The snapshot problem
A campaign analyst builds a strategy from what is, functionally, a track file. Registration rolls updated monthly. A tracking poll run over three days last week. Turnout models trained on the last comparable cycle. Media coverage tone scored by a contractor who filed the report on Tuesday. Each of these is a stream, and each was frozen the moment it was read. The strategy that comes out of the room is a belief about the electorate, and by the time it reaches a media buy or a canvassing script, the electorate has moved. Not metaphorically. Registration changes daily. Early voting reshapes the denominator hourly during its window. A debate, a resignation, a weather event on election morning can shift turnout propensity in blocks that no snapshot anticipated because no snapshot was live when it happened.
This is the same failure air traffic control was built to eliminate: a strategy fixed on a position report that the object has since left. The Federal Aviation Administration did not exist because more radar was elegant. It existed because on 30 June 1956 two aircraft, each flying from a filed plan and a periodic position report, occupied the same point over the Grand Canyon. Elections do not produce a wreckage report when a canvass plan goes stale. They produce a loss, quietly, weeks later, attributed to "ground game" or "the mood shifted" — language for a belief that was never updated because there was no architecture for updating it.
What ATC actually offers the analogy
Air traffic control's core discipline is separation: a quantified confidence that two aircraft will not occupy the same point, purchased by continuous surveillance and spent as reduced spacing. Translate that into the electoral domain and the quantity being purchased is not spacing between candidates. It is spacing between belief and reality — the gap a campaign can afford between what it thinks the electorate looks like and what the electorate has become.
That gap is priced the same way. A campaign working from a poll fielded over three days last week, cross-tabbed against a voter file updated once this month, is running procedural control: wide separation, because the intake is stale and confidence has to be bought with margin. A campaign fusing daily registration deltas, a rolling tracking poll re-weighted nightly, absentee and early-vote return rates by precinct, and media sentiment scored continuously, is running something closer to radar coverage: narrower margins, because the belief is closer to current. Neither is fantasy. Both exist today. The difference is spacing, and spacing is paid for in resources — field staff redeployed late, ad spend locked into a media plan that assumed a demographic split now three points off, a debate-night message calibrated to a state of opinion that no longer holds.
Two positions, honestly held
Position one: continuous intake is the only defensible basis for strategy in a contest this volatile, and anything less is procedural control dressed up as analysis. The electorate is not a fixed population being measured; it is a moving system with feedback loops — coverage shapes turnout, turnout shapes coverage — and a snapshot cannot represent a feedback loop, only a frame of it. A campaign analyst who fixes a strategy on a poll from ten days ago is doing exactly what TWA 2 and United 718 were doing: flying on a filed plan and hoping the last known position still holds.
Position two: this overstates what continuous data can buy in a domain built on secrecy and small samples. The ballot is deliberately non-cooperative. There is no transponder equivalent for voter intention — no Mode S reply that says "I have decided, and here is my heading." Polling is a noisy, expensive, intermittent interrogation of a system that has every incentive to misreport itself, understate certainty, or refuse to answer the phone at all. Response rates under 5% are common. Continuous re-polling does not fix a sampling problem; it multiplies a noisy signal at higher frequency, which can manufacture the appearance of movement where there is only variance.
"You are describing more polls as if more polls were more radar returns. A radar return is physics. A poll response is a person deciding whether to lie to a stranger about their vote. The analogy imports precision that the underlying measurement cannot supply."
That objection lands, and it should be conceded rather than argued around. The response is not that polling can be made as reliable as a Mode S transponder reply. It is that the architecture the thesis is arguing for does not require it to be. Air traffic control's advance was never simply "more observations." It was observations carrying provenance, integrity flags, and graceful downgrade — a track that says, explicitly, this position came from primary radar, this one from a pilot's voice report, and the confidence attached to each differs and is recorded. An electoral intake system built on the same logic does not treat a live-tracking poll and a three-week-old panel survey as interchangeable inputs to be averaged. It tags each stream with its own decay rate, weights registration data as near-certain and message-testing data as volatile, and lets the analyst see which parts of the belief are current and which are aging out. That is achievable without pretending the electorate is instrumented the way a runway approach is instrumented.
Where the second position actually wins ground
Position two has a stronger case on diminishing returns than on measurement quality, and it deserves to be stated plainly. The North Atlantic ran safely for fifty years on position reports every ten to fourteen minutes; most electoral decisions similarly do not need hourly refresh. A field director scheduling canvassers for next weekend does not need a stream updated every hour; a monthly voter-file pull and a weekly tracking wave is enough resolution for that decision, and running continuous sentiment analysis against it adds cost without adding usable signal. Most of a campaign's operational tempo is genuinely closer to procedural control, and that is not a failure of nerve. It is correct sizing.
Where continuity earns its cost is at the margin that actually breaks campaigns: the final ten to fourteen days, when early and absentee returns start flowing daily, media coverage cycles in hours rather than weeks, and a single debate performance or resignation can move a demographic bloc before the next scheduled poll would ever have caught it. This is the electoral equivalent of the North Atlantic gap closing from ten minutes to real time when Aireon's space-based ADS-B receivers came online in 2019: the certainty that had previously been bought with wide margins could suddenly be spent on tighter ones, and NAV CANADA and NATS took the return not as extra accuracy for its own sake but as recovered spacing — fourteen nautical miles where an hour of procedural buffer once stood. A campaign that fuses daily early-vote return rates against registration cohorts, in the closing fortnight, is not buying more accuracy either. It is buying back the margin it would otherwise have to hold against a strategy going stale mid-flight.
The narrowed claim
The thesis does not resolve cleanly in either direction here, and it should not. Continuous intake is not vindicated as the correct operating mode for every campaign decision — most of the calendar tolerates, and should use, batch-style snapshots, because the cost of continuous fusion is real and the electorate's tempo for most of a cycle is slow enough to bear it. But the specific failure named at the outset — a strategy fixed on a snapshot the electorate moved past — is real, recurring, and structurally identical to the failure air traffic control was built to eliminate. It concentrates in the closing window, where turnout signals, early returns and coverage move faster than any scheduled poll can track them, and it is there, not everywhere, that a provenance-tagged, continuously revised belief earns its cost. The claim narrows to: continuous intake is not owed to every decision a campaign makes, but it is owed to the decisions made in the interval where the gap between belief and electorate is closing fastest and the cost of being wrong is no longer recoverable by the next scheduled wave.