Seven months of watching. Mostly, it found lunch.

An earlier version of this analysis found 58 episodes rather than 44. Twenty-two of them were an artifact of our own scoring: when fewer than three venues are reporting, the index is capped at 49 rather than allowed to claim anything from a tiny sample. Overnight that is most rounds — so the baseline for midnight to 2am became a stack of synthetic 49s, and any night three venues did report scored as a dramatic anomaly against it. Thirteen of those false episodes were Fridays.

Those 456 capped rounds are now excluded from the baselines entirely. We are documenting it because a tracker that only publishes its hits is not measuring anything — it is marketing.

On this evidence, no — not as an indicator of anything happening inside the Pentagon. Across 171 days, 68% of everything it flagged happened between 10am and 3pm, and the single largest departure in seven months was the Fourth of July. The rest are Friday nights and busy weekday lunches.

That is not a reason to stop measuring it. It is a reason to be accurate about what is being measured: restaurants near the Pentagon get busy at mealtimes, and this is what that looks like at 30-minute resolution over seven months. If the pattern ever genuinely breaks, this archive is what would let anyone tell.

Everything here is recomputed from raw readings with the current formula, so the whole record uses one definition rather than whatever the scoring happened to be at the time. The full methodology is here.