Return to Blog Page
How can marketing agencies leverage the recent progress in artificial intelligence to grow sales and cut costs? Let’s talk about it. The initial idea to think about is that artificial intelligence is all about data, and having data, especially proprietary data, is the main driver of a good AI project. As a marketing agency, you have access to vast stores of data, which you carefully protect and defend. This same data can generate valuable insights for your clients, including customer segmentation, ad spend optimization, keyword selection, and many others.
Lookalike models can help your clients to identify prospects that look similar to existing customers, and optimize discount offers. The format of this data may not be text and tables. You may find that you have to choose among a variety of images to run with, and that selection can be assisted using a recommender system, trained on past ad performance in your vertical. The models that do this kind of work are typically custom, as your data is structured to best suit your own goals and attributes. There isn’t a one-size fits all solution that can ignore the vertical of the client’s product or service. There are limitations too. Artificial intelligence doesn’t tend to have a sense of humor, and so sometime the deep play on words and nuance of an ad is simply lost in translation. For that reason, we look at applications of artificial intelligence to the marketing vertical as enabling humans, rather than replacing them.
The value added by a custom AI model is the actionable insight you get from it. Let’s look at customer segmentation with AI as an example of unlocking value in your data. Customer segmentation is about mapping out the customer base using transaction records according to certain key features, like recency (latest purchase), frequency (e.g. purchases per month), and lifetime customer value (often estimated by the net present value of the customer account), and other hand-engineered features. Common features to add for segmentation are geographic (e.g. zip code of customer address, and the stuff we know about that area), demographic — such as the estimated customer age from their purchases, spend per month/year which is a proxy for socioeconomic status, behavioural features, which overlaps with frequency somewhat, and can include factors like how often they log onto a website to look at items, and finally, there are psychographic features like customers who prefer to purchase only wine and not beer. We may want to have several features for each aspect mentioned above, such as monthly purchases but also purchases by season, because people and businesses often exhibit seasonal patterns like closing for the winter, or having a big summer tourism season.
Clearly, there is a difference in how you segment customers per-campaign, and that’s where you can decide to build out a suite of artificial intelligence models that best fits your data and clients. Lemay.ai is specialized in this area of data science and custom artificial intelligence model development. Come by and say hi!