What is a nowcast in equity research?

A forecast is a model of what will happen. A nowcast is a read on what is already happening, just not reported yet. In equity research, that distinction shows up most in retail coverage, where the quarter closes weeks before the company tells anyone what was in it.
Your DCF has a forecast baked into every line. Your same-store-sales estimate for the quarter is a forecast, built off comps, macro assumptions, and whatever the last call told you about traffic trends. A nowcast checks that estimate in real time, using data that exists independent of management's calendar.
Nowcasting vs forecasting a retail print
The forecast answers "what will Q3 comps look like." The nowcast answers "what are Q3 comps looking like, right now, based on signals I can actually observe." The forecast is built in September for a quarter that ends in October and gets reported in November. The nowcast updates daily through all three of those months.
This matters because the gap between forecast and reality usually only closes on the print date. By then it's too late to do anything but react. A nowcast narrows that gap while the quarter is still open, which is the only window where the information is worth anything to a position.
Credit card panel data was the first widely used nowcasting input for retail, and it's still a reasonable proxy for spend. But panels sample a subset of cardholders and skew toward certain demographics and payment types, so the read can drift from the chain's sales mix, especially for concepts with a lot of cash, debit, or private-label card volume. App-based footfall panels have a similar issue: they only see the phones running the panel's SDK, which is a thinner and thinner slice of the population every year as privacy settings tighten.
High-frequency indicators ahead of the print
The useful nowcasting inputs share three traits: they update often, they cover most or all of the locations you care about, and they don't depend on a sample that can shift composition quarter to quarter. Parking lot traffic fits that description for any retailer with physical stores and a lot. A parking lot doesn't require consumers to opt into a panel or carry a particular card. It's a physical count, and it's there whether anyone measures it or not.
The tradeoff with any high-frequency indicator is translation. A vehicle count, a card panel, an app ping: none of these are revenue. Each has to be indexed against a baseline and adjusted for the stuff that moves traffic without moving conversion: holidays, weather, store openings and closings, a calendar shift that puts an extra Saturday in one quarter versus last year. Skip that step and you're just staring at noise with more confidence than it deserves.
That's the gap Sales Nowcast is built to close: daily, seasonally adjusted vehicle counts across a chain's parking lots, delivered as a time series you can drop next to your comp model instead of waiting on the call to tell you if your number was right.
What a nowcast can and can't tell you
A good nowcast tells you direction and rough magnitude of traffic ahead of the print. It does not tell you ticket size, mix, margin, or anything happening online. A chain can see flat-to-up traffic and still miss on comps if average ticket falls or promotional activity eats the gross margin line. Traffic is one input into same-store sales, not a substitute for it. Treat a nowcast as a sanity check on your thesis between now and the print.
The other limit is store-level noise. Daily counts at a single location swing with local weather and one-off events more than a chain-wide index does. For a read you can trade on, look at the aggregate trend across the box count, not any single lot on any single day.
If you cover a retail chain and want a traffic read before the 10-Q lands, that's the use case we built.