They found that paid (i.e., ads) and unpaid traffic were nearly perfect substitutes for brand names: people will either to get to eBay. For non-brands, clicks go down, but sales don’t; ads are mostly attracting “browsers”, rather than buyers.
The real answer is that it probably depends a lot on the industry, market, and individual business.
If your point is that effective advertising strategies vary by company then I agree, but that is a tangential and meaningless point in the context of the above conversation.
I found the eBay data interesting because AI had assumed that they would have a fairly sophisticated online operation, and yet neither the branded nor non-brand ads really helped (for the latter, removing ads led to -40% on clicks, but -2% on sales).
I can’t imagine that the neighborhood pizza place is really doing a top-notch job on ad attribution; it seems a heck of a lot more likely that they throw a few ads up because everyone else does too.
With respect to eBay, proving someone sucks at running ads isn’t the same as proving ads suck. (also it matters a lot more what % of sales the ads in this case represented, rather than just the absolute change in sales).
I agree with you about the neighborhood businesses fwiw.
Otherwise why continue to run the ads? A smaller but profitable marketshare is better than a larger but negative one.
All competition is a coordination problem.
Prediction and attribution are really hard. I linked a paper below showing that eBay was wasting money on entirely ineffective keyword ads. I can’t imagine that most small businesses (especially brick and mortar ones) are much savvier.
Whether something provides a return or not for your specific situation doesn't have anything to do with whether the product itself is working and valid.
Ebay situation is different because search is as close as you can get to "intent" in advertising. You don't need user-targeting when you have a search query telling you exactly what they're looking for. Targeting is a proxy for that kind of intent when it's not available.
What I'm not yet convinced of is that they're a panacea for all small businesses. Details aside, one moral of the eBay paper is that a $50B business, with a whole dedicated department, still managed to make a dog's breakfast of their online advertising; Uber reported similar problems earlier this winter.
It seems hard and it's not obvious to me that every business has the aptitude, market, or business model to do it well. At the same time, I can easily imagine people deciding "We need to do something about advertising" and then forging ahead with whatever something they land on. This is especially true with the ROI is difficult to calculate--and I do think it's probably harder to calculate for an ad than for something with very standardized cost/performance characteristics like new equipment.
Google, Facebook, and Shopify making money doesn't really answer this question. As an analogy: "Is there gold in them there hills?" "Sure, even the guy selling shovels is making money!"
Advertising can be complicated, but part of the value of these giant networks is helping to run those campaigns effectively. I'm not sure why you think everyone has to put in similar amounts of effort. That's not necessary at all. You have tools from simply setting the budget and targeting lookalikes of people liking your posts (what most small businesses do) to running your own bidder stack (for larger brands that have entire agencies working for them).
While attribution will never be perfect, it's not guessing either. Ecommerce only makes it easier as you can track all the way from exposure to checkout, and in some ways can be more attributable than other businesses expenses.
It's not impossible that Facebook and Google capture all of the value--or more--that their clients generate by running ads: Bob's Burgers runs $500/mo worth of ads, only generates $400 in (new) profit from them, but is afraid to stop because Paul's Pizza is also running ads and they might lose even more marketshare.