Tools and techniques will largely depend on your own email marketing strategy.
In this article, we’ll look closer at the three key AI techniques that can help you improve the results from your email marketing program.

The opportunities that AI brings for email marketing

Did you that according to RJ Metricsknow the top 1% of your shoppers are worth 18 times more than the average shopper?

This amazing stat begs two questions for retail marketers:
1: How can retail marketers engage the top 1% more effectively?
2: And how can marketers engage all consumers to convert them to top shoppers?

Happily, there are immediate steps brands can take to improve revenues and customer lifetime value. In the past few years, a trifecta of powerful computing, big data, and machine learning have triggered a big shakeup in personalisation.

Why has AI-based personalisation suddenly become so important? Because it works:

• At Netflix, more than 80 per cent of the TV shows people watch are discovered through the platform’s recommendation system
• At Google, news recommendations improved click-through rate (CTR) by 38%
• For Amazon, 35% of sales come from recommendations

Personalisation is now a critical issue. And yet, email marketing often still operates with legacy technologies and segment-driven campaigns. According to E-consultancy, just 17% of brands and 21% of agencies are innovating with artificial intelligence to create personalisation at scale. So there are currently opportunities for marketers to differentiate themselves through an AI strategy.

For retailers and e-commerce brands, personalisation technologies are all about understanding the customer so the product sells itself. Think about our examples from the retail and entertainment industries. From Spotify suggesting a playlist to Amazon recommending products to buy, they first understand each consumer’s tastes then curate content for that consumer.

A Spotify-style consumer taste profile, as applied in email personalisation

Today, shoppers demand relevance and personalisation in order to remain engaged. Statista found that the more personalised the email, the higher the email open and click-through rates. However, when we talk about hyper-personalisation, we’re going beyond the “Dear Emily,” intro and even beyond personalisation per audience segment. Nowadays, data-savvy marketers are using AI to adapt content for each unique consumer.

Nobody phrases it this way, but machine learning in retail is primarily a marketing issue. It’s really an attempt to understand the consumer so we can serve them better.

The 3P key techniques of AI in email personalisation

AI techniques can be used to achieve multiple goals. So the best AI email personalisation technique will largely depend on your email marketing strategy.

Email marketers can use the 3P’s of key AI techniques, namely:

1. Profile-based personalisation
2. Product-based personalisation
3. Predictive Analytics

1. AI for Profile-based email personalisation

Great for newsletters and other campaigns to large audiences

There’s a trend toward more in-depth recommendation systems online. Spotify and Netflix have refined their recommendation engines to create more meaningful engagement with their customers through hyper-personalisation.

For example, Spotify provides each individual listener with a tailored and personalised selection of songs, and while automated recommendations aren’t new, Spotify’s complex algorithm isn’t just based on a user’s saved songs; it builds a taste profile of each user ’s musical tastes.

These entertainment brands have raised consumer expectations. Now retailers must also provide each shopper with curated content – not just based on a user’s last click or purchase, but delve much deeper into their shopping tastes.

Profile-based personalisation is particularly suited to email newsletter personalisation for the following reasons:

1. Results tend to be highly relevant. Because profile-based AI relies on each consumers ‘taste profile’, they are likely to be highly relevant to a consumer’s interests. This makes profile-based AI especially valuable for organisations with lots of products and lots of consumers (think retailers, e-commerce and travel brands).

2. Encourage product discovery. Another advantage of this technique is that email marketers can use the shoppers taste profile to encourage browsing and shopping – even when the shopper doesn’t have a discernible need. By continually showing fresh and relevant content, email marketers can maximise product discovery and revenues.

3. Marketers can get started quickly. Profile-based personalisation avoids the cold-start problem that often bedevils product-to-product recommendation technologies.

While the technology still requires some initial data to begin making recommendations, the quality of those early recommendations is likely to be higher than with a platform that only becomes statistically valid after millions of data points have been added and correlated.

Recommender systems have the effect of guiding shoppers in a personalised way to interesting products in a large space of potential product recommendations.

Profile-based AI systems recommend products based on the attributes of products a given shopper has liked in the past. A profile-based recommender matches the attributes of each shopper’s tastes and preferences, with the attributes of products, to recommend new and interesting items to the shopper.

So for email newsletters, profile-based personalisation is typically the better option. It creates a nuanced picture of every shopper’s preferences, so the marketer can precisely tailor email content to every subscriber’s individual tastes.

2. AI for product-based email personalisation

Great for purchase confirmation triggers and other behavioural campaigns)

Product-based AI is based on the relationship between different products, with no information about the consumers required at all! All you need is a correlation between all of your products.

In marketing, product-based AI is excellent for cross-selling in behavioural campaigns such as purchase confirmation triggers. Users are shown matching or additional products. Amazon is the market leader in this type of AI personalisation.

Early on, Amazon realised that with the right product recommendations, customers buy more stuff.

You can see as a many as five different types of product-based AI personalisation at various points in the Amazon purchasing process

1. ‘Frequently bought together.’
2. ‘Customers who viewed this item also viewed’.
3. ‘What do customers buy after viewing this item?’
4. ‘Customers who bought this item also bought.’
5. ‘Your recently viewed items and featured recommendations.’

3. Predictive analytics

Great for nudging people into action, e.g. preventing churn

Predictive analytics uses historical data, statistical models, and AI to predict the marketing strategies that are most probable to be successful. Predictive analytics includes lots of techniques such as automatic winner selection and next best action.

Sounds terrific – so how does predictive analytics help your marketing strategy and tactics?

Predictive analysis helps to better understand your data and make informed decisions. Predictive techniques aided by AI helps predict future customer behaviours by uncovering patterns in your data.

For example, regression analysis uncovers any correlation between shopper’s previous behaviours to assess the likelihood of future purchase behaviour.

Predictive analysis can also identify dissatisfied consumers who are likely to churn and to identify consumers who are ready to purchase. Analysing customer data in predictive analysis can predict behaviour and help inform your marketing strategy.

Getting started with AI for email marketing is easier than you think

Deploying AI in your email marketing solution isn’t complicated. It doesn’t need any knowledge of Data Science. New AI-based software platforms are now available –  even for businesses that lack maturity when it comes to machine learning.

Because AI is still a new field, it is often best to begin with a Proof of Concept (POC). The POC is designed to prove that the AI actually delivers on your defined marketing objectives.

The POC is used as a pilot program to test a vendor’s technology in an operational environment, for an agreed timeframe. The purpose is to test the AI with your data and use it on a day-to-day operational basis.

Start with a list of possible POCs that align with your marketing KPIs. When you have a list of options, rank them in order of business impact, weighted if necessary, for ease-of-implementation. Then pick the top POC, and keep the remainder for future implementation. Importantly, aim to deliver early successes to build confidence and momentum.

Be sure to choose a vendor that has true AI, not merely a rules-based decisioning platform, which is impossible to scale for the volume of data and combinations of interactions that marketers are managing today. Also, be sure to choose a vendor who understands your industry vertical and who can integrate with your existing technology partners.

The time for AI-based personalisation is now

Personalisation is becoming an essential component of online marketing. This is not only because companies feel under pressure from their competition and have to continually work to stand out from the competition, but also because of shoppers’ raised expectations.

The increased competitive landscape demands that brands consider how to use AI to ‘surprise and delight’ shoppers on an individualised basis. The same data used by brands today can also be used by AI to create unique, relevant, and timely messaging for each individual shopper, not a targeted segment.

With the world’s biggest brands like Spotify, Netflix, Amazon and Google betting big on AI personalisation, we are getting closer to a day where all our email personalisation will be handled by AI.

The post Picking the right AI for email marketing personalisation appeared first on Email vendor selection.

Real-time access to ALL suitable data

Whether the connected consumer shops in-store, in-app, or online, or some combination, she rightly expects brands to recognise her and treat her individually and in real-time.

Google and others have written about ‘moments of engagement’ and there are lots of current opportunities for brands to leverage CRM data, merchandising data, EPOS data, Internet Of Things (IOT) data, apps/mobile data, third-party data, and other datasets. However, the volume, velocity, and variety of all this data can be a challenge for the modern marketer, which brings us to the next big lever: algorithm-based dynamic content.

RFP question:

Which of the following data types can be stored NATIVELY in the email platform? • CRM data • Pricing data • Stock & availability • Instore/EPOS data • Loyalty data • Social data • App data • Geo data • Web behavioural data • Fares & availability • IoT data Please explain how this data can be used, with practical examples based on our business objectives. How soon is this data available for use in segmentation and targeting, after it has been transferred (via API or suchlike)?

Dynamic content using Machine Learning

Every marketer has seen the rise of machine learning over the past five years, but few realise that machine learning can be used to increase email marketing performance by more than 50%.

Self-learning algorithms are now sufficiently advanced so they can mine data and dynamically populate email campaigns with greater accuracy and speed than a human ever can. This is a big step forward for email marketers because now marketers can concentrate on strategy and optimisation, rather than the laborious task of manually building email campaigns.

For example, eBay increased their revenue per opened email by 250% in some countries and 68% globally by using machine learning to insert content into their email platform. Similarly, US retailer One Kings Lane used machine learning to generate a 6% lift in total online revenues.

RFP question:

We’d like to use algorithms to improve email personalisation and speed-to market. Do you offer an algorithm for email personalisation? If so, please provide a non-technical overview of how it works, and how it will help us improve email performance. Please explain the criteria the algorithm takes into account.

So how does machine learning work for email marketers?

The components of a machine learning algorithm can be grouped into two broad categories: User behaviour analysis and content analysis. By analysing a user’s behaviour in real time to track mood and compare this activity to lookalikes, much like Amazon’s “Users that viewed this product also viewed X”. You can also use business rules for content filtering and content selection.

When analysing content, machine learning will analyse similarities in site-tagged metadata. It will also use Semantic Text Analysis to build a “topic cloud” beyond what is manually tagged in the website source. Popularity of content is tracked and used to further improve relevancy of personalized content recommendations and drive user engagement.

Trigger emails

Everyone knows that the revenue per email for triggered programs is somewhere between 3X and 7X that of newsletter campaigns, however triggered programs can be difficult to set up and maintain. For this reason, most brands typically only have a few triggered campaigns in place.

Often marketers start with campaigns like cart abandons and life-cycle programs, but there is a myriad of other triggers that generate huge incremental revenues – and nowadays these types of behavioral email campaigns are much easier to manage, particularly using ‘fire and forget’ self-learning algorithm-based triggers, which can really turbo-boost your performance.

Triggers to consider include price drop alerts, new product arrivals, end of line/back in stock notifications, in-store promotions – to name but a few.

RFP question:

We see an opportunity to further enhance our triggers. Please outline the types of triggers your platform supports, and how quickly fully-personalised triggers can be deployed. (-hours/minutes/seconds). Given what you know about our business, what types of triggers would you recommend initially?

Database growth

Potential customers come from different sources, types of advertisements, and have different customer histories and intentions… why would you treat them all the same? The best approach to increasing list size is to dynamically personalise your site to surgically tailor when and where you get opt-ins.

Intelligently serving sign-up campaigns based on visitor attributes and behavioural actions will increase your email signup rate by 3X-5X. Suggested strategies for increasing email signups include:

Timing: Ask for email before the customer leaves, or after they’ve looked at a certain series of pages.
Tailored based on source: Keep email signup message consistent with display ad clicked, don’t ask for email if the customer came from a marketing email.
Tailored based on current page: Signup request should reflect the users demonstrated interest.
Offer something in return for signup: Many retailers use a generic discount with signup, but often offering unique content can be equally effective.

RFP question:

A lot of anonymous visitors come to our website, visit one page, and then leave. Similarly, many anonymous visitors place an item in their cart and then leave. How would you help us engage/capture email addresses for these visitors?

Begin the journey

Assessing your performance against each of these 4 factors (data, machine learning, triggers, and database growth) will help you assess your current level of competency. Rather than setting a pie-in-the-sky vision like “world-class marketing,” this approach will enable you to assess where you stand in each area, and will help you identify the gaps in your current capabilities and uncover opportunities for improvement.

When you have assessed your current capabilities, of course, the next step is to identify a roadmap for continuous improvement in each of the four areas. This roadmap will help you create better experiences for your consumers, and generate more revenues for your company.

The post Machine Learning & the 4 Levers to Next Generation Email Marketing From the Start appeared first on Email vendor selection.