Why Customer Insights Data?

Why should a company choose customer insights data? This is a question I often get, especially when talking to people in the analytics department or with that type of background. “We already have customer analytical cubes in our warehouse/lakehouse, why build it again?”. With this article I will try to shed some light on this subject. I short, it is about making the business more versatile, faster to react and enabling use of large amounts of data in a way that is hard otherwise. It also enables advanced AI features that otherwise would take a lot of time to develop and the UI enables people in the business to experiment and try out things to see what works.

Customer Insights Data (CI-D) is Microsofts Customer Data Platform (CDP). CDP:s are a family of software that aggregate and measure customer data. It is in essence a datalake with some spark engines and a UI to control everything. The important thing to understand is that it is meant, not for DBA:s but for people working with customer data in the business. Some people with deeper skills in how to get data into CI-D are needed but once the data is there, measures, segments etc are meant to be created by “business” people.

Why not just use a data warehouse to solve this? Well, technically, the difference isn’t that large if you look at it from a strictly semantic datamodel perspective. However, the difference lies in who is supposed to be using it. Let’s take the scenario when you are in a B2C business trying to sell computers to customers. If the business wants to have a new measure called “Rate of gaming” where the products on the orderline are categorized as “gaming” or “business”. By calculating the number of order lines that are flagged as gaming compared to the total number of orderlines, for each customer, we will get a “rate of gaming”, a % that tells how many orders are gaming related. This can then be used in email marketing to send newsletters with more gaming oriented content.

With a classical data warehouse setup, the marketing department would order the measure to be calculated by registering at ticket in Jira/Azure Devops or similar. It would go into the backlog and then a few sprints (several months) later, it would be delivered. However, if the marketing department now realized that they forgot to limit this to just the orders placed in the last 24 months, then they will have to register a new ticket and wait several more sprints.

Using CI-D they could create the measures themselves (you have to create 3 to solve this), and then, after the next system synchronization, the measures would be there and working, typically around 24h or less, depending on how fast and often system refreshes are done. If they needed to change anything, like changing the filter, it can easily be done and after the next refresh, it would be live.

This is important for many reasons, the most important being that the marketing departments have to change with the world, and sometimes the world changes very fast. In those cases it can be very important to be able to act fast, to retain customers or to be able to exploit a marketing window that has appeared. Enabling this capability is key in the world of today

Why not use just dataverse? Why not just mirror all the data into dataverse and create a few rollups in dataverse? This is technically possible, not really a good idea for the following reasons;

1. Dataverse capacity is rather expensive. It is an operational database and is supposed to store data that is top be used operationally, not for analytical purposes like a CDP uses data. (with operationally I mean that you work with the data, like working a case in customer service to close it or working with an opportunity to close that)
2. Dataverse isn’t built to handle really large amounts of data while CI-D is as it is based on Delta lake data storage with Spark engines to do the heavy lifting.
3. Dataverse doesn’t really have a good way of doing measures. Yes, you can use rollup fields but these are very limited in functionality and you are limited to a few.

Experimentation is also a key way of working for a well functioning marketing department. As someone who like the scientific method being a MSc EE, I like this data driven approach of making hypothesis and trying them out to see what actually works and not. This does, however, require systems that enable the marketing department to experiment and measure outcomes. With experimentation times being weeks with traditional data warehousing and 24 h with CI-D, having CI-D is a true enabler to make a well functioning marketing department that can experiment as they see fit.

Machine learning and using AI in general is also something that is key today. For example CI-D has built in functionality to find similar customers to specific segments. If you start with a segment of all customers who bough gaming products in a computer sales, then, finding other customers that are similar from a data perspectiv who might also be interested in gaming products, can be a very valuable segment to approach with personalized offers.

Exporting selected segments to external systems for activation in these, is also a key functionality in CI-D. This unlocks some of the most low-hanging fruits like not marketing to some groups of people as they would never buy a particular product. With ad banners costing both actual money but also trust from customers when they are irrelevant, this is easy money to save. For instance, not marketing a gaming PC to a customer that just bought a gaming PC despite Google seeing that they visited that gaming PC:s product page.

Using measures and insights from CI-D outside of marketing can also be very powerful. This can be in scenarios like customer service when a customer calls and the customer service rep immediatly sees the total purchasing amount for business IT, consumer IT and gaming IT as well as the time between PC purchases, age of current PC and more. This can help the customer service rep not only be a better help to the customer but also be more relevant in recommendations.

There are many scenarios where CI-D give true value, but it is a bit more of a strategic product than CI-J which is a bit more tactical in giving faster wins. However, the long term wins of using a competent CDP like Dynamics 365 CI-D can be very large and I hope I have made the case for that clear.

Original Post https://powerplatform.se/why-customer-insights-data/