Interview: Teradata CTO Stephen Brobst

Teradata has been recognised by Gartner as the leader in its field thanks to a partnership strategy that aims to ensure its customers benefit from the best tools on the market

Teradata, a data warehousing and analytics firm that counts such giants of commerce as eBay, Wal-Mart and Dell among its customers, has recently been named as the industry leader in its field by analyst Gartner.

Gartner’s “Magic Quadrant for Data Warehouse Database Management Systems” places the company at the top of the pile in terms of ability to execute and completeness of vision, ahead of competitors Oracle and IBM.

“Someone asked me how much we paid Gartner to be at the top,” says Teradata’s chief technology officer (CTO), Stephen Brobst. “Teradata is a data warehouse platform. We don’t have an ETL [extract, transform, load] tool, or a BI [business intelligence] tool, or even a data mining tool. I can go to Oracle and they’ve got everything. How can Teradata have outperformed their products?”

According to Brobst, Gartner’s accolade is the result of a difference in strategy.

Best-of-breed partners

“Teradata’s strategy is that we want best-of-breed capability. No one vendor will be best of breed at everything. Our completeness of vision is the result of choosing best-of-breed partners,” he explains.

Teradata does not own a data mining solution, but it partners with business analytics software developer SAS, which does. Brobst says that SAS is widely recognised as a best-of-breed company in data mining.

“We want to be a best-of-breed rather than a stack player. IBM is more or less the same as Oracle, in this regard. They have the hardware and the database, they acquired Cognos for BI, and SPSS for the data mining.”

Brobst explains that his strategy is to fill the gaps in his portfolio with the best offerings from other vendors in the BI and analytics industry. And this approach can even extend to putative competitors, what he terms the “co-opetition” model.

“If you want to be best of breed, it means you can’t draw battle lines,” he says. “If we think SPSS is in the top three of data mining tools – and we do – then we will integrate with it.”

Most enterprise BI and analytic tools work to an ANSI (American National Standards Institute) standard, which also means they will interoperate with Teradata’s tools.

However, Brobst explains that this is not the same as the “deep integration” in which his company specialises.

“There are three approaches: deep integration, certification and ‘it will work’. When we see best of breed, we choose deep integration.”

Deep Integration

“Certification” means that the products have been tested together, and that the involved parties guarantee that they will co-operate, which is a step up from a pat on the back and a blithe “Don’t worry, it’ll work.” But deep integration means that the vendors are in partnership, and their product road maps are intertwined.

“Deep integration means I’ve met with their CTO and we have agreed what we’re going to deliver. We implement stuff to make them successful, and they do likewise. You can’t do this without partnership.”

He adds that SAS code runs inside the Teradata database.

“No other database does that. [Our competitors] claim integration with SAS, but that’s just at the certification level. Deep integration means you have engineering teams working together, not just combined marketing.”

He adds that what he terms the “stack companies” integrate their marketing and pricing lists with partners, but not the actual engineering. He argues, somewhat controversially, that the Oracle Data Integrator tool is more effectively integrated with Teradata than it is with Oracle itself.

Interview: Teradata CTO Stephen Brobst

Teradata has been recognised by Gartner as the leader in its field thanks to a partnership strategy that aims to ensure its customers benefit from the best tools on the market

But it is not always possible to perform a deep integration with every tool on the market, in which case the firm needs to decide where to invest its engineering resources. Regular tests and evaluations must be performed, not just on new products, but on long-standing solutions as they evolve and improve.

“A few years ago, I would have said Microsoft Analysis Services [part of Microsoft SQL Server] was a toy, unfit for the enterprise. But Microsoft has recently made significant investments and a lot of progress, and we’ve promoted them to what we would deem a best-of-breed tool. We now integrate deeply with them,” says Probst.

Start-up company Tableau, a developer of a data visualisation tool, was also spotted
by, and ultimately partnered with, Teradata. Again the strategy is one of partnership and deep integration, and not acquisition.

“People ask when we’re going to acquire MicroStrategy or Infomatica,” he says. “The answer is we’re not going to, it’s not our strategy. We’re not trying to be Oracle or IBM.”

The data tsunami

Brobst believes that rapidly increasing volumes of data will be a game changer for the analytics industry. Traditionally, analytics has been about transactions, such as the purchase of a flight ticket on the web from a firm such as Travelocity. A more sophisticated approach, in Brobst’s view, is to analyse not just the transactions but the interactions.

“What were the clicks that led up to the purchasing of that plane ticket? Moving from transactions to interactions leads to enormous data growth,” he says.

But how does this add value?

To illustrate the point Brobst gives the example of telecommunications companies and the way they analyse call detail records. Every phone call you make on your mobile will be stored in the data warehouse of your network provider, but the network provider now stores not just the call record (time and length) but what happened when the call was made.

“While you were driving down the M25, there may have been lots of interactions between your mobile phone and the network.”

Operational support systems (OSS) data at the network level would record how many times a call was made and what sort of response the call got, for example. Alternatively, billing support systems (BSS) data stores call detail records and helps companies understand the value of the customer because it records every billable event, but it does not reveal the customer’s experience.

“You don’t see that they were in a traffic jam and they tried to access the network five times and got a ‘busy’ response. You will see that in the OSS data, as well as the locations of all the dropped calls. That data matters. It answers the question: ‘Where should I build the new cell towers?’”

Brobst argues that moving from an analysis of transactions to analysing interactions is about really understanding the customer experience, not just the customer value. This will contribute to the huge growth in data volumes that is expected in the coming years.

EBay, a Teradata customer, has started down this path.

“Traditional data warehouses talk about terabytes. To eBay that’s loose change. EBay talks about double and triple digits of petabytes.”

According to Brobst, few other companies are exploring this field yet, only the “lunatic fringe”, as he describes them. But he warns that simply capturing the data is not enough, a level of analytic sophistication is also required.

“If your analytics are not sophisticated enough, and you just collect lots of data, you’re incurring a large cost for no reason. You need to be sophisticated from an organisational point of view in how to exploit data to drive decisions. Otherwise the economics don’t work.”