The Case Against The Surveillance Economy

Here’s a testament to the value of data: if Amazon, Apple, Facebook, Google and Microsoft, with a combined market capitalisation exceeding seven trillion dollars, were to set aside their differences and form a country, it would be the third-largest in the world by GDP. Four of those five companies – Amazon, Facebook, Google and Microsoft – share a common interest and major profit source: advertising revenues.

The digital advertising market tracks most of our daily activities for economic gain via a continuous migration of endless data related to users’ purchasing, social and personal lives. Artificial Intelligence deciphers this data and predicts an outcome, and the tech giants use that information to maximise their growth and revenue streams – kindly sharing it with marketeers who in turn collect even more data on their behalf.

In return, consumers have been given cookie consent forms we all hate when they interrupt our browsing, many of them listing the hundreds upon hundreds of unknown entities requesting access to our data in a Kafkaesque way.

How did we get here?

Overall, as marketers, we believe we are good people – hardworking and well-meaning. We want our clients and brands to succeed, and our campaigns to have a positive impact. We use all this data and AI to guide those goals.

But to work properly, machine learning requires enormous quantities of training data, with no oversight or limit on what was being collected. The trend of big data has further accelerated the phenomenon.

We have been sold the technology to power our surveillance with nice-sounding names like “Data Lake” or “Data Silo”. In the process, we have helped to build a surveillance apparatus so pervasive that to call it Orwellian would be reductive, and we named it “ad tech”.

Ad tech has enabled the largest transfer of wealth in advertising, to the tune of tens of billions of dollars every year, from marketers to middlemen. Around 60% of our combined programmatic advertising budget is siphoned by these ad tech providers.

As far as I can tell, they have enabled modest to moderate changes at most: for programmatic advertising, the overall engagement is usually reported of around six clicks per 10,000 impressions.

So, the question lingers: is all of this really necessary to achieve business performance?

A change of outlook

We all know we need to take user privacy to heart and stop relying on cookies, personally identifiable information, audience data and other identifiers in our endeavours.

We need to switch from the hoarder’s mentality of ‘keep everything in case it comes in handy’ to a minimalist approach of collecting only what we actually need.

But which data do we need? How can we identify the minimum amount of data needed for successfully reaching our business KPIs?

We can start by deciding what we won’t collect. I propose a bold approach:

• We will not collect behavioural data
• We will not collect third-party data from brokers
• We will not collect first-party data that we don’t have a specific use for

But if we do not collect all that precious data, you might ask, what’s left on the table?

We are left with only ephemeral data relative to a single, isolated ad impression. Therefore, how can we extract the maximum value from it?

First of all, we need to open the Pandora’s Box labelled “Data Quality”.

A recent study found that the accuracy of that AdTech targeting is often extremely poor. One experiment used six different advertising platforms to reach Australian men between the ages of 25 and 44. Their targeting performed slightly worse than random guessing.

In the third quarter of 2017, Procter & Gamble cut its digital marketing spend by $100m, with little to no impact on its business, proving that the targeting of those digital ads was largely ineffective.

Despite the extent of surveillance technology, a lot of the data that fuels advertising targeting is, frankly, extremely expensive and well-marketed garbage.

Does AI need personal data to make decisions?

Short answer – no.

If we are coding the AI, we can put our understanding of human behaviour into achieving the targeting objective of our marketing campaigns. By knowing who we want to target, we can reverse-engineer the data we require to reach those people and drive our business outcomes.

Before the digital age, all marketing was based on building the most accurate fictional marketing persona, an archetype of our best or most common customer, to whom we would tailor our advertising. In the digital ecosystem, this mentality has been translated to a 1 to 1 mirror image – a customised persona for every single consumer.

What if we reverted to the old model? What if we managed to drive digital advertising performances by relying only on a generic archetype?

To deliver on the promise, we need to ask ourselves two questions:

1. Do we have enough data from our single, ephemeral impression to identify our fictional persona?
2. By taking the ad tech middlemen out of the equation, does the doubling of our effective marketing budget eclipse the lack of targeted data?

But let’s not talk hypothetically – allow me to give a practical example

For one of our clients, we were tasked with reaching a specific target: individuals over thirty-five, with a high affinity for travel and an income of more than €100,000. The classical approach would have required a significant amount of data brokerage for demographic, historical location and financial data.

We took a different approach: we decided to extract, from the ephemeral data of an impression, enough proxy metrics to be able to reach the target audience without relying on any personal data.

We replaced the financial data with pricing and recency data on all smartphones, therefore training the AI to target only the most recent and expensive models.

We replaced the travel affinity with contextual data, placing the ads only on content related to specific travel locations, further filtered by exclusivity.

This approach generated an uplift of over 30% on conversions.

The brand saw a significant increase in sales, the finance team kept the budget as planned, the IT team removed unnecessary heavy trackers from the website, the marketing team over-delivered – and they could all tell their customers they cared about privacy.

And if better results and a clear conscience aren’t a good enough reason to renounce the surveillance economy, it’s hard to know what is.