Measuring for success

Alan Turing, father of the computer, proposed a test for a ‘thinking machine’ where a panel asks a computer a series of questions.

Based on the replies, the panel has to judge whether it is a computer or a human. If 30% of the panel are wrongly convinced that the replies indicate a human, then the computer is said to have passed the Turing Test. You can read Turing’s original discussion here. Success is one way on the path towards a thinking machine.

In the 64 years since that paper was written, no computer has yet convincingly passed the Turing test (there have been some disputed passes), and in any case, the threshold – 30% of the panel – is still very low.

Computers have no aptitude for understanding humans. For example, if you write a brilliantly punning headline for your article, Google’s algorithms will be baffled. If you use Microsoft Word, the grammar checker is frustratingly literal and near-enough useless.

This is all important to remember because if you are writing for a website, you will be judged on various measures – page views, click-throughs and so on. And here begins a problem. The measurements computers (Google Analytics etc) make will be completely accurate, based on the parameters you set for them. The devil is in the measurement and measurement can be a prison.

Measure for measures

If your campaign is there to, say, drive engagement, then you will adapt whatever you are doing to encourage sharing and clicking. You could, perhaps should, give £10 notes away. That would get people’s attention. What’s your budget? Measure it in £10 notes and give them away from your website until the budget has gone. It could be more cost-efficient than running a content plan. It would certainly be quicker. I predict a lot of traffic to such a site and rampant sharing on social media.

Would it achieve anything else? The metrics would show a resounding success. Visits, sharing, engagement, all sorts of good stuff. Everyone else would see someone giving money away º what’s the point? The computers can’t distinguish one from the other.

But many content plans are indeed giving money away. Content plans range from the brilliant (see the results from the recent CMA awards) to the bland, the dull, the awful, the boring.

The metrics

If your plan is defined by particular metrics, then it will become a slave to those metrics. This might not be a problem if your audience is very narrowly defined, but content can deliver far more than that. It’s good to measure – see what’s working and what’s not – but think beyond the numbers.

There are new and exciting ways to share information, like the famous listicles and Buzzfeed animal parade. There is also evidence that some of the most desirable attributes for content, like sharing, could be on the decline. Novelty works well as long as something is novel. When novelty turns to tedium then you will be searching for news ways to get back to those old metrics. Well argued, well written pieces will default as the minimum ingredient for content.

Wise of the machines

Maybe in the future machines can do all this work for you. Starcom Mediavest has a device that, er, deposits content on your behalf – effectively programmatic native content. Will this save you work and be more efficient than the £10 giveaway approach? I would say that is highly unlikely.

Meanwhile, the computers are indeed learning but we are miles from a Terminator-style dystopia, the technology is literally in its infancy. In New Zealand the Baby X project attempts to teach computers how to learn in a human-like way. The interface even resembles a baby. The technology is sophisticated, but the computer learns slowly. Very slowly, as this remarkable video shows.

Google might accelerate such technologies through its acquisition of UK-based DeepMind to create ‘neural’ computers, and have developed a ‘Neural Turing Machine’, essentially hardware designed to learn. The implications for Google’s search business could be staggering – search efficiency could improve with every click as the algorithm learns from you.

In the meantime, while you might be able to measure engagement, you cannot measure impact. Not yet. Legendary adman Leo Burnett said: “The public does not know what it wants… there is no sure way of finding out until the idea is exposed under normal conditions of sale. If people could tell you in advance what they want, there would never have been a wheel, a lever, much less an automobile, an airplane or a TV set.” Working to the numbers puts you down this path.

Burnett’s advice applies just as well to content. Remember these wise words before you start to look at the audience and usage statistics that drive so many campaigns. If your content efforts are shoehorned to fit an algorithm, your audience is a computer. Whoever they are, your targets will be human, not computers, for a long time to come.

Editor's pick

Most popular