SXSW: No ethics in analytics

Analytics can help predict hurricanes and oil price fluctuations.
Analytics can help predict hurricanes and oil price fluctuations.

SapientNitro's VP of global performance considers marketers' power over consumer data

Marketing exploits the data of customers because humans are predictable. But we are at a crossroads in the evolution in marketing. Simon James, VP global performance analytics at SapientNitro, sharing his SXSW15 talk, explores the ethics of whether we should be using data analysis to exploit customer habits just because we can.

AUSTIN — It was due to a classical example of chaos theory that Mary Shelley wrote arguably one of the first science fiction novels, Frankenstein. In 1815 Mount Tambora in Indonesia erupted, killing over 70,000 people and causing the "Year without Summer" in 1816, trapping Shelley indoors and into a contest to write the scariest story.

Exactly 200 years later, and we face another moral dilemma. Our desire to connect and share in a digital world means we tolerate the monitoring of our behaviors. If it means one less password to remember, who reads the terms and conditions? Yet, digital is data — the ones and zeros we all leave behind us in a thick exhaust. This data represents tremendous commercial opportunity to the canny marketer. But in the pursuit of these opportunities we are entering a world driven by ethics not statistics. Who decides what is ethical? How do authorities keep up with technology and sources of data that didn’t exist last year?

Data analytics is not infallible

Our mastery over data allows us to predict hurricane paths. It allows us to optimize the layout of "dark" grocery stores — stores that only exist to harvest ecommerce grocery shopping — to minimize the time it takes to process the order. It enables us to determine how price sensitive individuals are, and deliver personalized promotions designed to maximize sales and minimize discounts. Data analysts are highly qualified to tackle these problems. However data analytics is not infallible and algorithms are only as smart as the people who write them.

Analytics in marketing often relies on predicting outcomes. In predictive modeling there are two types of error that are generated, false positives (type i) and false negatives (type ii). For professional sports, drug screening is designed to predict who has taken banned substances. If a sportsperson was tested positive but later proven to be innocent, the reputational impact on the athlete’s career, and subsequent rumors would cause huge financial impact — and open grounds for compensation. This would bankrupt the majority of governing bodies in sports. Consequently there are practically no sports people who were wrongly "proven" to take banned substances. The only famous case of this happening is the former Commonwealth Games champion, Diane Modahl. The unintended consequence of this is that there will be many more false negatives — sportspeople who are doping, but will never test positive.

Positive and false negatives

On the other hand, the UK has one of the largest breast-cancer screening programs in the world. The consequence of false negatives here is potentially mortality. If the screening misses a single case of potential cancer, it would undermine confidence in the screening by making headlines in the press. As a result there are very few false negatives, but a significant amount of false positives — that is ladies who receive a positive test for cancer but are in fact clear. This is the consequence of not wanting to miss a single case. According to research conducted by Cancer Research UK, for every five lives that are saved by breast cancer screening, 17 women are treated for a cancer that would not have caused them any problem. 

A supermarket dark store

Problems of logic vs. moral questions

These are not statistical problems of logic such as optimizing a store layout, but moral questions that cannot be answered by the humble data analyst. The problem with marketing is that it has evolved in an environment without consequences. In the nineties the direct marketing business gave us "junk mail" before it was disrupted by email marketing. Email wrestled the mantle of intrusive media of choice, coining the term "spam" before mobile, social networks and instant messengers gained broad adoption. Today I am hassled by chuggers on the street, PPI claims on my phone, get rich quick schemes on message boards — in addition to the spam and junk mail.

We are now at a point where the consequences of getting predictions in marketing wrong is not the cost of a letter, or an email, it’s the besmirching of a brand in public via social. What was once limited to "white mail," private and limited in scale, it is now immediate, public and with a huge audience. Calm seas make poor sailors. Marketers are ill equipped to deal with the ethical conundrums that represent the gap between what we could do with a consumer’s data and what we should do.

Brands can be proactive

However, there is something brands can proactively do. The way brands treat their customers, their privacy, their security, and limit the use of their data is fast becoming the new CSR (corporate social responsibility). Now a staple of the annual report, a company’s approach to CSR and commitments to sustainability or environmental impact are seen as good corporate governance and a response to demands for greater visibility. But how long will it be before companies see the value in stating their data policy to shareholders, stressing their commitment to consumer privacy and security, and limiting the allowable use of their customers’ data for corporate gain? Surely this fits the agenda of transparency and social governance?

Time to step up

Science has always been guided by ethical boundaries, but these boundaries have evolved over time. It’s time for brands to step up and proactively communicate how they are protecting the data of their customers, and how they are limiting the use of such data. Brands should not be conforming to the minimum requirements of the Data Protection Act 1998 (sounds like a lifetime ago), but striving to the highest possible ethical standards of treating their customers’ data. Data is an asset and should be treated as such.

This article first appeared on