How to handle unit growth in a chain-level parking lot index

Build a parking lot traffic index across a retail chain long enough and you'll hit the same wall every time: the chain opens stores. A new unit adds a new lot to your sample, and if your index just sums or averages vehicle counts across every location you're tracking, the line goes up whether or not anyone's actually shopping more. You end up measuring square footage growth and calling it demand.
This matters most right before a print. If a chain guided to 40 new units this year and you're three weeks from the call with a traffic index that's run hot since August, the first question to ask is whether that's comps improving or just the store count catching up.
Separate the index you track from the index you report
Same-store sales reporting solved this problem decades ago by drawing a hard line between stores open a full comparable period and everything else. A lot-count traffic index needs the same split, and it needs to exist as two series, not one with a footnote.
Keep a same-store cut that only includes locations with a full trailing period of data, typically 12 to 13 months, matching however the chain itself defines its comp base. Keep a separate total-chain cut that includes every lot you're watching, new units included, so you can see the gap between the two. When the two series diverge, that gap is roughly your unit growth contribution. When they move together, unit growth isn't doing much and whatever the index is showing is closer to organic.
Most of the work is just not letting a new lot into the comp cut before it's earned its way in.
Rebasing after a store addition
The mechanical question is what to do with a lot the day it shows up in the data. A new store doesn't open at steady-state traffic. There's a grand-opening spike in the first week or two, often inflated by promotional foot traffic, followed by a drawdown to whatever the location's real run rate turns out to be. If you fold that lot into your index the moment satellite imagery first picks up cars in it, you're adding a traffic spike to the chain total at exactly the wrong moment.
The cleaner approach is to exclude a new lot from both cuts for an initial window, let the ramp settle, and then admit it to the total-chain index first and the same-store index only once it clears the comp-period threshold. Pick a window length and hold it constant across the whole chain. A 60- to 90-day exclusion before a lot counts toward anything is a reasonable default, and consistency matters more than the exact number, since you want unit-over-unit comparisons within your own index to mean the same thing every quarter.
Closures need the mirror treatment. A remodel or a quiet closure drops a lot out of your sample, and if you don't flag that drop you'll misread a falling average as falling traffic when it's really a shrinking denominator.
Reconcile against the unit count the chain already discloses
The chain's own 10-Q or investor deck usually gives you store count by quarter, sometimes by banner or region. Cross-check your lot inventory against that number before every print. If you're tracking 1,180 lots and the filing says the chain ended the quarter with 1,205 units, find the 25 you're missing before you draw any conclusions from the gap between your same-store and total-chain lines. A mismatch here is usually a data gap on your end, new builds not yet picked up, a banner you're not covering, not a real signal.
None of this is exotic. It's the same bookkeeping discipline the chain's own comp-sales methodology already runs on, applied to a traffic proxy instead of a revenue line. Sales Nowcast ships the daily, per-chain vehicle-count series with that same-store versus total-chain split built in, so new openings show up as a labeled addition to the series rather than an unexplained jump in the index. If you're building a nowcast ahead of a print, that's the data worth starting from.