One Second of Snapchat Video, and a Mile of GPS Track. None of It Told the Story Alone.
I read timestamps for a living. A fictional case turns on one device with one clock. Easy. A real case turns on four devices, four clocks, and four versions that don’t quite line up (not even your own phone and your own smartwatch can agree on what time it is, and they live in the same pocket). That’s exactly what happened in the trial that turned digital timeline reconstruction into a centerpiece of an American courtroom: State v. Alex Murdaugh, Colleton County Court of General Sessions, South Carolina, 2023.
What actually happened
On the evening of June 7, 2021, at the family’s Moselle property in South Carolina, Maggie Murdaugh and her son Paul were killed. Alex Murdaugh, the husband and father, gave his own account of that evening: where he was, when he left, when he came back. Prosecutors took that account apart using several independent digital sources: a short Snapchat video Paul recorded near the dog kennels about five minutes before the killings, step-count and orientation data pulled from Alex’s own phone (the phone had, in effect, been keeping notes the whole time; it just didn’t know yet that anyone would ask to see them), and GPS data from the vehicle he drove that night. A SLED agent testified from a compiled timeline document running dozens of pages of GPS points, phone calls, texts and orientation data (reported page counts vary by outlet; what matters is the volume, not the exact number), the kind of exhibit that makes a jury grateful someone thought to number the tabs. On March 2, 2023, Murdaugh was convicted, and the next day he was sentenced to two consecutive life sentences without parole.
That’s a familiar story: digital evidence “caught” a defendant in a lie. The more interesting professional question is different: how do you actually build a timeline out of sources that don’t speak the same clock language?
Every source has its own clock, and its own lie
This is the most common mistake I see from people who haven’t worked with multi-source evidence: treating every data feed as if it were one continuous, second-accurate report. It isn’t.
A Snapchat video carries a timestamp derived from the recording device’s own clock, and that clock can drift from true time by minutes, resyncing (if at all) only when there’s a signal. Step-count and orientation data from a phone come from a motion sensor (an accelerometer) sampling at fixed intervals, but the algorithm that turns raw motion into a “step” is pattern-matching: it sees a motion signature close enough to the one labeled “walking,” not an actual person walking. A vehicle bouncing on a rough road can generate exactly that pattern; motor-vehicle vibration is a documented source of false step counts in accelerometer research (somewhere, a fitness app is quietly congratulating a car for its cardio). Vehicle GPS samples on its own fixed interval too, and carries a position error that grows near buildings and tree cover because of multipath reflection, exactly the conditions of a wooded rural property. GPS, then, is least trustworthy precisely where you’d most want it to be precise.
Three sources, three sampling rates, three levels of precision, three different failure modes. Rather than blending them into one smooth story, the expert’s job is to work out how far each clock drifts from a verified external time reference (usually a cellular network clock or an NTP server), then report the result bounded by the uncertainty margin of the weakest source in the chain, not the strongest.
What a software expert witness actually tests in a multi-source timeline
This is exactly where a source code expert witness does work a lay witness can’t. Instead of presenting “at 7:39 p.m. exactly, X happened,” it presents what each source says, its sampling rate, its known measurement error, and where two independent sources corroborate each other inside a shared time window. When a video, a motion sensor, and a GPS record (three sources that don’t “talk” to each other and can’t have coordinated a story in advance) converge on the same window, that’s far stronger evidence than any one of them alone. (For once, the fact that your gadgets don’t gossip with each other works against you.)
But that’s also where a careful expert stops. Step data alone doesn’t prove a person walked, only that the sensor detected a matching pattern. A serious opinion says so explicitly. Instead of presenting one raw data point as freestanding proof, it builds its conclusion on the convergence of independent sources, not on whichever one is most dramatic.
What happens when the case itself reopens
On May 13, 2026, the South Carolina Supreme Court, in a unanimous 5-0 per curiam opinion, overturned Murdaugh’s convictions and ordered a new trial, holding primarily that the county’s court clerk had improperly influenced jurors with comments about Murdaugh’s testimony and credibility during the trial, conduct the court said “egregiously attacked Murdaugh’s credibility and his defense” and triggered a presumption of prejudice the state could not rebut. It matters to be precise about what that reversal does and doesn’t say: it says nothing about the reliability of the digital evidence or the timeline methodology. It’s about conduct inside the courtroom: the sensors held up fine; it was a person in the room who didn’t. A multi-source timeline, if it was built correctly the first time, doesn’t need to be “redone.” It needs to survive cross-examination again, at the same stated level of precision and no more.
That’s the actual lesson: methodological discipline doesn’t depend on the outcome of one specific case. It has to hold up even when the case comes back to court.
This article discusses a public U.S. criminal proceeding for professional-education purposes only. It is not legal advice. Verify the current status of the case (including any appellate rulings) and consult counsel before relying on it.