Now, 20 years later, virtual assistants have made a come back – this time without the distracting raising eyebrows (thankfully).
Today, a virtual assistant looks more like this:
AI is set to change the way we interact and work, and has even been described by Google as “The ultimate breakthrough technology”. And while the last ten years have been all about building a world that is mobile-first, “[…]the next ten years, will be dedicated to shifting to a world that is AI-first.”
There is already chatter that in 2018 alone, the investment into AI marketing technology is set to increase by 300%. AI is already powering products and services we use on a daily basis, even if you haven’t yet noticed it.
AI is an ambiguous figure in retail marketing, and with good reason. With much of the conversation focusing on the niche – (if it can beat us at chess) or the distant future (whether it will eventually take over the world as we know it) – little attention is being paid to how it can specifically help retail marketers, and why it’s something to champion not fear.
So how does AI stand to help retail marketers get their job done? The answer isn’t wildly different to the Google example above: far from making marketers redundant, we believe AI will come to be the marketing assistant of the future. This post will explain how.
Truth is, it doesn’t really matter how AI works – predictive algorithms, machine learning, deep learning, neural networks – ultimately, it doesn’t really make a difference. When you ask Siri to find the nearest burger place, Apple doesn’t go out of it’s way to tell you the ins-and-outs of the algorithms it uses to recognise your voice and verify your request.
Yet, when we talk about retail marketing and AI, we overemphasise the complexity of the applied technology and go into unnecessary granularity. This is often why retail marketers miss out on how AI can actually help them, and get lost in jargon and hype. Because they aren’t told.
People want to know whether Siri suggested the right burger place and not how it did it. Retail marketers should care about how AI is going to bring more value to their brand and consistently improve the customer online experience, not its technological inner-workings.
This is one of the common queries that arises any time AI is discussed in relation to retail and digital marketing. No one is going to lose their job; and even in the unlikely case that your job is automated in the distant future, it will be replaced by a more data-informed and creative role.
More importantly, it’s unlikely that there will be a humanoid super-minds that will take over the world. ‘The Singularity’ is a very long while away. Just as amongst ourselves as humans we are better at certain things than others, there will be AI designed to excel at certain tasks.
And finally, and most importantly, AI is not a “feature” – it’s not something that you can turn on or off. AI is a service; think of it as electricity, a grid that you can plug into at any time to gain access to power, or in this case insight and marketing personalisation. It’s not something that you retrofit into your existing tech stack to satisfy your objectives.
Let’s highlight a problem that is already affecting marketers today: data. The number of potential consumer touch points with any brand is growing exponentially. So too is the mound of customer data that retail marketers are sitting on. A decade ago, the average customer used two touch points online when buying an item,today, customers have an average of six (and growing).
Let’s think about the touchpoints you may encounter on your commute into work:
That’s 10 touchpoints. And this is a very simplified run through of a commute into work, excluding any other typical occurrences that could take place enroute.
If this wasn’t enough, customer expectations are increasing with every new disruptive technology and service – and these technologies don’t even have to be in any way related to retail to impact retailers. The fact that we can press a button in an app and someone gives us food or a taxi within minutes is no longer impressive. It’s expected.
So, for retail marketers AI has the potential to help them navigate the overwhelming amount of data that they have accumulating in their stack and use it to apply clear strategy to their campaigns. This is turn leads to customer-specific personalisation and an overall improved customer experience.
The true role of AI in retail marketing is to assist marketers: to help them sift through insurmountable data sets, to understand their needs and objectives and, perhaps most importantly, to enable them to focus on creative activities that improve the overall customer experience.
Assistance, however, requires true understanding. Every brand is different, and no one size of marketing strategy fits all. Some brands are heritage, some modern; some are solely data-driven or creative-led – some both. And there are those who prefer to maintain full control of their communications and have in-house teams, whilst others utilise automation or outsourcing for resources.
And this is why AI can’t be considered as merely an afterthought or feature. It must take into account your brand’s available resources, needs, objectives and goals. Imagine you employ a machine to cluster your database into multiple segments. If isolated as a feature, this algorithm will most likely spit out hundreds of different segments that could only be implemented by massive teams of data analysts, designers and marketers, simply because it doesn’t understand your needs. You want AI that works for you and makes your role smoother.
Here are three examples of how marketers might leverage AI in practice:
Cart abandonment campaign
Your assistant might prompt you with suggestions on how you can improve your existing campaigns, or suggest new ones to run. For example, it can analyse your cart abandonment campaign and automatically create a control group figuring out that sending emails or messages to your loyal customers might actually be worsening the customer experience, as they are likely to come back and buy anyway.
In this example, instead of operating like a ‘black box’, the marketing assistant would prompt you, the marketer, at the right time and make suggestions that are easy to understand – in this case not requiring much additional resource. And most importantly, the marketer maintains full control.
Example of a cart abandonment campaign (using dynamic content)
Another example: Imagine you already have a form of ‘win back’ campaign running. However, the “at risk” period is the same for everyone in the campaign – for illustrative purposes let’s say 26 weeks. The assistant can calculate individually-personalised send times that take into account the initial product purchased, engagement levels, CLV and many other factors of the recipient to best estimate when they are likely to become at risk.
Who has control over these decision? Based on how risk-averse you are, the assistant can either select a small group to start testing on, test on everyone, or make the switch automatically (without seeking your approval first). It can then let you know when statistically significant results come in and give you the option to revert or continue optimising.
It might sound like a small difference but when the assistant is automatically optimising or prompting suggestions across multiple campaigns and dimensions, having a brain of its own and understanding your needs can save you a lot of time.
Let’s continue and imagine you’ve implemented some of the assistant’s suggestions and have some cross-channel campaigns running – but you are concerned that you might be sending too many messages.
You probably have a few other questions: how many messages is enough? Is everyone getting the same amount? Are you sending too many messages on channels that cost too much? What if you want to look at all messages beyond email including mobile push notifications, direct mail, SMS, on-site etc – how does that influence the frequency of your communication?
On the surface it seems pretty straight forward – you have lots of data points on outgoing messages and how customers are interacting with them – surely a machine can easily figure out the optimal individual frequency based on everyone’s engagement.
But then you start asking more questions – or at least you should – how does a machine know which message to prioritise? What is the machine prioritising for? Is it revenue, profit, open rate, conversion rate, unsubscribe rate, repeat rate, predictive CLV? Let’s say you want to optimise the unsubscribe rate, it’s very likely that the machine will stop sending messages all together.
So now you’re probably thinking this is all a bit too complex. And you’re right – but this where a machine should do all the work, prompt you at the right moment and allow you to input key parameters and have enough visibility to monitor the output.
What the machine can actually do is analyse all possible touchpoints and patterns and provide you with several recommendations – for example anyone that has subscribed within the last week is most likely to convert if they receive x emails and x push notifications within a specific time frame. Or a lapsed customer, on average should receive 3 emails and one facebook retargeting ad.
Based on your brand knowledge and strategy, you can review some of these recommendations and decide how much control you want to relinquish depending on what you choose to implement.
So when you think of AI, don’t think of it as taking over the world or stealing your job.
Just as AI is currently changing our daily lives in the background, with minimal disruptions, we should let AI influence the way we market to customers in the same way. Its place will be in suggesting improvements and spotting trends that the retail marketer may not have time to spot, continually optimising campaigns in the background to create amazing customer experiences.