Hacker News Reader

Comment by nostrademons | Hacker News Reader

I think you're seeing a Simpson's Paradox effect with those. CPC for both mobile and desktop is going up (my supposition, I don't have numbers on this). However, mobile CPCs are less than desktop CPCs, and the query mix is continuing to shift to mobile. As a result, aggregate CPC goes down.

This also explains the Bing results. Google owns mobile search, so the shift toward mobile (with its lower CPCs) disproportionately affects them. Microsoft continues to flounder in mobile, so their average CPC will be higher right until it becomes nonexistent.

Strategically, this is a good trend for Google. They're being disrupted, but they own the primary disruptor.

View on HN

This is entirely possible. I will see if I can come up with ways to test that hypothesis. One puzzling bit of not well known information is that Google recently yanked their mobile ad feed from many (if not all) of their partners. One way to interpret that would be that they were trying to shore up their mobile CPC values. Really too bad they don't break out mobile and non-mobile.

If I'm not mistaken, the CFO mentioned that Simpson's Paradox was at work (by name) during the actual earnings call.

I didn't listen to the call, just read the statements and collected what information I could from them.

But I did download the slide deck, look at Slide #5 of 15. See how all the CPC numbers for all the properties are down.

Now look at how Paid Clicks (page 4) are up for the same periods. More clicks, less money per click. It includes their AdMob business (aka Mobile) but blends a bit because is a search ad on Android a O&O (Google Sites) click or a mobile click?

But there is a more interesting point here. If we stipulate that Simpson's paradox is at work here, and we stipulate that Pichette is interested in communicating clearly to the shareholders, would he not have the fiduciary duty to disclose the confounding variable?

After all if the appearance that the value of Google's search advertising business is eroding away, and that is, rounded to the nearest billion dollars the only business that makes Google any money[1], as an officer of the company you would want to say "Sure it looks that way but when we break it out you can see that really we're doing fine." otherwise you would leave people with the wrong impression, and they might sell or buy your stock acting on that misimpression which might form the basis for future litigation against the company. (This happened a lot in the 90's btw).

For me, I tend to be skeptical (its my nature), so if Pichette wanted to convince me that the ship is doing fine, he needs to let me know why there is water in my stateroom. :-)

[1] Youtube is catching up to be sure. But pretty much not a whole lot more is.

> See how all the CPC numbers for all the properties are down.

> But there is a more interesting point here. If we stipulate that Simpson's paradox is at work here, and we stipulate that Pichette is interested in communicating clearly to the shareholders, would he not have the fiduciary duty to disclose the confounding variable?

I checked out the transcript, and it turns out I was mis-remembering the context of the mention of Simpson's paradox: it was with respect to a question along the lines of "Why is the aggregate CPC growth higher than both the network and the Google Sites"? It would of course be comically absurd for a CFO to mention Simpson's Paradox and not say what the confounding variable is, and even more ridiculous for nobody else on the call to ask him to clarify. In context, the variable was the network/Google sites split.

Recently on Adwords, it's more difficult UX-wise to stop mobile targeting or bid less for tablets (or mobiles). I guess G is intentionally pushing people to bid as high for mobile & tablets as they did for desktops. As a careful observer (not as good as ChuckMcM), I see them popping up new and new tricks to give a better quarterly report.

> Google owns mobile search

Google does not own mobile search. Infact it is struggling[1] big time in mobile search. People do not use google to search on mobile. Infact, they use specialised apps like yelp, priceline etc to search for what they need.

[1] http://www.nytimes.com/2013/07/19/technology/google-misses-e...

Don't they own AdMob, one of the top ad platforms on Android?

http://www.google.com/ads/admob/

That is definitely true. On desktop, google is the ultimate gateway for everything, while on mobile I would be more willing to open individual apps for specific things.

> Strategically, this is a good trend for Google. They're being disrupted, but they own the primary disruptor.

Only if they can maintain similar margins in the long term. If it fundamentally eats a chunk of their profit margin, I'm not seeing that as a net positive for its share price.

Still much better than having somebody else do the disrupting, no?

Of course. Anything is better than being disrupted and ending up like, say, BlackBerry. ;)

Side note: this is the first time I've heard Simpson's paradox explained intuitively.

When I read the wikipedia page it had a pretty clear real-life example (inspection reveals: still there!) --

UC Berkeley was sued for accepting male grad students at a higher rate than female grad students. On examination, most departments at Berkeley admitted females at a higher rate than they admitted males, but males preferentially applied to easier departments.

As another side note: This still constitutes bias if you turn the attribution around. Rather than "men choose easier-to-get-into programs", you could say the university underfunds the areas of study that women are more excited about.

Even assuming there's some inherent difference in what women want to study (as opposed to cultural differences which could ultimately be removed) - funding isn't always the choice of the university.

You can't take a DARPA grant to research robots and then spend it on interpretive dance lessons.

That's societal bias then. And the university is accomplice to it.

interpretive dance for robots?

(I'd really like to see that...)

Its only good if mobile CPC is rising to replace it and Google is hopping from the sinking ship to the rising one. Otherwise "Simpsons Paradox" is just a description of the particular nature of the death spiral

Android has a lot of defensive value, but if you're saying having Android makes up for a new normal with lower CPCs then Google shareholders are in for a rough ride.

So, here's my hypothesis for what's driving the lower CPCs in mobile. This is based on my observations of what I see myself and other people using Google for, not any inside information (I worked in Search, and they generally kept me away from the Ads numbers).

People use desktop search to research big-ticket items: cars, vacations, mortgages, life insurance, mesothelioma, consumer electronics, etc. There's a lot of value in a sale there. A personal-injury lawyer may collect hundreds of thousands of dollars from a client with a solid case. As a result, these keywords get bid up really high, because the profit potential of a new customer is high.

People currently use mobile search for day-to-day local needs. What restaurant should I go to? Where can I find parking? What's there to do in this town? Most of these have sale prices in the $10-50 range, and so there's just not as much money coming out of the sale. As a result, the keywords don't get bid up as high, and CPC is lower.

These use cases are not substitutes. Just because people are searching for restaurants does not mean they aren't also searching for mesothelioma. We see this in the revenue numbers, which continue to rise. But as a percentage of the total addressable market, high-value keywords like insurance and lawsuits will shrink as the market becomes larger. If you're looking at aggregate CPC numbers this will look like Google is becoming less attractive to advertisers, but what's actually happening is that Google is becoming viable for less attractive advertisers. The total market size is expanding, and with that come more marginal customers.

This, BTW, is one reason why I think Wall Street analysts are idiots. They pay attention to the wrong metrics, because those are the only numbers they have to pay attention to. I have my own set of metrics that I use to judge when it's time to sell Google stock, but I'm not going to reveal them here.

I think CPCs are declining because mobile is inherently tougher to monetize with the smaller screen real estate and Google not being the default destination for everything (like they are on desktop). There's lots of competition for attention on mobile and apps are silos out of Google's reach (though they're trying to combat this with app indexing). Paid clicks are growing rapidly, but that seems to be more from them squeezing the pennies out of their desktop properties. There will be a natural limit to the amount of ad inventory they can fill it up with as well.

Facebook isn't having the same problems monetizing mobile. I haven't looked too deeply into their results, but last I checked their ad margins are rising rapidly. Facebook ads seem to be a lot more valuable to advertisers than a Google ad, and they have lots of runway with a lot of premier real estate on mobile devices (Instagram, WhatsApp).

If the "post-pc" trend continues and people increasingly do their computing on tablets or phones, the future doesn't bode well for Google.

The smaller screen space would show up in mobile revenue (which is not broken out by Google, so we don't actually know whether that's happening), but not in mobile CPC. CPC is "cost-per-click", so if fewer ads are shown per page, there will be fewer clicks, which might even drive CPC up if only the highest-quality ads are shown. CPC as a metric is usually used as a proxy of the value of a Google ad to an advertiser; because of the auction system, an advertiser bids only what the ad is worth to them.

Facebook has a very different ad strategy on both desktop and mobile. They target brand awareness much more than purchasing intent. As a user, I'm starting to see Facebook ads in my newsfeed of sites I never searched for or Liked on Facebook; AFAICT, they're using cookie data from the advertiser's website itself to inject the ad into your stream, so that you're reminded of the product on a regular basis. I can see why this would be worth a lot to the advertiser (targeted brand advertising!), but as a user, it is creepy and annoying as hell.

I'm aware of what CPC is. Screen real estate limits the display ads and are much less compelling on mobile.

> apps are silos out of Google's reach (though they're trying to combat this with app indexing)

They also own AdMob, which is one of the more successful in-app-advertising platforms.

For me the difference between pricing for mobile/desktop is mainly on the display side. I pay less for mobile due to lower awareness effect. Across different companies I have noticed that every click on a desktop display ad is accompanied by around 3-4 additional awareness driven clicks. Whiles on mobile there are very few additional clicks driven.

On the search side I am fairly indifferent.

How do you attribute the awareness-driven clicks to the previous desktop search? Is this at an aggregate level, or at a user level?

Not just Android, but AdMob. AdMob consistently beats every other mobile ad network I've tried for my apps by a large margin.

Reply on news.ycombinator.com