Field notes: the week curbside pickup made the lot count lie

I've been pulling daily vehicle counts off satellite for a mid-cap home improvement chain for about six quarters now, mostly as a sanity check against the app-panel footfall data everyone else on the desk already has open. Late August, the count at eleven of the chain's higher-volume boxes dropped almost nine percent week over week. No weather event, no regional softness in the panel data, nothing in the transcript from the prior quarter that would explain it. My first instinct was that this was the leading edge of a traffic slowdown the Street hadn't priced in yet.
It was curbside pickup.
The chain had spent the spring reconfiguring its front-of-store layout at a chunk of locations, carving out six to ten dedicated pickup stalls right along the building face, the kind with the painted numbers and the cone at the end. Those stalls sit inside the building footprint I'd been counting against, but a car doing a curbside pickup is in and out in under three minutes. A car on a real shopping trip, in a home improvement box especially, sits for twenty, thirty, sometimes longer if there's a lumber run involved. Same parking area, wildly different dwell pattern, and my daily snapshot was built on a single pass per day. Depending on what time the satellite caught the lot, I was either over- or undercounting the curbside traffic relative to the browsing traffic, and the ratio between the two had been quietly shifting all summer as the chain pushed BOPIS harder in its marketing.
The count was measuring something real, just not the thing I thought it was measuring, and that distinction is the one that matters when you're trying to get ahead of a print instead of explaining one after the fact.
What tipped me off was comparing the lot trend against loyalty-app order data a buy-side contact mentioned in passing: click-and-collect orders at those same eleven stores were up sharply for the period. Once I split the pickup bays out as their own zone and watched them separately from the general lot, the picture flipped. General-lot stalls were flat to slightly up. Pickup-bay turnover was way up. Total vehicle traffic at the property was healthy the whole time. It had just redistributed inside the same asphalt in a way that a single aggregate count couldn't tell apart.
This lot holds two distinct populations of cars at once. Curbside pickups average under three minutes. Browsing trips run twenty to thirty minutes once you count a lumber run. Net the two into a single daily number and a mix shift between them shows up as a traffic swing that was never actually there. Grocery anchors with curbside have the same problem, as do any box that's leaned into ship-from-store fulfillment, where delivery vans sit in a loading zone that can get lumped into the same count as customer stalls if you're not careful about where you draw the boundary. The fix is segmenting the lot before you ever build the index, then watching each zone's trend on its own instead of netting them all into a single number.
That's the part I wasn't doing well enough on my own spreadsheet, and it's the reason I started looking at a service that builds the index that way from the start rather than leaving the zoning to whoever's eyeballing the lot that week. If you want the daily count already split and trended before the print instead of after, a seasonally adjusted lot-count series built per chain is the thing to go check.
Next time a lot count moves nine percent in a week, check the bay layout before you check the macro.