Thursday, October 09, 2008

Google describes perfect advertising

In a post titled "Ad Perfect", Googler Susan Wojcicki describes targeting ads as matching a deep understanding of a user's intent, much like personalized search.

Some key excerpts:
Advertising should deliver the right information to the right person at the right time ... Our goal is always to show people the best ads, the ones that are the most relevant, timely, and useful .... We need to understand exactly what people are looking for, then give them exactly the information they want.

When a person is looking for a specific item ... the best ads will give more specific information, like where to buy the item.

In other cases, ads can help you learn about something you didn't know you wanted ... [and to] discover something [you] didn't know existed.

One way to make ads better would be to customize them based on factors like a person's location or preferences.

It [also] needs to be very easy and quick for anyone to create good ads ... to measure [and learn] how effective they are .... [and then] to show them only to people for whom they are useful.
What strikes me about this is how much this sounds like treating advertising as a recommendation problem. We need to learn what someone wants, taking into account the current context and long-term interests, and then help them discover interesting things they might not otherwise have known existed.

It appears to be a big shift away from mass market advertising and toward personalized advertising. This vision no longer has us targeting ads to people in general, but to each individual's intent, preferences, and context.

[Thanks, John Battelle, for the pointer to Susan's post]


Daniel Tunkelang said...

Thanks for the link, which inspired my diatribe: Search is Not Advertising.

Greg Linden said...

Hi, Daniel. Interesting argument, that "advertising is all about the advertisers, and the advertisers only care about providing value to users in so far as their interests are aligned."

I think Susan's post is claiming that advertisers get the most long-term value when they align their interests with users. That is, when advertisers make sure their ads are relevant and useful to people, the clicks are all for interested buyers with strong purchase intent.

There might be a counter argument here that deceptive advertising works and continues to work over the long-term. So, under this approach, the advertiser is trying to convince people who click to buy something products they don't want at prices that are unreasonably high, realizing extraordinary profits for the advertiser, without repelling too many users or hurting their long-term reputation.

In brand advertising -- which is the main thing I think you are talking about in your post, but not the main thing Susan is talking about -- we are not trying to create an immediate action, but trying to influence people's perception of a brand, the reputation of a company or product. Brand advertising sometimes is attempting to inform, but often seems to be trying to sell an image or emotion that causes people to buy things at higher prices than otherwise are available for similar products, so, in that sense, might not be well aligned with people's interests.

Susan's post does not directly address brand advertising, but I think the broad claim is that any advertising that is at odds with people's interests will be costly to maintain, and that advertisers generally will find advertising that is aligned with people's interests to be more effective in the long-term.

Anonymous said...

No, I think Daniel is also talking about "direct relevance response" advertising, and not just branding or marketing.

At the risk of putting words in Daniel's mouth, I think what Daniel is saying is that, relevance or not, what makes advertising different from search is the method or mode of delivery.

Search is pull, advertising is push. (Search is things that you ask for. Advertising is things that others foist upon you.)

That fact alone fundamentally changes the dynamic of the information, and gives rise to the problems that Daniel describes.

Let's put it this way: if there were only one single search result and one single advertisement in existence, for your particular query, then this wouldn't be an issue. But often there are thousands of results, and dozens of advertisers.

Suppose now that there are 500 web pages that are equally relevant, based on content alone, to your query. 500 pages all about iPods. Well then the thing that breaks the tie is an organic sort of popularity linking.. pagerank-like feature. So relevance is determined by the natural growth of the web.

On the other hand, suppose there are now 50 advertisers who all want to sell you an iPod. More or less, every single advertisers is equally relevant to your "iPod" query. There is not much Google can do to automatically determine quality of merchant service or something like that, to rank by that feature.

So what ends up happening is that the feature that determines how these 50 equally relevant advertisers rank is.. $$$. The person who pays the most money gets the most traffic.

So in one case, relevance was determined by organic web structure. In the other case it was determined by cold, hard cash.

We're not talking about deceptive advertising here. We're talking about 50 advertisers, all of whom are legitimately selling iPods.

And yet ads get ranked by $$$, and not by merchant service quality, by available discounts, by combined savings offers, by shipping costs, or by other things that the *user* would probably find relevant. Instead, they are ranked by how much money the advertiser is willing to pay.

That is the exact opposite of user-focused relevance. It serves the advertiser, not the user. Daniel has a good point, here.

Anonymous said...

For example, what if I, as a user, wanted to search for (and find) advertisements on Google on which the merchant had only spent a single penny. The idea being that the merchant would agree to pass the savings along to me.. the less advertising money that the merchant has to spend, the more he can cut his own costs, and therefore lower his prices.

So let's take that idea one further: The merchant who has had to spend $0.00 on advertising is even *more* relevant to me than the one who had to spend $0.01.

And there we are again.. back in the organic results! Only this time, all advertisements have disappeared from the page, because someone paying money to get my attention is not relevant to me, because they're only going to charge me higher fees, to recoup the cost of their advertising.

Perfect advertising (and I do mean advertising, not branding or marketing) is advertising that doesn't exist. If relevance is of such key importance, then the search engine should be good enough to put that relevant information in the natural results.

Daniel Tunkelang said...

Jeremy, the words you're putting in my mouth go well with the dumplings I just devoured. I agree at least 99%.

The only quibble I have is that search doesn't have to be all pull. I'm all for push that is driven by my information needs, even if I don't actively express them, e.g., alerting. But the motivation is 100% from the user, not from the pusher.

Anonymous said...

I don't even think we have a quibble on the push/pull thing. I think I mean the same thing as you, and am just using the terms slightly wrong.

Either way, I've always felt that the arguments that the big web companies make around advertising, Google especially, fall completely flat under even moderate scrutiny.

The argument that they make, about the relevance of advertising, is its own Achilles' heel. If advertisements were already correctly ranked by relevance, to begin with, why does there need to be $$$ in the system to change that natural relevance order?

Money seems to be, by definition, anti-relevance. This isn't even Greg's counter-argument about deceptive advertising. I'm talking about a situation in which all the ads are completely self-truthful, and ranked in a relevance order. Why would anyone need to pay more or less than anyone else, in such a situation? Why auctions? Why bids? It all directly implies anti-relevance.

Greg Linden said...

Hi, Jeremy. Actually, I basically agree with your last statement, that the ads should be ordered in relevance order.

We're not quite there yet, but, with the new penalties for poor ad quality Google and others are adding, I think web advertising is moving closer to ordering by relevance.

While it still is possible for people to compensate for low relevance ads with much higher bids, the incentives seek to push advertisers to maximize the relevance of their ads.

Anonymous said...

So my question is: Suppose we actually get to this vision, of perfect relevance-ordering for advertisements.

Why, then, would anyone pay more $$$ than anyone else? Why wouldn't everyone just pay $0.01, flat?

Right now, you have ads for which people are paying $0.50/click, $4/click and (even for certain ambulance-chasing lawyer scenarios) $50/click.

If money then no longer matters, and paying more won't change your relevance ordering, there isn't going to be bidding or auctions. No one is going to pay $55/click, if no matter how much they pay, they are still ranked 4th. No one is going to pay $50/click, if they can get away with being ranked 1st by only paying $0.01, because relevance dominates.

The moment that happens, the billions of dollars that the web companies are raking in turn into hundreds of thousands of dollars. Barely enough to keep up a staff of a couple dozen people.

So can you see why I might think that keeping the influence of money as a strong part of the ranking is in the best interest of the web companies? Without it, they really wouldn't earn that much money. They're ultimately always going to serve the advertiser, at the end of the day, not the user, because the advertiser is the one bidding up the click price.

Greg Linden said...

Yes, at the extreme where the bid no longer matters, there are problems, I agree.

Everyone seems pretty far away from that, though, mostly just looking at making the bid matter a little less and relevance matter a little more, not about making the bid not matter at all.

There is a question of how far this can be pushed. Even in the extreme case where relevance is weighted much more heavily than now, it still seems to me that we could see a lot of variation in bids if, for example, there is a lot of uncertainty around the relevance (which might require bids to be used as a signal of the advertiser's confidence in the relevance) or if there are a lot of near ties in relevance (which seems likely, especially if relevance is only known with a fairly wide band of uncertainty).

But, to be honest, my thoughts here are not well formed yet. I'm quite interested in this topic, but only have started to look at it at this point.

Daniel Tunkelang said...

Jeremy, I'm glad we agree, and I suspect that "push" and "pull" are imprecise concepts in the first place.

Greg, I'm not sure how far we are away from the extreme. For example, I experimented with using AdWords to advertise my blog, figuring it would be a cheap way to spread the word. I quickly found that I'd have to spend dollars per click to do so--which quickly persuaded me to stick to word-of-mouth marketing. But it's hardly that the competitors for those AdWords had more relevant content; rather they stood to gain more from the traffic.

Perhaps the right approach, in terms of optimizing user experience rather than Google's profits, is to use an attention bonds approach. I put down a deposit for each click (or perhaps even for each impression), but I don't pay unless you as a user feel that I wasted your time. Obviously such an approach would require some thought to implement, not to mention motivation on the part of the search engines!

Anonymous said...

Daniel, isn't that what, more or less, CPA (cost per action) is about? ONLY if you click through AND if you actually execute some action (e.g. buy a product from the advertizer/merchant) does the money move around?
Something doesn't sound 100% right about that, though, as a sloppy merchant who can't convert the visitor to a customer might then not pay the publisher (e.g. Google), even though Google did its part of serving the ad... but the basic idea is there.

benzrad华中朱子卓 said...

that would be a great changes in people's life with the aid of ai.