There are two quick flaws with this argument.

1) It assumes you can't track ROI on 'traditional' marketing. You can. There are a myriad of ways: Using advertising-specific numbers with call forwarding; Statistics modeling; asking the customer, &c. People generally don't spend millions unless they know it's going to matter, and traditional advertising has defended itself more than once.

2) It assumes the purpose of advertising is 'click-through'. Most advertisement is building name recognition: not everyone needs an ACME Widget right now, but I want to make sure they know to go ACME when they need it.

3) It forgets you're competing in a shared environment. If Pepsi shrinks its ad budget, Coke wins. "Many manufacturers secretly question whether advertising really sells their product, but are vaguely afraid that their competitors might steal a march on them if they stopped."

I agree with most of what you're saying, but would like to point a few things out:

1) Yes, you can track ROIs of that nature, and many people do. But as Steven Levitt says, it seems to not be a part of corporate culture to do so. And according to him, they often do waste millions. It's not the marketer's money, after all. They'll often say things like, "It's my job to spend the budget" when Levitt points out their losses.

2) There's more to online ads than clickthrough, I agree. I didn't say there wasn't. My point was that the stats are easier to come by, and they'll be harder to weasel up.

3) That is a legitimate concern, I agree. However, being in a marketing arms race should already have been factored into the ROI, IRR, or whatever all-inclusive metric or set of metrics you are basing the budget decisions on. I don't mean Pepsi should give up, or that its marketing budget is bloated, but that even in an arms-race you can go too far.

Good analysis, but there's still room to fudge internet numbers, especially when you're getting high priced CPM banner/skyscraper ads for "branding" purposes.

Exactly branding was our marketing department go-to excuse for ROI negative campaigns.

We ended up trying to measure the branding by tracking users who came to our site directly by typing the url into the address bar, or who googled our website by name, so we can estimate how strong the branding was.

The only problem is you can't really do this on a per campaign basis, but if you geo target your campaigns and check the brand strength stats on the same geo targets you can get a rough estimate.

Just out of curiosity, how would these events unfold?

1) They do the campaign and 2) rationalize [what I presume is] a failure, or 1) Rationalize a bad campaign and 2) get it going?

Also, how valuable was branding? Were the bad ROI's justified after all?

You may be very spot on there. There's an old adage in advertising, something like "Half of money spent on ads is wasted, but you can't tell which half." Online, you can, so you have to make smarter investments to maintain the appearance of being a good marketer.

I think you have some good points, I have friends in (offline) marketing - these people are the greatest bullshitters of all time...

But I think that there is another simpler explanation: supply and demand. Since everyone has a blog with banners, and one on five people runs a website with ads the supply is enormous, and this naturally drives prices down.

The supply of banners might be huge, but companies don't buy the banners as much as the eyeballs that see them. It's the same with TV; marketers buy a spot for its viewers; the price of an ad depends on the show's ratings.

That is not true. A marketing campaign is based on a media-mix, where it is decided what ratio of money will be spent on what. For instance 25% print, 25% outdoor, 25%television and 25% web.

You reach different demographics in different ways with each channel. Marketers know this.

I'm not sure that internet ads truly pay that much less... is it cheaper to deliver a 15 second video ad to 250,000 untargeted people on the internet or on TV?