The growth of automated decision making is explosive. But when algorithms act as the role of gatekeeper between consumers and brands, how can marketers ensure their brands are still part of the action? Tracey Porter investigates.
This article originally appeared in The Truth Issue, our October/November 2018 print edition of Marketing magazine.
Andrew Burt, a chief privacy officer at data management platform provider Immuta, refers to them as the silent failures. That’s how the legal engineer by trade describes what he sees as the biggest issue facing a world where increasingly more people are reliant on the use of algorithms to make critical decisions. Asked earlier this year whether he believed algorithms could be trusted, Burt claimed that as people began to rely more on complex algorithms, their makers’ ability to explain their inner workings will get progressively harder.
This isn’t simply because these models are hard to interpret, rather because the networks they’re being connected to are becoming more convoluted.
While researchers have long documented the many occasions in which algorithms outperform humans – including beating doctors and pathologists in predicting the survival of cancer patients, occurrence of heart attacks and severity of diseases – every day the world of IT gets harder to manage. “We have more endpoints, more data, more databases and more storage technologies than ever before,” Burt says.
“I believe our biggest challenge lies in being able to understand the data environments we are relying on. Because if we don’t, there’s a very real possibility that we’ll be constantly confronting silent failures, where something has gone wrong that we simply don’t know about, with very real and potentially devastating consequences.”
Some may accuse the Yale Law School fellow of being alarmist. But from a marketing perspective, the replacement of digital search mechanisms in favour of new interfaces powered by artificial intelligence (AI) – such as messaging, chatbots and voice activation – means customers are increasingly indifferent to the traditional branding efforts that influence buying decisions.
Adam Winchester, the global head of projects at Melbourne-headquartered digital consultancy Human Pixel, says the ‘new marketing environment’ presents marketers with a vast array of potential platforms, add-ons, modules, SaaS (software as a service), PaaS (Platform as a Service) and various other scripts and cloud offerings to choose from. Not only is this ‘third space’ able to help them to understand their market, it can also help increase content velocity, increase or understand ROI, improve various metrics and target correct avatars. As a result, algorithms now control much of the advertising and retargeting we see and sometimes even the price we pay, he says.
“The internet and the continued growth of online shopping gives us the ability to experiment with the algorithmic study ‘at pace’ of people, behaviour, reputation, geography and many other metrics, which all allow a narrow AI system to make assumptions, or even real-time changes to tip the scales in their favour – to engage with consumers in the right way, at the right time with the right price in order to increase the changes of a sale. A good algorithm can mean the difference between a buyer or a browser.”
Accenture Interactive ANZ managing director Michael Buckley agrees the increasing use of algorithms has changed the game substantially for marketers and that many customers have removed conventional online retail and bricks and mortar stores from their paths to purchase.
Buckley says tech giants Google and Amazon – through their all pervasive AI powered voice-activated assistants Google Home and Alexa – have worked hard at positioning themselves as our personal shoppers as much as our personal assistants. This erodes competing brands’ access to consumers to exacerbate Google and Amazon’s market dominance in many categories.
“Amazon is entrenched in the way people shop, particularly in Europe and the US. What people are doing when they search for something is only using Amazon as the search engine. As a result, Amazon has created a whole advertising marketplace, because there’s so much traffic on search results. A lot of marketers are now not just trying to include their products as part of the Amazon marketplace but they’re actually advertising them [on Amazon].”
The claim is backed up by findings by ecommerce data firm Jumpshot, the research of which shows that Amazon in the US now owns more than 90% market share across five different product categories. These include home improvement tools (93%), skin care (91%), batteries (97%), golf (92%) and kitchen and dining accessories (94%). In addition, Q1 2018 figures show Amazon was also eating up market share across other categories, including men’s athletic shoes (74%) and cleaning supplies (88%).
It seems this is just the tip of the iceberg when you think about the all-important personal data that firms like Amazon can exponentially glean from their users. Earlier this year Amazon filed for a patent that included an algorithm that would allow an Amazon device, such as its Echo smart speaker, to listen to a conversation and analyse it for certain words. Referred to as a ‘voice sniffer algorithm’, the technology works by building a personality profile on the user. The algorithm uses positive trigger words such as ‘liked’ or ‘loved’, ‘prefer’ and ‘bought’ or negative trigger words such as ‘hate’ or ‘disliked’.
The device can then capture additional audio that can then be analysed for keywords, gauging interest levels in various products. The patent application states that the “identified keywords can be stored and/or transmitted to an appropriate location accessible to entities such as advertisers or content providers, who can use the keywords to attempt to select or customise content that is likely relevant to the user”. If the patent is approved, Amazon could then offer “personalised offers on products, encourage [a user] to take action or better persuade someone to buy a product”. The data could also be made available to friends of the user for gift buying, according to the patent. It has not yet been approved by the US Patent and Trademark Office.
Those interested in the technology say the move to analyse conversations as a means to discern users’ interests may amplify Alexa’s intelligence and that this algorithm could eventually feed from Alexa into the rest of the Amazon consumer product offerings. The belief is that it could ultimately help “drive purchasing and buying behaviour of [Amazon] Prime members”. Those against the move argue the patent may also allow video cameras on devices to capture image information to attempt to determine which user is speaking – the ultimate data capture or privacy invasion depending upon which side of the fence you sit. Amazon argues that there is provision in the patent that a user “can have the option of activating or deactivating the sniffing or voice capture processes, for purposes such as privacy and data security”. It also states that users must indicate a “willingness to have voice content analysed” for the trigger word algorithms to work.
“We take privacy seriously and have built multiple layers of privacy into our devices. We do not use customers’ voice recordings for targeted advertising.”
Google has submitted patents for a similar technology, but has defended its position to US media outlets by stating that “all devices that come with the Google Assistant, including Google Home, are designed with user privacy in mind”.
“For Google Home, we only store voice queries after a physical trigger or after recognising a hot-word trigger like, ‘OK, Google’ or ‘Hey, Google,’” the company has told media.
But while that type of technology may not be quite here yet, Buckley argues the challenge now is how marketers can somehow get their brands to supersede the Amazon and Google marketplace and in doing so invite themselves into customers’ homes.
Buckley ponders, “I have an eight-year-old boy and if his remote control car runs out of batteries, he will just say, ‘Hey, Alexa, can I order more AA batteries?’ He, like most consumers, doesn’t request a particular brand of battery, for example Duracell or Eveready. So instead what arrives in the mail is an Amazon-branded battery. Is the eight-year-old as the consumer happy? He’s ecstatic, because it solves the problem of his non-working toy car. The paradigm then becomes how do Eveready and Duracell break into that market?”
A key to the answer, says Buckley, is being on the spot with personalised content. “If you start to order milk through Amazon, how do you make sure that it’s the brand that you love and not the brand that Amazon recommends? You’ve got to find a way to actually get your customers to say ‘I want product A’ and not just allow Amazon to recommend it for you.
“I think everyone has now recognised that their brand needs to become part of the home. It can’t just be the brand at a retail store or online. The brands that do get into the home are the ones that will win.”
Human Pixel’s Adam Winchester says it’s possible for marketers to influence the way these algorithms work to ensure they are in their clients’ favour, but to do so they must rely on data and trained systems to read and understand this data.
“Big data and the rise of business intelligence provides marketers with the tools to get the formula right. If you really understand your perfect target customer through buyer persona creation and behaviour mapping, and you understand the triggers that end in a sale, then with enough data and the right tools, today’s marketers can create the ‘perfect storm’ as it were, to supercharge any marketing campaign.
“The power of AI is the ability to sort through massive amounts of data and find trends that can be used to deliver extremely relevant and timely messaging, advertising or content generally to the right person at the right time and sometimes in the last millisecond when dealing with someone browsing a website.”
But the AI is only as good as the research behind it, he says.
“Think about the case where Target figured out a girl was pregnant before she had told her father – in this case, Target had done painstaking research in determining triggers in its marketing. AI will help marketers find these trends and act upon them in the right way in order to help their customers find the right products just before they need it.
“Like anything, though, poor planning results in poor results. So all of these tools are for naught if the marketer doesn’t put in the right amount of the right type of work in the beginning.”
For Winchester, finding the right tools to work with their business, and getting a large enough dataset to use in these tools to use them effectively is paramount to success. Before any machine learning or AI is implemented in a business, the business needs to know what to train the system to do, he says.
“Without actual data to study (and lots of it), it’s basically guessing.”
What many marketers don’t realise is that most of the technology doesn’t do “exactly what it says it does on the box”.
Winchester says from problems such as integrating with existing systems, or being told that the product works in a particular way when it doesn’t, companies are faced with a dizzying complexity of tech to negotiate in order to get their products, first, in the right context and, second, transacted smoothly with the right logistical back-up to get into customers’ hands with minimal delay and fuss. As a result, the current marketing environment – while offering amazing opportunities for companies – also presents extreme risks.
“Much of the enterprise level kit can cost millions and years to integrate into existing systems while the actual impacts to the organisation may not be known until after the integration,” says Winchester. “This means that companies risk losing months or even years in a market that changes on a weekly basis.”
* * * * *
To purchase a subscription to Marketing magazine, visit the online shop »
* * * * *
Image credit:Zulmaury Saavedra