Almost everyday of the past 12 years I’ve looked at web
analytic data. It’s never been a completely satisfying experience. Log files
don’t paint a vivid picture and javascript’s palette of data points is often a
mess. At the end of the day, it is all “just numbers.” It doesn’t tell me what
to do. It doesn’t do it for me. It doesn’t let me shape it as I want to.
Reports are the output of modern web analytics. Sites are layered with tags on top of the page and tables are created for the data to sit and be compiled. There have been recent advances in reporting itself, namely solutions like Chartbeat (one of the great startup names) that provide realtime data streams. This leapfrogs data delivery over the hours of latency that exists in the site analytics space. But even realtime analytic data is still just a report, only a little more interesting by being timely. The Achilles heel of all site analytics continues to be the ability to mashup datasets from the report tables, an essential first step for segmentation and thus actionable analytics. Not only is this useful for look backs but it also delivers the flexibility needed to create new measurement tables.
Whether it was intentional or not Site Optimization tools emerged a few years ago and began to fill this gap. While they were used to optimize website performance they were marketed as Testing tools. Who could argue for the need to test? Their liftoff was an exciting time in the market. Initially these tools paid great attention to the content targeting aspect of the product, the object. This was done not to the neglect of the product but to the simplicity of early versioning. More and more though it became clear that segmentation was what would really exponentially improve the application of the technology.
The past few years Web analytic tools have evolved to some degree and added more robust segmentation ability (AdWords being the ultimate segmentation creation and reporting tool). As everyone involved in optimizing the web in one way or another knows, it is the keystone to success. I do believe however that success for these solutions are still limited by the underlying data collection and warehousing methods that ultimately affect their ability to make the analytics actionable.
To be more specific, as in other applications, there is limited flexibility in data collection approaches and tables requiring a top-down approach to creating data sets that form the building blocks of intelligent segment creation. Top-down is the classic breaking down of systems to look at the sub systems compositional structure. It informed software engineering until the rise of object-oriented programming in the mid 90’s. For the most part it is the way web analytics works today.
Like the exciting rise of the OOP paradigms that now power the web, at Yieldbot we are really exciting about new bottom-up approaches to data set and segmentation creation possible using these same paradigms. We believe data needs to be as flexible and realtime as the web itself. There is great power in the ability to take data, make it a sub-system of the platform and enable it to be shaped based on the goals of campaign. Just as the semantic web has opened up metadata with bottom-up RDF and microformats we will see a new wave of analytics will bring the same mashup ability to data. We’re already doing it.
As if that’s not enough incentive for bottom-up data approaches, as more and more data inputs/outputs are API generated (another area current web analytic solutions are lagging) the ability to take bottom-up approaches become even more mission critical. The best example of this is Display advertising. The speed to power new markets using APIs driven data are beyond showing the potential, they are generating returns for the DSPs, Exchanges and RTB using it.
Interestingly, unlike the data sharing advances in media buying, the analytics of advertising is spreadsheet wasteland with scarce little BI. One wonders how this situation exists when it is marketers and advertisers that need segmentation most of all. Segmentation breeds efficiency. Segmentation delivers relevance. Segmentation elicits emotion. Segmentation and twin sister matching sit in the middle of all that’s relevant in digital. Over the next few years it will increasingly become the difference of success or failure. Who’s matching you? And now the new question, how are they doing it? Bottoms up!
