Validation · field-tested

How early can it warn you?

We don't just claim accuracy — we measure it. We replayed a real severe-weather outbreak straight from the archives and scored our lightning logic against satellite-detected lightning. Here's how it did.

~1 hr
average lead time
before lightning reaches you
6 hr
longest lead
on long-tracked storms
5 regions
tested
Texas to Florida to the Midwest
1.4M+
real flashes
scored against satellite truth
The headline result

Time to get to safety

For every location that was eventually struck, we measured how far ahead the app first raised a lightning watch or approaching alert. That gap is the time you'd have to act. We ran it on five outbreaks in different parts of the country:

OutbreakRegionMedian leadAverage leadLongest lead
Jun 2, 2026East Texas (Southern Plains)30 min59 min5.8 hr
Jun 19, 2026Lower Mississippi Valley15 min56 min6.0 hr
Jun 21, 2026Indiana & Illinois55 min76 min5.6 hr
Jun 24, 2026Eastern Iowa (Upper Midwest)70 min74 min5.8 hr
Jun 24, 2026Florida peninsula60 min70 min5.8 hr
Across all five events~65 min averageup to 6 hr

Lead time depends on how a storm behaves: an organized line marching in from a distance (June 21) gives far more warning than fast pop-up convection developing nearly overhead (June 19). Either way, you're alerted before it reaches your 10-mile ring.

The test

Replaying a real outbreak, minute by minute

Backtesting means re-running the app's exact lightning logic over a past storm and checking it against what actually happened — using an independent source of truth.

The events
Five severe thunderstorm outbreaks in different regions — East Texas (Jun 2), the lower Mississippi Valley (Jun 19), Indiana & Illinois (Jun 21), eastern Iowa (Jun 24), and the Florida peninsula (Jun 24), 2026 — each a 6-hour window.
Ground truth
Independent satellite lightning detection — nearly 1.5 million real flashes across the windows, recorded separately from anything the app computes.
Model input
Archived high-resolution weather-model forecasts, fed through the exact production logic.
Coverage
~2,250 grid points across the five regions, scored every 5 minutes.
What we score
For every point and time: did we raise the alert before lightning actually arrived — and how far ahead?
1,462,722 lightning flashes ~2,250 grid points 5 regions 30 hrs replayed satellite-verified truth
Built to be trusted

Alerts worth acting on

A warning is only useful if you can trust it. Two design choices keep Before It Strikes' lightning alerts meaningful — not noisy.

🎯

“Approaching” means it's coming to you

The approaching alert only fires when a storm is genuinely closing in on you — never when it's stationary or moving away. So when you get the heads-up, it's one worth acting on, not just noise from a storm that's passing by.

⚖️

An honest chance

The lightning “chance” is calibrated to your exact location, so a distant threat never overstates your risk. High confidence is reserved for real, nearby lightning — the number you see reflects the threat where you actually are.

How we keep it honest

The fine print — because safety

Read the numbers with these caveats

Before It Strikes surfaces official National Weather Service products as the authoritative source of truth. The lightning features described here are experimental situational awareness only and must never be used in place of official warnings or lightning-safety guidance. When in doubt, seek shelter.

← Back to Before It Strikes