Why nearby camera evidence disappears in 72 hours — and what to do about it.

The most valuable evidence in a personal injury claim is usually already gone by the time anyone thinks to look for it. Not because it never existed — but because it expired on a server in a back office somewhere, overwritten by the next day's footage on a 72-hour retention cycle.

Most retail security systems retain footage for 24 to 72 hours. Many traffic cameras run a 30-day buffer; the rest are far shorter. Dashcams in nearby vehicles? Usually a 12-hour loop unless the driver explicitly saves a clip. By the time a claim is filed, intake is completed, and an attorney sends a preservation letter, the camera that would have proven liability has already written over the moment in question.

This is the forensic crisis nobody talks about. Below is what we learned about it — and the engineering response we built into IncidentApp.

The retention cliff

Here's what we found surveying retention practices across the categories of cameras that actually matter at a crash site:

• Retail security (gas stations, convenience stores, restaurants): 24–72 hours. Often shorter at smaller franchises with consumer-grade NVRs.
• Municipal traffic cameras: highly variable. 7–30 days is common, but city budgets and the specific intersection determine retention.
• Highway DOT cameras: usually not recorded at all. Most state DOT cameras are streaming-only — they exist for traffic management dashboards, not evidence.
• Dashcams in nearby vehicles: 6–12 hour loops, sometimes longer with high-capacity SD cards. Drivers must manually preserve clips or the footage is gone by the next morning's commute.
• Doorbell cameras (Ring, Nest, etc.): 30 days if the homeowner has a paid subscription. Otherwise, often live-only with no retention at all.

The pattern is clear: the cameras that capture the most valuable evidence retain it for the shortest time. The cameras that retain footage longest (corporate networks with cloud storage) are usually the hardest to access without legal process.

What this means for personal injury claims

The traditional claim timeline goes something like this:

1. Day 0: Crash happens. Police take a report. Both drivers exchange info.
2. Day 1–3: Injury becomes apparent. Initial medical treatment.
3. Day 7–30: Decision to retain counsel. Initial intake.
4. Day 30–60: Attorney begins discovery, sends preservation letters.

By Day 30, when the preservation letter goes out, the security camera footage from the gas station across the street was overwritten three weeks ago. The dashcam clip from the bystander who pulled over to help? Looped over by the next day. The Ring doorbell from the corner house? Gone unless the homeowner happened to download it.

The economic value of evidence collapses on a half-life of about 36 hours. After 72 hours, the median piece of camera-based evidence is gone forever.

This isn't a discovery problem — it's a capture window problem. By the time the legal system notices the evidence exists, the evidence has already expired.

The engineering response

When we set out to build IncidentApp, this 72-hour cliff was one of the hardest constraints to design around. We needed to do three things at the moment of the crash, before the user has even processed what happened:

1. Map nearby cameras automatically

Within seconds of incident capture, IncidentApp reverse-geocodes the crash location and surfaces every potentially camera-equipped business, intersection, and residence within a ~150 meter radius. We layer in commercial property datasets, traffic camera registries, and known dashcam-friendly route patterns.

The user doesn't need to remember to look around. The app already knows.

2. Generate camera owner outreach

For each surfaced camera, IncidentApp generates a tailored preservation request — phrased as a polite ask, not a legal threat. The user gets a one-tap option to send it: SMS to the property owner if we have a number, otherwise a printed letter or in-person handoff prompt.

The legal effect of a preservation request is limited at this stage (the user isn't an attorney), but practically: most owners cooperate when asked nicely and given clear context. Formal requests via counsel come later if needed.

3. Cryptographically seal what's captured locally

Even if no third-party footage survives, we make sure the user's own evidence is bulletproof. Every photo and document captured through IncidentApp is integrity-hashed at the moment of capture, with GPS and timestamp baked into the metadata. The seal is generated client-side using SHA-256 over the binary plus the metadata blob:

// Pseudo-code — actual implementation runs in Swift
const seal = sha256(
  imageBinary +
  JSON.stringify({ gps, timestamp, deviceId })
);

If the file is ever modified later — even a single pixel — the hash breaks. The packet either matches its seal or it doesn't.

The hidden engineering problem

The hardest part wasn't the camera-mapping or the cryptography. It was something more subtle: the user can barely think.

Anyone who has been in a crash knows the cognitive fog that follows. Adrenaline hijacks attention. Decision-making collapses to the next ten seconds. Designing a UI that asks someone in this state to do anything requires a complete inversion of normal product thinking: every screen has to assume the user is smarter when calm, dumber right now, and may switch back at any moment.

We'll write that one up separately. For now, the design rule that emerged: when the world goes wrong, the app gets simpler.

The takeaway

If you're an attorney working a personal injury claim, the highest-leverage intervention you can make is happening before the case exists — at the scene, in the first hour, while the cameras still hold the moment. The preservation letter you're sending on Day 30 is too late more often than the system admits.

If you're a consumer who's just been in an accident: tap the cameras you can see. Take photos of them. Ask the owner now, while they're still nearby. The IncidentApp flow does this automatically, but the principle is more important than the tool.

And if you're an engineer: the 72-hour retention cliff is one of those structural problems that's been sitting in plain sight for fifteen years, waiting for someone to build an opinionated platform around it. We're glad to be doing it.