Posted on: Monday 12th of January 2015
This is a condensed version of a talk given by Reuben Binns at the 31st Chaos Communication Congress in Hamburg, Germany in December 2014.
The web is immensely valuable, but many of the free web services we use every day are funded by advertising. As well as being the cause of our current concerns about privacy and surveillance, this business model is also a surprisingly inefficient way of achieving its supposed purpose; that is, matching consumers with products and services they actually want and need.
Targeted advertising – the kind of advertising that funds much of the free web services – is based on tracking down consumers who might be susceptible to a message, making them aware of the product, and nudging them to buy it.
The value to consumers, so the theory goes, is that they get more relevant advertisements.
So does this system actually work?
We do know that only about 56% of ads served over the internet are ever actually ‘in view’ – defined as being on the screen for one second or more. And that doesn’t include those that are technically ‘in view’ but ignored, or blocked by a browser plugin. Then there’s click fraud, where fraudsters generate automated fake views and clicks on adverts, in order to cheat marketers out of their money (depending on the type of advertising, between 11% and 52% of impressions are fraudulent). Finally, the consumer profiles that they use for targeting are often riddled with errors. So it’s unclear how well this system actually works at influencing consumers.
Unfortunately there are also some major problems for consumers themselves.
First, the online advertising infrastructure funds and encourages the mass collection of personal data, which in turn facilitates surveillance by governments and crooks. For instance, a data broker might accidentally sell consumer profiles to criminals. The advertising technology infrastructure itself also means scammers can deliver personalised ads, specially designed to trick particular users to click on adverts that result in malware attacks.
Aside from the risks of surveillance by spies or criminals, there are potential ethical implications. The whole point of profiling is to categorise and discriminate between different people; what looks like helpful personalisation could also be harmful discrimination. And it is driven by algorithms which are difficult to challenge and hold accountable. There may also be a subtle psychological effect. In the 1970’s, robotics researchers made a surprising discovery: people like interacting with robots that have human features, but only up to a certain point. If they get too life-like, we get creeped out – they called this point the ‘uncanny valley’. Some of us are starting to wonder if there’s an equivalent effect when it comes to personalisation on the web. If targeting appears to know us a little too well, we get freaked out.
An alternative approach
So what’s the alternative? Is there a better way to match people with the goods and services they want and need?
I think there are a few ideas worth exploring.
First, It’s worth looking at how large organisations, like public authorities and big businesses go about finding the right suppliers and buying the right goods and services. They don’t sit around browsing the web, waiting for a relevant targeted ad to pop up. They put out a request for proposals, specifying exactly what they want, with criteria like service provision, quality, and in some cases even social and environmental criteria, inviting potential suppliers to bid for their custom. Suppliers have to compete against each other, sometimes via a reverse auction, to demonstrate that they are capable of delivering the service or good at a competitive price.
In fact there are already a few services aimed at ordinary consumers to make personal requests for proposals (such as FluBit and Autobutler). These services have one important difference with the digital advertising model. They allow the buyer to initiate the process and set the agenda in the lead up to the sale. Instead of the seller asking ‘who can I find to buy my product?’, it’s the buyer asking ‘who will sell me what I’m looking for, at this price, on these terms?’
Second, as an individual, my demand for a good or service doesn’t give me a great deal of bargaining power with a supplier. But if I club together with others, and we negotiate with suppliers en mass, we have much greater power. This has worked quite successfully in home energy markets. Working through trusted intermediaries like local councils, charities, or even religious communities, consumers can entice energy suppliers to compete against each other for their custom. This system can work well for the suppliers too, because they can cut their marketing and customer acquisition costs. This idea has its roots in the industrial revolution. Worker’s co-operatives, who had recently discovered how to increase their power in the labour market through collectivising, decided to try organising their everyday shopping that way too. They formed Co-operative wholesale societies in order to arrange bulk purchases, and, where possible, organise production.
Another mechanism worth considering is the ‘assurance contract’, a term economists use to describe the kind of threshold pledge system that’s used by sites like Kickstarter. Individuals pledge to put their money into something but only if a sufficient number of others do the same. If the threshold is reached, the purchase goes ahead, if not, it doesn’t. It’s a way of unlocking latent demand that might not otherwise be visible from just observing what consumers browse and buy.
The final piece of the puzzle, I think, is decentralised infrastructure for these mechanisms. So far, things like intent-casting, assurance contracts and collective purchasing have tended to be run in a centralised way by a single intermediary organisation working in a silo focusing on one area, each making their own proprietary software. This model has been successful and will probably continue for the near future, but it places a lot of power in the hands of these new intermediaries. In the long run, these mechanisms can only reach their full potential if they are based on open protocols, where power is distributed amongst, and negotiated by all of the stakeholders.
So, combine all these different elements together, and you could have a decentralised, privacy-preserving, consumer-driven system for unlocking rich data about demand and matching it to supply.
Let’s take a moment to imagine how someone might use this kind of system in practice. Perhaps I want to buy a new laptop, with certain hardware specifications features. I would anonymously broadcast my personal request out to a distributed public network. From there it can be crawled by sellers or third parties acting on their behalf. Any sellers who think they are capable of providing a product that meets my criteria can bid to fulfil my request. But my request could also be crawled by intermediaries who could aggregate similar requests together and identify opportunities to create assurance contracts and dynamic buying co-operatives. I could browse through these opportunities anonymously at my leisure without giving away any information, picking the one that best meets my criteria. My request would only become associated with my identity when I begin communicating with my chosen seller.
The system we have now is like a one-way mirror which allows marketers to secretly monitor consumer’s behaviour, targeting them with algorithms whose logic is indecipherable. Perhaps there is an alternative system that isn’t based on the surveillance of consumers, but on the transparency of suppliers, who are encouraged to compete on the quality of their goods and services, rather than on their ability to grab consumer’s attention. A system that’s much more efficient for suppliers because it’s based on proven demand rather than guesswork; and that’s not just about convincing people to buy stuff, but about serving their genuine needs and values.