Yesterday I read, The World’s Longest Recruiting Process: Josh McFarland is Joining Greylock as a Partner. My pukoradar began to boot up at the sight of that headline but I ignored it. Then I came to:
TellApart applied cutting edge data analytics and predictive modeling to the field of marketing. In the time it takes you to blink — literally 100 milliseconds — TellApart’s technology can respond to an impression in an ad exchange, crunch thousands of model features and serve the right ad at the right place and the right time. And it does this hundreds of thousands of times per second, with great insight and accuracy.
Really? This is the best you can do? I'm by no means the first person to point out the inanity and insanity of absurdly smart people reduced to doing this. I also freely admit that I used to be one of the flock myself, being fascinated by the potential for algorithmic ad placement, trading, optimisation, and so on (and I still am fascinated in some ways). The parallels with, say, high-frequency trading are obvious and have already been seized upon.
Then today I read, Methbot: a $3M-$5M/day video ad-tech fraud:
White Ops, a security firm, has published a detailed report on a crime-ring they call "Methbot" that generated $3M-$5M by creating 6,000 fake websites to embed videos in, then generating convincing bots that that appeared to watch 300,000,000 videos/day -- running virtual instances of various browsers (mostly Chrome) on virtual machines running MacOS X, from a huge pool of IP addresses that they fraudulently had assigned to US locations, deploying clever grace-notes like limiting access to "daylight" hours in their notional locations; simulating mouse-movements and clicks and more.
It's bad enough that prodigious brains and huge amounts of capital are being deployed towards optimising digital ads. But what stories like this second one tell me is that even the fundamental premise of digital ads is invalid (let alone trying to optimise it).
Page views? That's so 20th century.
Click farms? Old hat.
Automated click farms? Yawn.
Measurement "errors" and over-billing? Surely. Not.
And so we arrive at purely simulated, never-even-existed views and clicks. Ah, yes, let's do that.
A race to the bottom.
Surely we can come up with something better than this? A system that actually creates real value? We're missing the woods for the trees.
I'm sure Josh is a perfectly wonderful human being – and the Greylock blog post goes on to tell you so in nauseating detail – but that's not really what set my pukoradar off.