Telecoms, Media & Technology is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 3099067.

Media + Networks

Where Content Meets IP

Bayesian Inference (or why my 5-year-old son gets recommended Pepper Pig)

Some years ago, the now-discredited Autonomy was putting some wonderful smoke-and-mirror demonstrations out in the market, showing exactly how we could use their technology to make video voice-searchable, or infer recommendations for product vendors etc etc.

While Autonomy may have been masters as selling futures, the reality took some years, and evolution in computing and a couple of very big (Amazon?) successes to really generate the focus, investment and resource to turn these ideas into reality.

Now, as we all know, recommendation engines are a pain, because they flood every clickable promotion slot possible in every web-page we look at with adverts promoting cheaper sources for whatever it was we just bought.

I tend to decide what I want, get onto a search engine or two, compare a few prices, take an emotional ‘punt’ with who I trust to deliver it in good condition, and make my purchase. Often within just a few minutes.

At that point the recommendation engine kicks in from all those cookies and shared datasets, and for the next few days or weeks I see nothing but promotions for the same thing I just bought, but cheaper, or bundled with extra goodies. Totally useless. In fact it makes me resentful. I resent the company promoting the goods, and I resent the company who actually sold them to me. I also resent the fact that this distracting and now-irrelevant data is distracting me from my activities on line.


Somewhere out there I am sure there is someone with a vacuum between their ears, one that NEEDS the advertising to prompt a thought. They start every morning empty, and follow adverts all day, and feel that the content that surrounds the adverts is a pure distraction. They buy instinctively as promotions offer discounts. They return more expensive purchases and buy them again from cheaper vendors. They fill their lives with recommendations.

That is not me. Nor anyone I personally know.  I am not that ad-target.

I’m not sold on recommendations by inference. While it probably means my ad experience is tuned to me – it is yet another echo chamber: it would be much better if I was promoted endless things I had never heard or, or had not yet bought. That would actually be more likely to draw me away from the content that the advert is paying to deliver. But perhaps that is not the aim here.  Perhaps the aim is actually to simply justify how advertisers are pissing away their clients’ money with ever more data to occlude the fact that there really is little return on investment.

While watching the TV the other day, as he gently burned off a fever, my five-year-old boy asked me why his Netflix account was recommending Pepper Pig "because he had watched the nature program life".

Bayesian inference at its best. And worst. If there were two programs more disconnected I don’t know them. One is exquisite – the very highest end of documentary production and focused on the broad topic of nature. The other was Pepper Pig. Only a computer would make the connection between a cartoon pig and a nature documentary.

I did contemplate explaining inference to him. But I found myself saying "you know what dude; Netflix has got it wrong. That’s crazy. You are right. Trust your instincts …Do you want sausages for dinner?"