What is a target deck in imagery intelligence?
A target deck is the working set of locations an imagery shop is actively tracking: the sites worth a repeat look, the ones that feed a nomination list, the ones someone upstream asked about by name. If you've built one, you know it's less a single document than a living folder structure: imagery clips, coordinates, a line or two of context on why the site matters, and a revisit cadence that depends on how hot the target is.
Ask five analysts what belongs in a target deck and you'll get five slightly different answers, because the term gets used loosely across shops. Some mean the literal slide deck that goes into a briefing. Others mean the underlying database of named areas of interest, sometimes called a target folder or nomination list depending on the unit. Both usages are correct in context. What they share is the same core idea: a curated list of places, kept current, that someone decided deserve recurring attention.
What goes into a target deck
At minimum, a target deck entry carries a location, a reason the site was nominated, and enough imagery history to show change over time. Shops that do this well also track collection requirements against each entry, so a tasking officer knows when the last usable pass happened and when the next one is due. The deck is only as good as its upkeep. An entry with imagery from eight months ago is stale: the site may have changed completely since that pass, and nobody watching the deck would know it.
The nomination process is where most of the analyst time goes. Someone has to look at a site, decide it matches a pattern worth watching, write the justification, and get it added. That step is manual almost everywhere. A target gets nominated when an analyst happens to spot it. The archive itself doesn't get searched for every other site that looks the same.
The target folder problem
Say your deck has a surface-to-air missile site, a suspected staging area, or a processing facility with a distinctive layout. The deck tells you to keep watching that one site. It doesn't tell you whether the same layout shows up fifteen other places across the region you're responsible for. Finding those takes someone scrolling the archive tile by tile, comparing shapes and shadows against the one example they already have, across an area too large to browse end to end.
This is the gap between a target deck and a complete target folder. The deck holds what's already been nominated. A complete folder holds every matching site, nominated or not. Most shops don't have the second thing, because manual tile-by-tile search doesn't scale to a province-sized area of interest, let alone a country.
Broad Area Search is built for that specific gap. Mark the one site already in your deck, and it searches the rest of your imagery archive for places that match its visual pattern, using imagery embeddings instead of a human eye working tile by tile. The nomination write-up and the judgment call on whether a match belongs in the deck still sit with the analyst. What it takes off someone's plate is the physical work of looking at every tile in the search area to find the other candidates in the first place.
A target deck will always need a person to decide what belongs in it. The search for every other site that looks like the one already on the list doesn't have to be manual once the pattern is known, and that's the difference between a deck of noticed targets and a folder that's actually complete.
If a known site in your deck might have twins you haven't found yet, that's the use case Broad Area Search is built around.