Blog· October 26, 2026

When a yard won't give you a clean brand count

A backlit excavator silhouetted on a hilltop at sunrise, details lost against the glare.

You pull the frame expecting a tidy row of units you can tag by brand, and instead you get a jammed corner of the lot: three machines nosed in tight, a trailer roof throwing shadow across the one next to it, and a stack of attachments that could belong to any of four OEMs. This happens on real yards more than any benchmark chart admits. Contractor laydown areas, rental return lots, and dealer back rows are not built for a clean overhead read. They're built to fit iron in, and that works against you every time.

The instinct is to force a number anyway. Squint at the pixels, guess at the paint color, write down a count that feels defensible. Resist that. A forced count that's wrong looks identical to a correct one on your tracker, and it will poison every trend line downstream of it, because nothing further down the chain knows the difference between a confirmed read and a guess.

Why some sites never resolve to a clean count

A handful of conditions show up again and again on the yards that give analysts trouble:

Stacking and nesting. Excavators backed bucket-to-bucket, skid steers parked bumper-tight to save space. From directly overhead, two machines can read as one footprint, or one machine can throw enough shadow to look like two.

Low-contrast paint matchups. Yellow-on-yellow between brands that share a similar hue at a distance reads differently than it does standing next to the machine. At 0.3 m resolution you get shape, proportion, and rough color, not a logo.

Partial occlusion. A canopy, a parked trailer, a stand of trees at the lot edge. Anything between the sensor and the unit removes information that doesn't come back no matter how long you stare at the crop.

Transit and staging churn. Yards that turn over fast for rental or project staging can have units mid-move at the moment of capture, half on a lowboy, half off, with nobody having agreed in advance whether that counts as present on site.

None of these are flaws in the capture. They're just what a working yard looks like from above, and any count discipline that pretends otherwise is going to break on the first busy site.

What to do with the frame you've got

Treat unresolved units as their own bucket, not as a coin flip between two brands. If a site has 40 units and 34 attribute cleanly, write down 34 attributed and 6 unresolved. That's a real number. A guessed 40-way split across two OEMs is not, and it will show up as noise the next time you try to explain why the mix "changed" quarter over quarter.

Where you can attribute with less confidence than you'd like, flag it as such rather than rounding it up to certain. A rough brand attribution that's labeled rough is still useful for spotting a shift. A rough attribution dressed up as a precise one is what gets a competitive-intel read thrown out in a room full of people who know the account.

Don't chase a single frame for certainty it can't give you. If the lot was stacked tight this quarter, check whether the next quarter's pass caught it rearranged. Yards move. A jammed corner in July is often a clear row in October.

Why the trend matters more than any one read

A single pass of a hard site tells you less than the same site read four times a year. One cluttered quarter might leave you with an unresolved count you can't do much with. Watch that same yard pass after pass, though, and the pattern that matters, a competitor's iron creeping into a corner that used to be all yours, or your own units thinning out on the back row, shows up even through the noise of any one messy frame. The actual job is catching the direction a site is moving before a rep notices it from the gate.

That's the logic behind the quarterly brand-mix tracking Share of Yard runs on named accounts: a repeatable read on the same site, quarter after quarter, so one bad frame doesn't cost you the trend. If you're tracking a book of accounts where the yards don't always cooperate, it's worth seeing what a quarterly read looks like on your own list.