a link to
https://pulscar.io/blog/car-clicking-when-starting

SOURCE FILES (all public, downloaded from https://static.nhtsa.gov/odi/ffdd/ in September 2026)
- Complaints: COMPLAINTS_RECEIVED_2015-2019.zip, COMPLAINTS_RECEIVED_2020-2024.zip, COMPLAINTS_RECEIVED_2025-2026.zip
  777,040 vehicle complaints received 1 January 2015 to 8 September 2026 (one row per ODINO).
- Recalls: FLAT_RCL (all campaigns on file), one row per make/model/year per campaign; campaigns counted once.
- Manufacturer communications (TSBs): FLAT_TSBS, one row per make/model/year; counted once per NHTSA ID.
Loaded into SQLite with an FTS5 index on the complaint free text (CDESCR) and on bulletin summaries.

1. COMPLAINTS THAT DESCRIBE A CLICK WITH NO CRANK
Step 1, full-text candidates (1,920 complaints):
  complaints_fts MATCH 'NEAR(click* start*, 12) OR NEAR(click* crank*, 12) OR NEAR(click* key, 12)
  OR NEAR(click* ignition, 12) OR NEAR(click* "turn over", 12) OR NEAR(click* "turns over", 12)'
Step 2, Python filters applied to the text (case-insensitive), keeping 677 complaints:
  a) a no-start or no-crank statement must be present:
     (won't|wouldn't|will not|would not|does not|did not|doesn't|didn't|not|no|failed to|unable to|
      refused to|couldn't|could not|can't|cannot) (up to two words) (start|crank|turn over|turning over|fire up)
     OR \bno[- ]?(start|crank)\b OR \bnot? (starting|cranking)\b OR (just|only|simply) (a )?click(s|ed|ing)?
     OR click(s|ed|ing)? but (won't|wouldn't|will not|does not|no)
  b) at least one "click" mention that is NOT negated ("no click", "not even a click", "or even click",
     "without clicking", "doesn't click" ...), NOT about clicking a button, fob, screen or "click here",
     NOT "clicking (noise) while/when driving, running, accelerating, turning, braking, idling",
     and that sits within 90 characters of a start/crank/turn-over/key/ignition/push-button/solenoid/starter/jump word.
Reading tests (all random samples, seed recorded in the scripts):
  - First loose filter (749 kept): 30 read, about 10 were off topic, nearly all "no click at all" reports.
    That is why negated clicks were excluded.
  - Second filter (661 kept): 30 read, 21 clearly a click with no crank, 5 about something else
    (a horn, turn-signal relay, tailgate, screen, driving noise), 4 cut off or ambiguous.
  - Final filter (677 kept): 100 read in full and categorised (sample100_reading.csv):
    13 off topic, 87 genuine. Treat the 677 as roughly 590 genuine reports; make and model shares
    are not corrected for this because the off-topic rate did not cluster by make in the 100 read.
What was counted (complaints_by_make_model.csv, and the figures in the article):
  - by make: n and n per 10,000 of ALL complaints for that make in the same file (denominator shown).
  - by model: make + model across model years, models with >= 6 complaints; top three model years listed.
  - model year, vehicle age (year received minus model year, model years 1990-2027), mileage
    (MILES field, 0 < miles < 400,000, n = 447), towed/crash/fire flags vs the all-complaint baseline.
  - text features by regex (share of 677): battery replaced/new 107 (15.8%); starter replaced/new/failed 55 (8.1%);
    both 11; a "still/again/continues/did not fix" word in the same complaint after a starter mention 38 of 55
    and after a battery mention 82 of 107; intermittent wording 250 (36.9%); "lights/radio/everything works"
    167 (24.7%); jump mentioned with a start 74 (10.9%); cold/freezing/winter 35 (5.2%); shop could not find
    or duplicate 43 (6.4%); stranded/towed/roadside 197 (29.1%); rapid-click words 128 (18.9%);
    single-click words 48 (7.1%); neutral safety/range switch/shifter 55 (8.1%); ignition switch/steering lock/
    security/immobiliser 78 (11.5%); relay/fuse/TIPM/module/computer 101 (14.9%); alternator 64 (9.5%);
    engine seized/blown/replaced 59 (8.7%).
  - Honda Accord 2013-2015 (looser NEAR query, 109 complaints): 72 mention the starter, 49 the battery,
    43 a push button, 4 "V6", 3 "grind". Dodge Journey 2009-2010 (55): 27 starter, 24 battery,
    15 neutral, 15 "recall", 7 key fob/WIN, 2 TIPM.
Limitations: complaints are self-reported, unverified and not failure rates; owners of some brands file more
than others; a complaint can carry several component categories; regex features count words, not confirmed
repairs; "no diagnosis reported" dominates because people file when nobody has fixed the car.

2. STARTER RECALLS (starter_recalls.csv)
  Rows: every campaign in the recall file whose COMPNAME contains "STARTER" (35 campaigns, all product types).
  vehicle_type assigned by reading make and model: 19 light-vehicle campaigns (1,568,007 vehicles),
  2 motorcycle (24V557, 15V439), 14 heavy truck, bus, RV, fire apparatus or equipment.
  remedy_type from the remedy text: replace the starter 9, inspect and correct 6, relay or relay kit 2,
  add a fuse 1, software update 1. fire_or_overheat_risk = defect or consequence text contains
  fire/thermal/overheat/melt/ignite: 18 of 19 light-vehicle campaigns.
  Vehicle counts are the manufacturer's "potentially affected" population.
  Part 573 reports read in full: 26V438 (RCLRPT-26V438-2276.pdf), 25V636 (RCLRPT-25V636-0319.pdf);
  dealer letter RCMN-17V114-0847.pdf; NHTSA recall API summaries for 17V114 and 24V107.

3. NO-CRANK MANUFACTURER COMMUNICATIONS (no_crank_bulletins.csv)
  tsbs_fts MATCH '"no crank" OR "no-crank" OR "not crank" OR "won''t crank" OR "will not crank" OR
  "does not crank" OR "click no start" OR "click no crank" OR "clicks and does not turn over" OR
  "starter clicks" OR "no start no crank" OR "no crank no start"', then a regex re-check on the summary,
  one row per NHTSA ID: 675 communications dated 2015 to September 2026 (all product types; Freightliner 8).
  remedy_class by regex on the summary, checked in this order: software or reprogramming (132);
  battery drain or discharge (64); wiring, ground, fuse or connection (190); starter or solenoid (35);
  module or relay (41); cold weather (15); other or unclassified (198). A summary is a manufacturer's
  one-paragraph abstract, so the class is indicative; only 31 of 675 summaries use the word "click".
  Bulletins read in full: GM PIT5266C (MC-10112241-9999.pdf), Honda 16-002 (MC-10115802-9999.pdf),
  Honda 23-023 (MC-10234629-0001.pdf).

No personal data: the CSVs carry NHTSA reference numbers, makes, models, years, counts and short
document excerpts only. Scripts: q3.py (filter), stats.py (counts), tsb.py (bulletins), available on request.