Collection capacity is not what limits an airborne intelligence program. Exploitation is. Sensors and flying hours are comparatively easy to add, while trained analysts, a working tasking process and a dissemination path into the hands of someone who can act are slow to build. Size the program against the analysts, not the aircraft.
Every year the sensors get better and cheaper. Resolution improves, turrets shrink, wide-area payloads that used to need a large aircraft now fit on a single-engine turboprop.
None of that has changed the thing that actually limits an airborne intelligence program, which is how many people can look at the product and how quickly what they conclude reaches somebody empowered to act.
The arithmetic nobody runs before signing
A single full-motion video feed, monitored live, needs roughly one analyst per shift plus supervision. Run that capability around the clock and you are staffing three shifts plus relief — call it five people minimum for one orbit, before anyone does post-mission exploitation, writes a product or briefs a customer.
A wide-area payload does not multiply the analyst requirement by two. Depending on how it is tasked it can multiply it by five or ten, because the entire value of wide-area collection is the ability to go back through it and follow things that were not the original target. That is analyst time, and it is time that scales with hours collected rather than hours flown.
Programs routinely buy the wide-area sensor and staff for the single feed. The result is recorded imagery nobody has capacity to exploit — technically collected, operationally invisible.
The second constraint: the last hundred meters
Even where exploitation capacity exists, product frequently fails to arrive in a form the recipient can act on.
An analyst working in English produces a written assessment for a commander who works in Spanish and needs it in twenty minutes. A ground station produces a clip with coordinates embedded in metadata for a unit whose radios do not carry data. A weekly intelligence summary lands on Monday for an operation that was decided on Friday.
Each of these is a format-and-cadence failure rather than a collection failure, and each one makes a technically successful program operationally useless. They are also the cheapest failures to fix, if anybody looks for them — which requires sitting with the end user and watching what they actually do with the product, not asking them what they want.
How to size the analytic cell honestly
Start from the decision, not the sensor. What decision is this capability meant to inform, who makes it, and on what cycle? A border interdiction decision made hourly requires a completely different analytic posture from a pattern-of-life assessment delivered monthly.
Then work backwards: cadence sets the shift pattern, shift pattern sets headcount, headcount sets facility size and power and recruiting timeline. Only then does the sensor question have a defensible answer, because now you know what exploitation capacity the sensor is allowed to generate.
Done in that order, some programs conclude they should buy a cheaper sensor and hire four more analysts. That is usually the right answer and it is almost never the one that gets proposed, because the sensor is a line item somebody can sell and the analysts are a recurring cost somebody has to defend.
Nobody has ever lost a program because the sensor was slightly worse than it could have been. Programs die because the product arrived too late, in the wrong format, to somebody who could not act on it.