People install a dash cam expecting to save the whole drive. What they actually need is about 30 seconds: the ten seconds before impact, the impact itself, and the twenty seconds after where everyone's position tells you who did what. Those 30 seconds are the difference between a claim paid in nine days and a claim argued for four months.
The catch is that the 30 seconds you need are almost always stored in three separate files. Loop recording splits footage into fixed-length segments, and a collision lands wherever it lands — usually straddling a boundary. If you don't know which files matter and how to save them before the loop overwrites them, you end up with a partial story: the crash but not the approach, or the aftermath but not the cause. Here's how to get the full sequence and make sure it survives.
Understanding the file structure is the whole trick. A dual-channel 4K setup writing 3-minute segments produces a constant stream of files, roughly 1.2GB per minute across both channels at high bitrate. When the G-sensor detects the impact signature — a sharp lateral or longitudinal deceleration spike — the camera does three things at once:
Two things go wrong here constantly. First, the pre-buffer is short — 10 seconds on most cameras — so the "before" footage often starts after the other driver's lane change already happened. Second, event files are locked but not permanent; most cameras hold a fixed number of protected files and start deleting the oldest locked file once that count fills. If you don't retrieve the footage within a few weeks of driving, it can age out even though it was "protected."
Pull the card and sort by timestamp. You want a contiguous run, and it should look like this:
| File | What It Contains | Why It Matters |
|---|---|---|
| Segment immediately before | Approach, lane position, your speed | Proves you weren't the cause; captures the other driver's maneuver |
| Event file (front) | Impact plus ~20s after | Shows point of contact and your reaction |
| Event file (rear) | Same window from behind | Independent angle; often catches the plate the front lens missed |
| Next 1–2 segments | Post-impact, both vehicles at rest | Establishes final positions and damage at scene |
Copy all four before doing anything else. Don't review them on the camera's screen, don't trim them, and don't send them through a messaging app — every one of those steps re-encodes or discards data. Copy the raw files to two separate places: a laptop and a USB drive, or your laptop and cloud storage.
Insurers and, if it comes to it, courts care about integrity, not drama. A file that's been through three apps, cropped, and re-saved looks like it's been through three apps, cropped, and re-saved. Preserve the original filename, the original timestamp, and the original file size.
The practical version of a chain of custody for a normal driver is simple and takes four minutes:
2026-10-27_main-st-rear-end.That folder is what you send. If a claims adjuster asks for a clip, send copies of the originals, not a trimmed version. If someone asks why you preserved it the way you did, "these are the files exactly as the camera wrote them, here's the card they came off" ends the conversation.
Here's the failure nobody thinks about until it's too late: the timestamp has to be right. A dash cam with no GPS or with a battery that's drifted will default its clock on every cold start. Footage stamped 2023 in a 2026 claim is a gift to the other side's adjuster, who can argue the recording isn't from the incident at all.
Check your camera's date and time once a month, and immediately after any period where it sat unplugged. Cameras with a GPS module sync the clock from satellite time automatically and rarely have this problem. Cameras without one drift, especially in the first week after the backup battery dies. If your camera lets you set the timezone, set it — a clip stamped an hour off is a small crack that a determined adjuster will widen.
Sometimes you get out of the car, look at the damage, and realize the G-sensor never triggered. It happens more than people admit. Low-speed parking lot impacts are the classic case: a 3-mph bump produces a deceleration curve the sensor's threshold ignores, and nothing gets locked. The footage still existed in the loop — it just wasn't protected.
In that situation you have a narrow window. The loop buffer holds several hours of driving, so if you pull the card within a day or two and haven't driven much, the clip is still there in the regular folder. Copy every file whose timestamp brackets the incident and review them on a computer. This is also the argument for setting the G-sensor to medium rather than high sensitivity — high sensitivity fills your locked-file quota with potholes and driveway ramps, and once that quota is full the camera starts deleting real events. Medium catches actual impacts and ignores curbs.
While you're in the menu, make sure parking mode is writing event files separately from continuous parking footage. Cameras that record parking mode as one long low-bitrate file make extraction painful; cameras that split it into event-triggered clips make this whole process take two minutes.
A short list, learned the hard way by people who lost usable evidence:
None of this is complicated, but all of it has to happen in the first 24 hours, when you're rattled and standing in a parking lot. The people who handle this well are the ones who rehearsed it once, on a calm afternoon, before they needed to. Pull the card, find a segment boundary, copy four files, name the folder. Five minutes of practice turns a bad day into a routine task.
If you're setting this up for the first time, the two things worth configuring properly are G-sensor sensitivity at medium and a hardwire kit so the camera is awake in parking lots — both are covered in the same setup flow as the current RedTiger models, and both matter more than resolution for this particular job.
← Back to Blog