Blog· October 26, 2026
What counts as a visit in a quarterly brand-mix read

If you've ever argued with a regional manager about whether "we checked that yard in April" means anything, you already know the term "visit" gets thrown around loosely. Before you can trust a quarterly brand-mix read, you need to agree on what a visit is, because the whole trend line depends on it.
A visit is one pass, one date, one count
A visit is a single imaging pass over a named site on a given date, used to produce one count and one rough brand attribution for that site. Not a week of activity. Not an average. One pass, one date, one read.
That sounds obvious until you start pulling real imagery. A satellite tasking a site on a Tuesday and coming back cloud-free on a Thursday doesn't give you two visits for the quarter, it gives you one usable visit and one wasted tasking attempt. If an analyst reprocesses the same Tuesday pass a week later because the first pull had a georeferencing error, that's still one visit. The visit's date is the date stamped on the imagery, independent of when anyone opened the file to look at it.
This matters because a rep's windshield survey doesn't have this discipline. Someone drives past a gate, glances at what's parked, and calls it a read. There's no fixed date attached, no repeatable framing, and no way to tell six weeks later whether the lot looked the same or had turned over twice. A visit, properly defined, is a snapshot pinned to a calendar date and a fixed footprint over the site, which is the only thing that makes quarter-over-quarter comparison honest.
Why a skipped quarter isn't a broken trend
Cloud cover, smoke, a tasking conflict with a higher-priority order. Any of these can knock out a pass for a given site in a given quarter. When that happens, you don't have a visit. You have a gap.
Mark the quarter as missing instead of interpolating a brand split nobody observed. A gap doesn't invalidate the trend the way a bad data point would. Three real visits over three quarters with one gap still tells you more than a rep's memory of what he thinks he saw in March. What it shouldn't become is a quarter quietly filled in with "probably similar to last time," because that's exactly the kind of guess the whole exercise exists to replace.
This is also where cadence discipline earns its keep. A quarterly cadence schedules four attempts a year per site, and how many of those turn into usable visits depends on weather and tasking conflicts. Some years you get all four clean. Some years a site gets three usable passes and one washout. The count of actual visits, not the calendar quarters that passed, is what the trend is built on.
Counting units, not reading nameplates
One more distinction worth pinning down: a visit produces a count and a brand attribution by visit, not a certified identification of every unit on the lot. At 0.3 m GSD or finer, from VHR satellite or aerial RGB imagery, you can count distinct pieces of equipment and make a reasonable call on brand based on shape, color, and footprint. That's attribution by visual read, done consistently pass over pass. It's not a nameplate inspection, and nobody pulling satellite imagery should pretend it is.
What makes the read useful is consistency: the same method gets applied the same way every quarter, so a shift from three visits ago to now reflects a change on the ground rather than a difference in how two people eyeballed the lot.
A few quarters of visits logged against a site turn into a per-site share reading that moves, which is the thing a yard walk or a rep's drive-by was never built to catch. A repeatable quarterly read across every account that matters is the whole point of tracking it this way instead of trusting whoever happened to be driving past that week.
If you've got a short list of accounts where the count and the mix would change how you plan next quarter, that's the place to start.