Over the last few years, marketing automation has evolved from a ‘nice-to-have’ technology to an essential component of the martech arsenal. But what does it mean without a data-driven strategy, and is machine learning the answer?
A few weeks ago, I was at a conference on artificial intelligence where the panel discussed the implications of AI for the workplace. Anthony Painter, one of the speakers, had an interesting response to the question of whether AI would put us all out of a job.
“Hopefully, yes”, he said.
His point was that we shouldn’t be afraid of the concept AI doing our jobs for us – that’s what they’re designed to do. Reaching a stage where AI did everything we need for us would be a utopia, not an apocalypse.
Philosophical interludes aside, many would argue that automation is a major step on that journey. In a marketing context, it allows practitioners to focus more of their time on high-level strategic thinking and worry less about the execution. Today, automation software can be applied in myriad capacities across sales and marketing, with top-tier platforms offering everything from lead scoring and segmentation to social media scheduling and auto-SEO.
With a wealth of customer data now available to marketers, including everything from demographics, preferences and website interactions to user clickstreams and social media activity, marketing automation has helped marketers put data to work in new and interesting ways .
Rules are made to be broken
But rules-based marketing automation has an inherent handicap: it’s only as smart as the human operating it.
Take a key use of marketing automation: audience segmentation. Marketing automation systems  can be applied to CRM data to split customers into segments based on things like on-site behaviour, demographic data or stated preferences.
However, the rules which determine the segmentation are chosen by the marketer, meaning they rely on human assumptions about which data points are worth looking at. This leaves room for faulty assumptions, and doesn’t allow the marketers to take the whole data set into account.
What’s more the structured way data is collected limits the potential for nuanced analysis. It can be difficult to accurately segment based on factors that doesn’t correspond to one of the pre-defined fields – e.g. by business size, income or education level – but that might be significant. One author described this as a ‘two-dimensional ’ view of customers.
Enter: machine learning
Machine learning could be the solution. Through an analytical process called ‘clustering’, machine learning can look at a full set of a customer data, identify patterns and organize it into ‘clusters’ of similar data. The advantage of this is that it doesn’t take into account the marketer’s assumptions about what data is important – that information is determined by the analysis instead. This leaves the door open for trends and connections that might have been missed by analysing individual parts of the data one at a time.
Another benefit is that it enables predictions to be in real time. For example, a machine learning system might find that customers within a certain demographic, who has visited 3 or more of your product pages are twice as likely to purchase. Combine this insight with marketing automation, and the prospect could be converted by sending them an ultra-relevant deal or offer at the optimal point in their customer journey.
A further example might be optimizing email send times. A marketing automation system could split test email sends at different times of day. A machine learning algorithm could then take the resulting data on opens and click-throughs, combine it with historical data and alter the next email send based on the results. Over time, the campaign would self-optimize for success, without relying on the marketer drawing insights from the data and actioning them manually. Sound utopian yet?
Machine learning has other interesting applications within marketing, too – such as churn prediction. This uses an algorithm to compare new customers to existing ones in the database. The rationale is this: if similar customers have ‘churned’ in the past, the new customer is likely to churn also. The system takes the full data universe into account when making the comparison, meaning factors which may not have been perceived as relevant (and therefore ignored by the marketer) may be revealed as significant.
Sales enablement is another use-case. Also utilizing CRM data, this technology is designed to deliver sales teams the most relevant content for each sales opportunity, helping the to ensure the buying process is as frictionless as possible.
Top-end software will tell you which sales content is being used most frequently, by whom, and for what purpose. It will also analyze the data to reveal which content has driven the most revenue, giving you a sense of how effective your sales documents are and where there are opportunities to optimize.
Machine learning is a hugely exciting prospect for the marketing world, with massive potential for improving the efficiency and effectiveness of teams and campaigns. Which means it’s critical that marketers work with their automation vendor get the most out of the technology.
Much has been made of the fight between mobile apps and the mobile web. But the line between the two is no longer as clear-cut as it used to be with the introduction of Progressive Web Apps and Android Instant Apps.
Chatbots offer retailers a cheaper, smarter and more efficient way of engaging customers, helping brands deliver everything from 24-hour customer service to personalized product recommendations – all without human intervention. So which retail brands are using chatbots, and how?