Date

By now you’ve probably noticed that most people, most of the time, simply don’t click on ads. That’s a pretty easy conclusion when you and everyone you know ignores (or blocks) ads almost all the time.

And the numbers back you up: the average AdSense click-through rate is under 1%, with roughly 5% of those clicks leading to "conversions". In other words, Google has to display roughly two thousand ads to generate a single conversion for their clients. That works out to an efficiency of 0.05%, which is mind-bendingly wasteful when you think about it.

You may have also noticed that the situation hasn’t improved much over the years. Online advertising has never been efficient, and you wouldn’t be the first to think that something is amiss when 0.05% efficiency is normal here in 2016 after 20 years of online ads.

I’d like to share my hypothesis as to why this is:

AdSense is merely 0.05% efficient because Google has every incentive to keep it that way. Put another way, Google’s business would be severely diminished if AdSense was much more than 0.05% efficient.

How’s that for a perverse incentive! Are they fighting to keep it this way?

It certainly seems so when nearly everyone would be happier with a huge efficiency improvement but Google simply does not and cannot care. They’re making money hand over fist and it would be stupid of them to kill their Golden Goose.

Why, though? Why should we expect no efficiency improvements from Google?

Imagine a world...

Begin by sitting back in your chair and then take a slow deep breath. Close your eyes to imagine an online experience where the few suggestions you see reflect exactly what you want in the moment. The suggestion for ice cream comes a day or two after finishing your last pint, on a sunny day, for a flavor you like, on sale, and near where you already go. Yummy. The suggestion for a house cleaning service comes only after your home has reduced itself into a total mess, at a price you’re more than happy to pay. Handy. Imagine a suggestion coming in that you and three friends are available and interested in going mountain biking on Saturday, with nobody having to text anyone. Take a moment to envision that wonderful fantasy. I’m not sure you even can!

In this fanstasy every suggestion you saw would help you get what you already wanted, except sooner, more conveniently, and perhaps at a lower than expected price. You would no longer be bombarded by dozens of irrelevant, distracting ads. Let’s consider this heretofore unfathomable fantasy to be a 50%-efficient connection-making system. A thousand-fold improvement, in other words.

When we imagine a 50%-or-higher connection-making efficiency, we’re imagining a digital system that knows pretty much everything about us, all the time. The amazing predictive ability needed to "read a person’s mind" to the point of making perfectly timely suggestions is totally dependent on the quality of information the system has to work with. When this connection-making system has only low quality information, its predictions will be bad and its suggestions will fall flat (or even cause harm). In stark contrast, when the system has extremely high quality information the predictions flow freely with high accuracy and the connections it suggests are valued.

Consider Carol, who fancies herself a cupid and knows both Alice and Bob very well. If Carol carefully considers their personalities, quirks, and interests and still imagines they’d make a great couple, her carefully crafted suggestion-filled introduction--at the time both Alice and Bob are single and looking--has a surprisingly good chance of them connecting wonderfully. This scenario involves high quality information: Carol knows Alice and Bob very well and puts in the effort to construct a situation that has a high probability of leading to a connection. As you might expect, successful connections like this are few and far between and worthy of note. It isn’t out of the question for careful Carol to see a 50% or higher success rate.

When cupid Carol has only superficial knowledge of her targets or puts little effort into her introductions, she would be lucky to inspire any good connections at all. This is the low-quality-information scenario we experience so often in our daily lives, with online advertising in particular. When Carol knows very little or thinks mostly of herself, her efforts are likely to fail repeatedly. Not quite as low as 0.05% efficiency because she is not a robot, but low enough to be told to cut out the constant cupid shit because it’s a huge waste of everyone’s time.

The best information is privately held

Transforming the fantasy of 50% efficiency into reality would demand the impossible of Google: they would have to put the self-interest of people like you and me above their own self-interest. That’s never going to happen and we’d be foolish to expect it.

Any amazingly effective connection-making system--the kind the serves the interests of individuals--will depend on high quality information: highly personal, extremely detailed, and continuously updated. How could it work otherwise? While it’s in Google’s self-interest to acquire this type of information and they already do an OK job of it, it’s in the self-interest of regular people to keep the most intimate and detailed information about themselves private. Or else they’ll get taken advantage of.

This impedance mismatch means that the digital information underlying a truly efficient connection-making system will be owned, controlled, and kept private by the individuals involved. Definitely not centralized into Google-controlled databases. Many people hold cavalier attitudes about their digital privacy so far, but there’s a limit and also potential for push-back.

As far as I can tell, Google’s revenues and stock price are based largely on the power it can derive, now and far into the future, from its massive database of continually-updated personal information about billions of people. When we say "information is power", we’re describing their business model.

If Google were to develop and release a high-efficiency AdSense replacement, its design would have to involve decentralized information capture and storage, thus loosening their grip over information ownership and compromising their source of revenue and power. If Google were to thus legitimize a decentralized approach to connection-making, it would find itself on a steep and slippery slope toward irrelevancy. With the cat out of the bag, regular people would quickly realize that they could transform their digital information into power for themselves. If there’s one thing you can count on, it’s that people act according to self-interest. Eventually.

We’re stuck with the 0.05% efficiency of Google AdSense because their self-interest is only fractionally aligned with the self-interest of its billion-strong audience. What’s good for them is mostly bad for the rest of us while what’s good for us is mostly bad for them. Google would have to be suicidal to raise the efficiency of AdSense; the rest of us would be wise to create a high-efficiency alternative.