The law of large numbers problem in retail comps, explained

Every retail analyst runs into this once a chain gets big enough: the same 3% comp that looked great at $5 billion in revenue now takes an extra $1.5 billion of incremental sales to produce. That's the law of large numbers problem, and it has nothing to do with the health of the business and everything to do with denominator math.
The term gets tossed around on earnings calls as a built-in excuse. A CFO says comps are decelerating "due to the law of large numbers," and the room nods, because at some point a chain's base gets too large to grow at the pace it did when it had 800 stores instead of 4,000. The mechanical reason is real. Whether it's the whole story, or management is leaning on it to explain something else, is the actual analyst question.
Why a bigger base drags down the comp
Same-store sales measure growth against last year's sales at the same locations. As the base dollar figure rises, each percentage point of comp takes a bigger pile of dollars to produce. A chain doing $200 million per average unit volume needs far less incremental revenue per store to post a 4% comp than one doing $900 million. Scale that across thousands of doors and a mega-cap retailer decelerates even when execution hasn't slipped at all.
That's also why comping a mega-cap chain against a smaller, faster-growing peer on headline comp percentage alone doesn't hold up. A 2% comp off a $100 billion base and a 6% comp off an $8 billion base can represent similar, or even smaller, absolute dollar growth for the smaller name. Skip that math and you end up overrating momentum at small-caps and underrating steady compounding at scale.
Unit growth versus comp growth
The second half of the problem shows up in how a chain blends two very different levers into one revenue number: new unit count and comp. A chain opening 150 net new stores a year can post a healthy headline revenue figure even while comp is flat or negative, because new boxes ramp into the base against a prior-year number of effectively zero, which inflates the top line.
This is where models get sloppy. Skip backing out unit growth to isolate the comp, and you're not measuring like-for-like demand. You're measuring square footage growth with a sales number attached to it. A chain that grew units 8% and comp 1% tells a very different demand story than one that grew units 1% and comp 8%, even when both land at roughly 9% total revenue growth in the same quarter.
Cannibalization compounds it. New stores opened near existing ones pull share from the existing fleet, so comp can go negative in a market even as total unit count and total revenue there rise. Management rarely breaks this out by market on the call. Analysts are left inferring it from aggregate figures that net the effect out before anyone outside the company sees it.
What this means heading into the print
None of this means scale kills growth. It means the comp number by itself, without context on unit growth, store age mix and the base it's measured against, tells you less than it looks like it tells you. A 1-point deceleration in comp at a chain with 4,500 stores means something different than the same 1 point at a chain with 400.
The practical problem for anyone modeling a mega-cap retailer ahead of a print is timing. The comp figure and management's guidance both describe a quarter that's already closed, and the 10-Q itself usually lands weeks after the quarter ends. Traffic at the parking lot level starts moving before the sales commentary does, and that gap is where a lot of mispriced quarters get caught. That's the gap a daily read on parking lot activity across the whole chain is built to close, store by store, independent of how the comp math is getting dressed up on the call.
If you're trying to get a read on this quarter's traffic before the print lands instead of after it, that's the whole idea behind Sales Nowcast.