How to avoid the data discrimination trap
- 25 October, 2018 07:23
An accusation of using data to discriminate a potential audience is the stuff of nightmares for marketers. However, without the right safeguards and processes in place, it’s a very, very real possibility for most out there.
If you can keep your data clean and your moral compass pointing in the right direction, you can avoid accusations coming your way.
With Facebook making news because its platform has been used to target discriminatorily, sharp focus has been brought upon how brands are using data to target the right people for their brand, without facing accusations of ulterior motives.
To ensure you’re doing the right thing for all concerned, you need to look at the data itself firstly. Is it painting a true and accurate picture?
To start with, you need to know how the data has been collected, Institute of Analytics Professionals of Australia (IAPA) managing director, Annette Slunjski, says. “It may have gaps. It may have been collected selectively, or in a such a way that the data set is actually skewed towards a particular belief, whatever it might be.
“If that’s the case, it doesn't matter how unbiased anything you do with it is, the data, to begin with, is already biased.”
Power in your hands
Once the issue of data integrity is addressed, focus comes onto how it’s used. And while marketing has always been regarded as part art, part science, the balance is quickly tipping towards the latter.
Marketers today need a strong understanding of data and how to interpret it and, ideally, have firm rules and regulations in place to ensure a consistent and systematic approach.
Robust governance policies around what data should be used for, how it should be analysed and which data sets should be used together are increasingly common in analytics teams. These policies should guide the use of data and ensure a brand is using it consistently.
There’s still a human element, however, which needs to be overlaid to ensure brands are doing things with integrity, and with common sense. Because even if data is clean and free from bias, it can potentially rule people out of subsets incorrectly – and possibly leave you open to accusations of discrimination – if taken at face value. This is simply by virtue of the fact there’s too much of it.
“You can take targeting too far, to the point where you have profiled a customer to a place that isolates them from a greater offering, and you, therefore, eliminate your brand from the opportunity to sell them the stuff they either didn’t know they needed, or we didn’t know they wanted,” Catch Group, CMO, Ryan Gracie, tells CMO.
In order to prevent this, Gracie suggests running parallel programs to ensure your marketing activity and data are balanced.
“It’s important to maintain a program of mass or more generic marketing that builds brand and forms associations with products from a diverse range for when that person comes into market for that product, placing your brand on the consideration list,” he continues.
“We sell over two million products. If we were to target only based on recent browsing behaviour and then target based on this category propensity, we would create a self-fulfilling prophecy by selling them more products from the targeted category, but in fact, we narrow down the wider sales opportunity.”
PwC senior marketing manager, Peter Hionis, says it all comes down to knowing your audience and getting the right products and messages to them.
“It’s got to be appropriate for both parties. If you’re Rolls Royce and you’re actively targeting someone who you know earns $15,000 per year, it’s probably not of interest to them, and it’s going to reflect poorly on your brand,” he says.
“Depending on the platform you’re using, you can potentially give context with some upfront messaging around why people are receiving that offer or product, letting them know this is why we think this would be good for you.”
In the marketer’s hands
Data discrimination is rightly a hot topic, and as referenced earlier, has been somewhat triggered by complaints about Facebook. In August, Facebook removed 5000 options that can be used to exclude religious and ethnic minority groups by advertisers on its social media platform in order to try and minimise misuse.
The key thing to note, however, is the complaints raised against Facebook are about enabling marketers to discriminate in their targeting. This still needs a willing marketer to take advantage of the loophole available to them. Ultimately, the responsibility is the marketer’s.
“In all of the conversations around how data is used, I keep coming back to the fact that the marketer is ultimately responsible,” Slunjski says. “What data and analytics allow you to do is to make a better decision, rather than you making that decision by gut feeling.
“But marketers still do need to make that decision. It's not just that the computer says yes, or the computer says no.”
The litmus test is around harm, Slunjski says, “and whether the decision you're about to make, be it through analytics, through an automated system, some machine learning or artificial intelligence, is going to cause someone harm or distress.
“There are a lot of things you can do with data, and that’s different from what you should do. You have to have that ethical element in there as well.”
To paraphrase Spiderman, access to all-powerful data creates great responsibility.
Misuse it – whether that’s with malice or ignorance – and it could have serious consequences. Not just for you as a marketer, and not just for your brand’s reputation.
“That consequences can range from someone just not getting a particular offer, to changing the course of their life,” Slunjski adds.