Machine learning in Marketing
One of the areas where machine learning has had a tremendous impact already is marketing. With the transformation of the enterprise into an online entity and with the gradual transition of marketing channels from offline to online, the abilities of ML to affect marketing have greatly increased over the last decade. Traditional marketing relies on more general principles – like the famous 4P principle – product, price, place, promotion, where by addressing all of these elements you have successfully assessed the company’s marketing needs. There are many more similar thought models – like the later extended 5P, where you add people to the mix as a key variable, as well as many other marketing models, that still focus on simplifying reality to manageable portions (often divided further to even smaller portions), that we can address as part of our marketing effort. While it is great to rely on such concepts that offer an useful framework, we feel the game has changed substantially with the combination of Internet and the modeling capabilities of machine learning.
Today’s marketing campaigns are not run like they used to be. While some marketing channels have been kept – e.g. print advertising, TV advertising, major new competitors have arrived in the form of social media advertising, google ads, youtube ads and influencer engagements. On top of that there is the capability of approaching customers in many other digital media, like e-mail, SMS, IM, online chats, through product placement in games, etc. Importantly the majority of these new channels offer several significant changes as compared to the traditional channels. For once, the contact with the user is usually very direct, consequently the measurement of the result is also more reliable. Secondly, the contact is much more personal. Airing a commercial for millions of viewers is just not enough for many products and services now, especially the ones who target niche markets. You would rather have targeted access to potential buyers than shooting blind to the general public and expecting to get results. Third, you now have the means to quickly combine multiple sources into singular campaigns. This way instead of having several campaign lines, offering several messages through print or perhaps TV channels, you now have the capability to approach the user with essentially the same message at the same time through multiple channels.
When you add the capabilities of machine learning though, things become even more interesting. Understanding your data, your product segmentation, your market segmentation, the behaviour of each segment and evaluating the potential effect of changes to the market provide material opportunities. Do you really need focus group tests for months, when you can first assess your expected marketing campaign effect by AI models and then run a test campaign on a small subset of real users? You can do it faster and with more reliable results. Do you really need to worry about how each of the Ps of your 4Ps would be affected specifically when you can test the final effect of your campaign directly as end result? We are not saying the old techniques are not useful, they certainly are – at minimum they create a very useful thought framework for the marketing in general. However, we feel ML techniques and new marketing channels come with their own set of rules and understanding them would be key for successful marketing in a fast evolving and more complex world.
Here are some marketing trends that we feel will become more and more important in the near future.
- Personalization. The new marketing channels allow specific message, style and even price to be displayed to each user. They also allow information to be analyzed with machine learning techniques, that could make a decision on selecting the best message, style and price for each user. This gives marketing an unprecedented ability to target individual users or market segments. This segmentation allows for getting the right message for each user, so that the campaign can achieve better cost-benefit. We believe this trend will increase significantly in the future.
- Active segmentation. ML tools allows us to do segmentation analyses, run individualized campaigns within each segment, check campaign results almost immediately and then adapt behaviour accordingly. The marketing cycle is now quick, the results are easier to measure and effective improvement measures could be taken in a very short period. This allows the marketing departments to run simultaneously a set of different and highly targeted campaigns – in essence portraying different messages to different users depending on the segmentation decision. Results for different groups can be compared so that the marketing professionals have the ability to adapt based on immediate results and run the next campaign. Marketing becomes an area of continuous improvement and ML tools are a key driver in that process.
- Multi-channel access to users. While the idea of multi-channel marketing is not new, the combination of segmentation and personalization on one hand, with the ability to identify the user based on social media IDs, e-mail, behavior, etc., allow running increasingly integrated marketing campaigns. Importantly since such user preferences and past choices affect search engines, youtube search, social media feed, etc., then marketing can take advantage of that and offer the same specific message through all of these platforms. Multi-channel marketing now has the ability to target each individual with a consistent message, while messages for various segments remain different.
https://blog.hubspot.com/marketing/4-ps-of-marketing
https://corporatefinanceinstitute.com/resources/knowledge/other/5-ps-marketing/
https://builtin.com/artificial-intelligence/machine-learning-marketing