Artificial intelligence opens up new possibilities for optimization in digital dialogue marketing. AI processes help to identify the context in which a customer is currently using the product and to select the appropriate communication for this context.
This works even if the data quality is poor, because self-learning algorithms are becoming increasingly better at interpreting even erroneous data or supplementing missing data that is necessary to recognize a context. The most important use case for AI in digital dialogue marketing, or the use case in which optimization by AI provides the greatest uplift, is multivariate testing.
Using AI, it is possible to test a theoretically unlimited number of factors and factor characteristics of a campaign and to optimize the campaign for different segments, automatically and in realtime.

Even in marketing, everyone is currently talking about artificial intelligence. And according to an Adobe study, 87 percent of marketing experts share the opinion that AI should be used more in marketing. No wonder, because according to a study by the Boston Consulting Group, 85 percent of executives, managers, and analysts think that AI brings competitive advantages. However, use cases are currently mostly limited to a few areas: Chatbots, retargeting, and website/shop personalization are the predominant use cases. Digital dialogue marketing can also benefit from artificial intelligence. The promise to bring together the diverse data obtained through tracking to form a 360° customer view and to derive relevant communication from it has been made years ago. So far however, it has only been implemented to a limited extent, if at all, by most companies. One reason for this is that such optimizations require a high manual effort as well as a comprehensive data view. AI methods open up new possibilities here to create highly differentiated, highly relevant communication without ongoing effort and with only a small amount of data.

AI helps to identitify contexts and adapt communication

As stated in Trend 1 (Realtime Customer Centricity), truly customer-centric, digital dialogue marketing must adapt to and respond to the customer’s current usage context. But how can a context be identified? What data in what form (e.g. metadata from websites, the customer’s location or the weather at that location) must be available in order to be able to conclude that the customer is in a certain context in which he might be receptive to a certain information? Marketers should establish an archetype for each relevant context, describing what data in what form suggests that context. A simple example:

  • Time: weekdays 06:00-09:30
  • Device: smartphone or tablet
  • Location: in motion
  • Web use: access to online shop

The context behind this is most likely “Customer is sitting on the train on the way to work and spends their time browsing through the online shop”. After creating the archetypes, a self-learning AI mechanism can be trained to recognize contexts from a variety of different live data. It should be noted that in the rarest cases

  • all data necessary for the clear recognition of the context
    are available,
  • all data is available in the required quality,
  • the combination of the different data and their characteristics
    is as unique as in the example described.

A suitable AI can, however, learn to supplement missing data, interpret data of inferior quality, and even derive the correct context from an unclear data image with high probability. The self-learning algorithm continuously improves itself by incorporating results from tracking customer reactions into its calculations. It is even possible that insights are gained that have not yet been considered when defining the archetypes. If no realtime location data is available in the above example, the algorithm could instead conclude from the fact that a customer surfs their smartphone in the morning at typical commuter time and, from the typical start of work, uses online services from a desktop PC that is not their private PC (recognizable, for example, by IP address or browser configuration) that the customer is on their way to work when they surf their smartphone on the train in the morning. The algorithm automatically uses such new findings to improve its results.

More efficient multivariate testing

The recognition of contexts and control of the appropriate communication is not the only application for AI in digital dialogue marketing. Multivariate testing is one of the most important use cases. For example, the more sophisticated an email campaign is designed, the more assumptions have to be made about the various factors of the campaign and the possible characteristics of these factors: Which subject line leads to the highest opening rate? Which offer leads to the highest click rate? Which type of approach best makes the customer read the email? At what dispatch time is the customer most likely to notice the email? Multivariate tests are an established procedure for optimizing campaigns with various possible combinations of factor characteristics (for example, five different subject lines, three types of address, ten offers and 20 delivery times).
The problem is that the number of different test variants and thus the complexity of the procedure increases with the number of factors and factor values tested. Classic marketing automation already elevates the complexity to a level where no marketer can control the multitude of variants manually anymore. The complexity increases even further when different customer segments are addressed. Here it is no longer a question of determining the best variant overall, but rather the best factor characteristics for individual segments. In addition, it is usually not known which segment characteristics are decisive for the respective test results. Is offer 1 clicked because the customers are women over 40 or rather because they come from Düsseldorf? Moreover, the answer to this question can typically only be evaluated after the test has been completed, so that the learnings can only be used for the next mailing. Another problem is that previously unknown segments cannot be taken into account or it is questionable whether the correct criteria for segmentation have been chosen at all. These problems can be solved by artificial intelligence. AI procedures independently recognize correlations between factor values and segment characteristics and automatically form segments that are as granular as possible depending on which segment characteristics work best with which factor values. The recognition and optimization already works during the campaign runtime, i.e. immediately and in realtime. Content is adapted in realtime for the newly identified segments or even individual customers, if necessary. If a customer opens an email one hour after it was sent, they may receive different content than if they had opened the same email five minutes after it was sent (see also Trend 1 “Realtime Customer Centricity”).

Additional fields of application

There are many more use cases for the use of artificial intelligence in digital dialogue marketing and there are lots of ideas how AI could be used for specific use cases. These include, for example, dispatch time optimization, recommendations, lead-nurturing, or the automated creation of content. Here, at this point it should be left at that. Marketers should find out which possibilities already exist to optimize digital dialogue marketing with the help of AI or which of their use cases can be further optimized by using AI. In general, this could apply to many use cases where procedures such as marketing automation or data analytics reach their limits with increasing complexity.