The significance of data in digital marketing - part 2

Okay, we have the data, we know where it comes from and how to categorise it. Next, we want to use this data for segmentation. With segmentation, you start from a message and decide who will receive that message.

When thinking about segmentation, most marketers assume that it is only used to decide who will receive a certain email. However, segments can be used in personalisation as well. 

Minimise frustration

Based on the types of data described in the previous blog post, you can refine the contacts that will receive the communication, assuring that only relevant recipients are selected. The difficulty lies in the correct application of the data – knowing that part of the information may be missing or even incorrect – so that the correct number of contacts receive your communication. 

Basically, you want to achieve maximal results (opens, clicks, information requests, sales appointments and/or purchases – depending on channel and your type of business) while minimising frustration by the incorrectly targeted recipients (frustration that might result in churn).

Examples of segments

Historically, segmentation has been done by creating fixed (you guessed it) segments. At first, segments were based primarily on a combination of attributes, geographical data and/or basic aggregated data. Think:

• Everyone subscribed to our newsletter
• All customers in the Ghent region
• All female customers in the Ghent region
• All female customers in the Ghent region having a AAA-score (with that score based solely on historic purchase data)
• All female customers in the Ghent region subscribed to our newsletter

Technology of today allows for even better segmentation, combining the examples above with behavioural and/or location-based data:

• All female customers in the Ghent region having visited our website in the past 3 days
• All customers with a score higher than 40 (one contact that qualifies having 10 points for being a purchase director, 10 points for downloading a white paper and 25 points for having a scheduled meeting with our salesperson)
• Anonymous visitors of our website that look at sneaker products
• Identified female visitors that look at sneaker products on our website
   

Storing segments is useful allowing them to be re-used when needed in future communications and marketing programs. On top of that, it is always possible to add selections to existing segments when the need arises. For example, when there is a special exclusive event in one of your stores, you could decide to use the stored segment of newsletter subscribers, adding the extra selection of including only people in that store’s region, but only for that specific event communication.

Tailor your message to your segments

When communicating with your contacts, it is also advisable to tailor your message itself in function of the person you are communicating with. Once again, this personalisation is irrespective of the channel you are communicating with. For example, with email the content of the mail will be changed while on website visits banners and product pages might change and ‘your store’ might be displayed.

• The most basic way to personalise your communication is to insert attributes. Starting an email with the first name of the recipient is a good way to break the ice. Another example could be the inclusion of the current points on the loyalty card.
• Next, think about showing content based on rule sets. This differentiation can start from something simple, such as showing a different picture for male and female recipients. Another example is to show only certain info based on the type of customer, e.g. showing subscription data only for contacts who have purchased that service.
• Another way to personalise (and already more advanced) is to dynamically add content based on attributes. Newspapers do this in their daily news emails, where the content of ‘in your local region’ is populated automatically with the last 5 articles based on your postal code.
• Behavioural information can be added in your communication. Think about last visited product on the website or a mail with abandoned shopping cart information included.
• Dynamic information allows for the inclusion of information that changes based on inferred data (e.g. online weather info based on place of residence) or a time ticker showing how long the sale period is still running.

Another important instance of dynamic information is the showing of product recommendations based on behaviour on the website and/or based on purchases. Note that the use of product recommendations will even be more powerful when you can show this when visiting the website combined with the inclusion of the same product recommendations (based on the same behaviour) in emails!

And what about A.I.?

The rise of A.I. is happening, and this will also have an impact on digital marketing. What can we expect from Artificial Intelligence in this area, now and in the coming years?

• Send time optimisation is the process of defining the moment that a given contact is most likely to view a certain communication (which can be different across channels) and using this to decide when the message will be delivered. Basically, this will throttle the sending out of the communication (regardless of the channel) over a longer period since contacts will no longer receive the message at the same time.

  The incorporation of this function is already being included in some software solutions. Disadvantage of send time optimisation is that it becomes harder to perform A/B testing because longer lead times needs to be taken into account.

• Channel selection concentrates not on deciding when to send a communication but over which channel. Per contact and per (type of) message, the algorithm will decide what the optimal channel is to maximise the chance the message will be read. As with send time calculation, A/B testing is harder to set up as the message will necessarily be different across channels with possibly different content.
• Product recommendation is a technology already in place for some years but where we see a switch from rule-based calculations to A.I. supported algorithms. This switch allows for more advanced and less error-prone recommendations to be served to your consumers (e.g. websites stating: ‘people who have shown interest in a headphone of 500€ also has shown interest in headphones of 50€’, risking a loss of 450€ turnover).

  Modern digital marketing strategies for companies where product recommendations are useful should include the option to serve a contact the same recommendations across different channels, at the very least on the website and in emails.

• Recipient selection is the availability of an algorithm that – based on a message with certain content – decides which contacts in your database will receive the communication. This selection of contacts can use attribute, geographical and behavioural data.
• Content selection will select articles to be used within the communication based on the attributes, geographical info and the behaviour of the contact receiving the message, so that every sent message will be tailored specifically for the recipient. As such, it is similar to product recommendations but goes further as it will have impact on the complete message.

  The already mentioned inclusion of local content in a newspaper daily newsletter is a non-A.I. (because rule-driven) example of content selection, but with AI the selection of content will become more powerful.

• Content generation, the holy grail of digital marketing. Using Natural Language Generation (NLG), the complete content of the digital communication will be ‘written’ by A.I. Whether it is the content (and layout) of the website, an email or the app, the complete message will be constructed specifically for the recipient and every communication will be unique and to the point for the recipient.