Building a same-store sales proxy from parking lot counts

Same-store sales is the number every retail REIT earnings call circles back to, and it's also the number you don't get until the tenant files. A parking lot count series won't hand you an exact SSS print before the quarter closes. What it gives you is a parallel read, built from the lot itself, that you can run against your own model instead of waiting on management's language.

Here's roughly how analysts who do this put the proxy together.

What a parking lot count measures

A count is vehicles parked in a lot at a given pass, not foot traffic and not basket size. For a big-box anchor or a grocery-anchored center, parked cars correlate with visits reasonably well because most trips are car trips and most visits leave a car sitting for the length of the shop. For a mall with heavy transit access, or a center with a lot of walk-ins from adjacent residential, the correlation is weaker and you should say so in your note rather than pretend the proxy is clean.

The other thing a count measures is parking supply changes, whether you want it to or not. A restriped lot, a new outparcel pad, a temporary construction fence around forty stalls: all of these move the count without a single customer behaving differently. Know the site before you trust the series.

Building the proxy, step by step

Start by picking lots, not the whole tenant footprint. A national retailer with 1,800 stores gives you 1,800 possible sites, and you only need a sample of them. Pick a set of anchor locations in the markets you cover, weight toward stores that report into the comp base (exclude anything under two years old, same as the tenant's own comp rules), and track those specific lots over time rather than swapping sites in and out.

Pull a daily occupancy read for each lot. One pass a day smooths out the noise of an individual bad-weather afternoon; what you're after is the trend across weeks, not a single Tuesday.

Normalize before you compare anything. Day-of-week matters more than people expect: Saturday counts at a grocery-anchored center can run two or three times a weekday count, so a raw week-over-week number is mostly noise unless you're comparing like days. Strip out the obvious seasonal shape too, holiday weeks, back-to-school, the week after Thanksgiving, so what's left is the part that actually tracks demand.

Index the normalized series to a base period, then lay it against the tenant's last several quarters of disclosed comps. This step turns a count series into a proxy, because it tells you whether the two move together for this tenant, in this box format, in these markets. Some tenants will show a tight relationship. Others won't, and that's useful information on its own, since it tells you where the proxy is worth building a forecast on and where it's just a sanity check.

Once you've got a relationship that holds across a few quarters, you can watch the current-quarter count series against your model's comp assumption and flag the gap before the print. The exercise gives you a reporting-before-the-tenant-does read, one you reconcile against the filed number once it lands.

Where the proxy breaks

Curbside pickup and buy-online-pickup-in-store traffic can inflate parking counts relative to in-store sales, since a BOPIS customer parks for ninety seconds instead of walking the floor. Delivery-heavy formats break the relationship the same way from the other direction. Parking structures and underground garages are outside what an overhead pass can count at all, so any tenant leaning on structured parking needs a different approach entirely. And a single bad quarter of construction noise at a shared lot, a renovation, a lease-up next door, can throw off a whole site's baseline for months.

Listing out the failure modes gives you the boundaries the proxy actually works inside. Treat it as a model input with known limits and it earns a place in the forecast; treat it as ground truth and it will eventually embarrass you in front of a portfolio manager.

If you're building this out across a watchlist of properties, Parking Lot Counts delivers the daily occupancy series per site so the indexing and normalization work above is something you do on a clean feed instead of a raw image archive.