Eventually, I could imagine an equilibrium in which a store like Giant pares back on the number of items it sells in the store, keeping only the most popular items available. You would have to order less-popular items on line. That way, they could cut back on those restocking costs.
Of course, for all I know, Giant’s business model is to charge a big markup on stuff, and they figure once they get you into the store they make a profit. And if stocking a great variety of items gets you into the store, then that is the right strategy.
I don't think it can work that way. The grocery business is a lot like an old rule of thumb we have in software: 80% of users only use 20% of features. The trick is that everyone uses a different 20%. You can't implement 20% of the features from your software and expect retain 80% of the users.
Similarly a grocer is forced to stock everything a customer might expect to get for their weekly shop. If there are one or two items out of 100 that I can't get, then I have to go to another store. And once I'm there, I'm getting everything I need there. When you stopped stocking that $4 jar of olives you haven't just lost $4 in revenue from me, you've lost several thousand dollars annually because I've taken all of my custom elsewhere.
I believe a few stores (Aldi, IIRC) have managed to pare down the selection and have customers put up with it, but that's because the customers understand what they gain is exceptionally low prices. There's an explicit deal being made between them. That's not a model a median store like Giant could go to. Similarly Trader Joe's has managed to carve out a niche of stocking very few SKUs, but virtually no one expects it to be a comprehensive shopping trip. They can get away with that specifically because they're the exception to the rule. There's not room in the market for everyone to play that niche game.
The idea of having people order in advance online is interesting. It could be nice to have a system a little like a comic book shop, where you go in and the store already has a sack of the things you've pre-ordered waiting for you. Ultimately I don't think it will work for two reasons: (1) most people want to pick their produce out themselves, which has hampered PeaPod and all the other online grocery delivery businesses; (2) the vast majority of people will not put any effort into planning their shopping trip in advance.
Almost no one actually writes out a comprehensive grocery list, to say nothing of firing up a computer and pre-purchasing. I saw some stats on grocery planning on a personal finance blog at some point, but I can't find them. Luckily we don't need them: just keep an eye out for how many people in your local store have anything resembling a list. In my experience the intersection in the sets "people who have a grocery list" and "people without babies/toddlers with them" is usually just me. And even I've only got four or five items scrawled on an index card, not my whole list.
I could be quite wrong about this. Most of what I understand of the grocery business comes from extending micro-econ first principals and reading Management in Ten Words, by Terry Leahy, the ex-CEO of Tesco. Still, the margins on groceries stores are famously thin — about 2% IIRC — so we know the market is very competitive. The adoption of a near-universal strategy in an environment like that means that we should at least start with the assumption that that strategy is very efficient, if not near optimal for current conditions.
The McArdle post which Kling quotes in the post above had this very good line:
If you want Wal-Mart to have a labor force like Trader Joe’s and Costco, you probably want them to have a business model like Trader Joe’s and Costco — which is to say that you want them to have a customer demographic like Trader Joe’s and Costco.
It's good to keep in mind these things are not mutually orthogonal.
See also this week's EconTalk with Mike Munger, which is largely about grocery stores and consumer sovereignty. (And, incidentally, how naive Michael Pollan is about business.)