Websites, call centers, mobile apps, email, ad click data—customer interactions with these channels all generate a tremendous amount of valuable data. The challenge lies in mining this valuable event data to turn it into actionable information to improve marketing, sales, and service outcomes. Doing so requires the right infrastructure, data management techniques, and applications to extract meaningful signals from cross-channel interaction data.
Today’s analytic tools are adequate for analyzing customer behavior within individual channels, but not across channels. Analyzing customer behavior across multiple channels usually requires custom engineering work within a big data environment and a standardized approach to holistically understand cross-channel customer activity. To accomplish this, some companies are building a cross-channel event repository in Hadoop that sorts data from multiple channels by user, with both batch and real-time versions, each designed for different use cases.
The goal of omni-channel marketing is to take all the data from each event and channel, put it in a form to be able to view the entire customer experience, report on it, and then optimize and personalize the customer experience. One key best practice to accomplish this goal is to join event data as it comes in with enrichment data. Enrichment data could be information about the person (age, gender, etc.), geographic data or product information associated with them for personalizing the user experience in real-time or for future interactions. This data can also then be joined with other information, such as weather data as an example. Weather data could be used to decide if the customer or prospect might be interested in buying a swim suit or a parka, which could also be relevant for use later in making personalized offers.
Another important best practice for successful event analytics is to properly manage user IDs. User ID hierarchy mapping by channel is often used to be able to join data from multiple channels to get a complete picture of each user regardless of how they interact with your company. The goal is to collect multiple kinds of cookies and user ids, and resolve identities and update an event repository as identities are learned.
User-ordered data is stored by user, which requires determining the highest user ID. All events of the same user are seen as a line, sorted by time with each channel interlaced. User-ordered data allows you to analyze the paths users take to get to various web pages or sections of a website, or how they move through various channels. Marketing attribution (path to conversion) is another important application of user-ordered data, analyzing emails being sent, ad clicks, postings on Facebook, etc., and then a conversion event. All of these user histories help determine which interactions result in a conversion and which ones don’t. You can then assign weights to channels to determine where to focus resources.
The call center is a good example of where user-ordered data can be a big benefit. Let’s say a customer calls in because they are having a problem using the website. After learning their identity, the call center operator can immediately see what the customer did most recently on the website to help them resolve their problem. Personalization is another use. By knowing what the customer did most recently on a website, for example, their experience can be personalized in real-time.
With custom engineering work within a big data environment and a standardized approach, an event repository in Hadoop can successfully be used for cross-channel reporting, path analysis, channel attribution, customer journey visualization and other descriptive analytics applications to improve marketing, service, and sales outcomes.
Watch our webinar on using Hadoop to build a single repository for logging customer relationships across multiple channels.