Teradata's bet on subscriptions is paying off

At the opening keynote of the company's annual industry analyst summit, COO Oliver Ratzeberger displayed a chart showing share prices up almost 40 percent over the past year, and or almost double over that of a couple years ago. Wall Street likes the fact that Teradata's subscription business is ramping up much faster than expected, accounting for over 60 percent in Q1 vs. the 40-50 percent that was expected. Mind you, Teradata is pulling off this transition while remaining publicly-traded, unlike midsized software counterparts like Tibco and Informatica that felt the need to go private.

Beneath the surface, Teradata has made a number of changes to its business that have facilitated the transition to subscription. Many of those pieces were announced last year.

First is sharpening its target market. As we noted back then, Teradata formally declared that it was limiting its fire to the top 500 opportunities with the most demanding analytic needs; it acknowledged the reality that it never made sense for Teradata to compete for business better suited for Redshift or Snowflake.

The next element is platform portability, addressing demand, not only for cloud, but also the reality that commodity hardware is getting sufficiently powerful to become "good enough." So, while Teradata will still sell you an appliance, its prime focus is the Teradata Everywhere strategy. The same subscription can be applied, regardless of whether you deploy Teradata on commodity or specialized IntelliFlex infrastructure, on-premises, or in the cloud. A key building block for Teradata Everywhere work is simplified pricing, using the TCore metric that is based on CPU and IO bandwidth.

The hybrid, portable play is the obvious one for on-premises incumbents faced with emerging cloud analytic platforms that are vying to become the next-generation default choices. For that world, there is considerable daylight between Teradata and Oracle, in that Oracle is clearly optimized for the Oracle Public Cloud rather than all the others.

Regarding cloud, for now Teradata is on the public cloud. It offers service through a few data centers in the US and Western Europe, but the prime focus is AWS and Azure; Teradata will add Google Cloud when enough customers ask for it. As for managed private cloud, watch this space. Teradata does have specialized, software-defined infrastructure, IntelliFlex, that provides the technology. As IBM, Oracle, and Microsoft are already planting stakes with managed private cloud offerings, we expect that Teradata - with its elite target market - will provide an option there. It's not a matter of keeping up with the Joneses, but instead, acknowledgement that among top customers, there will likely be scenarios where policies rule out public cloud.

There's also another shift in the works that in the short run is fairly modest, but promises bigger returns over the long haul. Last fall, the Teradata Analytics platform was announced. Teradata is shedding its former data warehouse branding because it is expanding its reach to big data and AI workloads and not restrict the developer to SQL. For now, the changes to the product are incremental. They include support of more Aster analytics functions in the mother ship: machine learning, graph, time series, along with temporal capabilities. Teradata kicked off this process last year that we expect will hit the most popular of Aster's library of over a hundred functions.

But there's more basic change in the offing. Teradata has disclosed that it will start supporting R and Python programmers natively rather than going through SQL - which was how Aster handled it.

Teradata also stated that future support for Anaconda libraries for the R and Python crowd, Spark, and TensorFlow will be in the offing. Teradata is hardly the first to follow this path; for instance, Microsoft has added in-database capabilities for R and Python over the last couple versions of SQL Server. We also expect that there will be connectivity with the notebooks that have become the favored development environment for data scientists. In so doing, Teradata, like SAS, realizes that it must meet data scientists in the tools and environments where they live.

TechRepublic: How to get a job in cloud computing: 10 skills to master

Containerization will be the key to making this all work, as the resource consumption footprints for R, Python, Spark, and TensorFlow workloads are likely to be far different from those for running SQL queries. In some cases, that also means supporting heterogenous hardware. Given that Teradata is expanding its focus on AI, what's missing at this point is support for GPUs, but we expect that will be a matter of time.

Teradata has also disclosed that cloud storage is in its future. It reflects the reality that we identified at the start of the year: cloud storage is becoming the de facto data lake. And while cloud storage was designed for just that, necessity being the mother of invention has driven providers like Amazon to support direct query and extended data warehouse integration to S3, while Hadoop vendors rapidly embrace cloud storage in place of HDFS.

In so doing, Teradata is positioning itself as the more concurrent, curated, and managed alternative to Hadoop and cloud-based dedicated AI analytic services. Teradata distinguishes itself in supporting hundreds or thousands of concurrent users - that's something where Hadoop and Spark fall short, as they were not originally designed for this scenario.

So, it's certainly more than coincidence as Teradata expands its aspirations to handle multiple analytics workloads, that its primary voice has become the person who ran arguably the largest and most diverse Teradata implementation while at eBay. Singularity was a multi-petabyte data warehouse at eBay project that dating from around 2010 that was one of the first Teradata instances to accommodate log and other forms of data. Teradata Analytics platform is clearly Descends from that concept.