Building the Right Digital Measurement Infrastructure: Setting the Table for Big-Data

0
193

Share on LinkedIn

In my last post, I described four challenges to creating a robust digital measurement infrastructure in a real-time environment: tag-based governance that demands exhaustive pre-planning, poor intra-page data capture, lack of a robust data model, and long load latencies that make real-time decisioning impossible. Celebrus Technologies (formerly Speed-Trap) has a technology that tackles every one of these challenges.

First some back-ground. Celebrus has been around almost as long as Semphonic. Founded in 1999, they are based in the U.K. and while there’s a pretty good chance you haven’t heard of them if you are here in the U.S., their technology is good enough to have become a key ingredient in SAS’s Customer Experience Analytics product (probably the most advanced solution in the market today) as well as Teradata’s digital measurement strategy. Because of our focus on high-end customer analytics and digital segmentation and personalization, I’ve been working with Celebrus for several years now and have referred them into several of our database marketing partners.

The Celebrus technology has suffered, I think, from lack of a category. It is not a Web analytics solution. It’s not a reporting solution or an analyst’s work-bench or a statistical analysis package. It’s specifically targeted as a measurement infrastructure. In some ways, it’s like a Tag Management System (TMS) for the data warehouse. You supply the end-user, database management, and analysis tools. Celebrus supplies the data. Until recently, that left them with a tiny market since virtually every large enterprise was meeting their digital measurement needs via a Web analytics solution. With the growth of digital data marts, there’s suddenly a new reason to think about Celebrus.

As with my last post, my goal is to lightly touch on the core issues here. For a deeper account, I strongly encourage you to download the free white paper.

Let’s start with governance. Celebrus provides a solution similar to a basic TMS – a single line of code that is dropped on every page. Unlike a TMS, Celebrus doesn’t handle multiple tags and doesn’t provide a GUI for customization. While I’ve likened it to a TMS for data warehousing, I admit that’s partly misleading. It covers only a piece of TMS functionality. Indeed, if you live in a tag-laden environment, you might choose to drop the Celebrus tag onto your pages using a TMS. That being said, Celebrus offers up much more robust data collection that a standard Web analytics page tag. The system was originally built to measure client-side Web performance – so it features true asynch data capture of nearly every client side event – without any customization. This gives it a significant advantage when it comes to governance. Celebrus greatly reduces the amount of pre-planning you need to put into your tagging – something a TMS really doesn’t do.

That robust data capture also guarantees that Celebrus provides immediate access to key customer re-marketing events and data. You don’t need to customize the tag to track forms or internal search or DHTML. It happens automatically. Some of the most interesting customer targeting and segmentation data is routinely lost because Web analytics tags are so poor at tracking intra-page events. Celebrus isn’t quite as comprehensive as, for instance, Clicktale. They made a conscious decision (and I think a good one) not to include mouse movement data. It’s designed to work in a high-volume environment and capture nearly everything you might actually target on – not to provide the sort of GUI analysis at which Clicktale excels. This is truly a data-capture system that makes it much less likely you’ll every have to go tell your stakeholders you don’t have the data to answer their questions or meet their requests. Never having to say your sorry is the ultimate goal of a good data capture system. Of course, there’s a trade-off here too as well. Celebrus captures a lot of low-level data and while not all of it lands in your warehouse, you’ll need people to structure and make sense of it. I believe that’s a reasonable trade-off for a truly advanced enterprise.

Which brings me to the problem of structuring all that digital data. While capture and data model are theoretically separate, in reality, they are bed-fellows. When you capture data, you have to store it somewhere and you have to some kind of model. Web analytics data feeds provide an interchange model, not a usage model. Celebrus does much better. The Celebrus system comes with a pretty substantial data model – one that includes excellent support for real-time decisioning. It’s open, it’s adaptable, and it can help you bootstrap a data warehousing effort in much less time than might otherwise be the case. This makes it a lot easier to actually get a data warehousing solution up and running. Is the Celebrus model complete? Not really. I’m actually working on a second Semphonic-only white paper that will delve into our data modelling framework based on our Two-Tier Segmentation and which I view as a companion-piece to the Celebrus white paper. No data model is every really finished, but at least with Celebrus you get something that is well on its way.

Finally, there is real-time. Celebrus does an outstanding job supporting real-time. With very low latency collection, a data model specifically designed to accommodate true real-time decisioning, and a set of tools that can move you into site personalization almost immediately, this is one of the true strengths of the product. Keep in mind, this isn’t about real-time reporting or real-time analysis – it’s about supporting real-time customer personalization decisioning. Celebrus has done some deep-thinking here (because real-time decisioning is very challenging) and has a built a solution whose trade-offs make eminent practical sense.

Yes, given the sunk cost of the multi-year investment in tagging infrastructure at most of today’s large enterprises, it’s easier and less work to source a data mart from a Web analytics solution’s data feed than to implement a whole new collection system. And depending on your needs and the balance of Web analytics tool vs. data mart, that may remain a viable long-term path. If you see yourself shifting heavily into the data mart space, however, that path looks considerably less attractive. Web analytics tools are expensive if their only role is data collection. They present governance challenges that often demand another layer of complexity and expense in a TMS. They provide inflexible and, around intra-page data, rather poor data capture. They saddle you with the necessity of building, from scratch, a data model for digital in your warehouse. And, of course, they essentially preclude any possibility for real-time or near real-time decisioning by introducing a huge and essentially unavoidable latency.

If you’re goal is to source digital data into a warehouse, Web analytics tools may be very convenient in the short-term just because of the investment you’ve already made in tagging. They aren’t the technology stack you’d choose if you had a clean slate. We’ve all gone down technology paths because they were convenient even when we knew they weren’t ideal. But warehouse initiatives are expensive and highly visible – they are the wrong place to hit a dead-end because of your sourcing mechanisms.

Check out the entire white paper here

Republished with author's permission from original post.

Gary Angel
Gary is the CEO of Digital Mortar. DM is the leading platform for in-store customer journey analytics. It provides near real-time reporting and analysis of how stores performed including full in-store funnel analysis, segmented customer journey analysis, staff evaluation and optimization, and compliance reporting. Prior to founding Digital Mortar, Gary led Ernst & Young's Digital Analytics practice. His previous company, Semphonic, was acquired by EY in 2013.

ADD YOUR COMMENT

Please use comments to add value to the discussion. Maximum one link to an educational blog post or article. We will NOT PUBLISH brief comments like "good post," comments that mainly promote links, or comments with links to companies, products, or services.

Please enter your comment!
Please enter your name here