The truest and most memorable insight into IT industry dynamics I ever heard was delivered by Steve Mills, senior VP and group executive of software and systems at IBM. During a 2010 interview we were discussing the relative importance of various layers of the stack, and I asked him, "what's really driving customer technology selections: hardware, middleware or applications?"

"It's money," Mills replied. "That's the No. 1 motivator. And money is not a single-dimensional factor because there's short-term money, long-term money and money described in broader value terms versus the cost of a product. The surrounding costs are far in excess of products."

That line hit me like a ton of bricks, so now no matter how hyped or religious technology debates become, I try to remember that money will determine the outcome. Customer loyalty, inertia, vendor market share, cozy relationships and other factors may intervene, but as in natural selection and economics, the idea is that the best, most cost-effective approach will ultimately win the day.

The context of that 2010 question was whether companies would choose Oracle's engineered systems versus IBM's offerings. After sharing his money line, Mills said IBM is seeing customers switch from Oracle Database to IBM DB2 because the former is expensive by comparison.

[ Want more on how big data might impact incumbent technologies? Read Big Data Debate: End Near For ETL? ]

Money cuts both ways where vendors are concerned. On Monday I had another interview with Mills, this time to talk about IBM's latest big data initiatives. I asked him whether he thinks Hadoop and NoSQL databases might change how or even whether people continue to use relational databases and extract, transform, load (ETL) technologies -- a topic we've explored in recent InformationWeek debates.

"Back in the 1990s the assertion was that object-oriented databases would replace relational databases," Mills responded. "Periodically we get all this excitement about technologies, but it's not a one-size-fits-all world. There are point-product companies that offer a thing, but those things don't necessarily solve the complete problem."

At that point I had already completed our collection of "13 Big Data Vendors To Watch In 2013," but the conversation with Mills confirmed in my mind the decision not to include any incumbent vendors from the relational database world.

It's not that I disagree with Mills. In fact, I'd guess that some of the vendors that did make my list won't be around five years from now. All of these vendors are focused pretty much on one thing, and if it doesn't work out, well, then there's always the next startup. On the other hand, their pure, undiluted and uncompromised attention to one product tends to focus the vendor on building a better, more cost-effective product. Hadoop and NoSQL platforms are proving to be more scalable, flexible and affordable in big-data deployments, and that's why they are the focus of the "13 To Watch" collection.

IBM does, in fact, have a Hadoop option. I recall Rod Smith, VP of emerging Internet technologies, presenting on IBM's research on Hadoop way back at Hadoop World NYC in 2009. But it then took two years for IBM to release its InfoSphere BigInsights Hadoop software distribution.

Mills insists that "thousands of customers" are using BigInsights, but there are few signs of a big commercial push. I haven't heard many customer names. It feels like window dressing. Similarly, Mills said Monday that NoSQL structures can be managed by IBM DB2, but I never hear about it being used as a NoSQL database.

DB2 was the incumbent database at Constant Contact, but the digital marketing services company chose the open-source Cassandra NoSQL database when it decided to provide social marketing services as well as e-mail campaigns. By Constant Contact's estimates the relational route would have required a $2.5 million investment and nine months of development. It deployed Cassandra, supported by DataStax, within three months at a cost of $250,000. In short, money won.

IBM isn't the only relational incumbent that seems to be taking an arm's length approach to new big data platforms. Teradata's thinking about Hadoop has clearly evolved. I recall conversations with CTO Stephen Brobst in 2010 in which he downplayed the importance of Hadoop, slammed its performance and served up the well-worn "open source isn't free" argument. One year later Teradata acquired AsterData so it could support MapReduce processing within a relational database environment. Then in October, Teradata announced it will support Hadoop nodes alongside relational nodes on a single appliance, mirroring a strategy EMC Greenplum adopted at least a year earlier.

Oracle and Microsoft are the latest big names to offer Hadoop, in this case through partnerships with Cloudera and Hortonworks, respectively. I'm guessing those moves helped persuade database newbie SAP to jump on the big data bandwagon by adding a Hadoop connector.

At SAS I've heard the open-source-ain't-free argument from CEO Jim Goodnight more than once. Nonetheless, SAS added support for open-source R-based statistical models a couple of years ago at the behest of customers who didn't want to have to spend time and money developing stuff that was already available from the R community.

I could cite other examples, but my point is that incumbents all too often suffer from not-invented-here syndrome. If they embrace potentially disruptive technologies at all, it tends to be a late, shallow or half-hearted embrace. That's because what might get disrupted is the flow of money from incumbent technologies.

There's a lot going on in the world of big data, and I agree with Mills' point that companies using new platforms will still have to address data integration, data quality, data governance, backup, recovery and so on. They'll also still need conventional relational databases and business intelligence and analytic systems for all the proven applications and new applications to come that will run on those incumbent platforms. But if the question is which vendors will really champion and innovate big data platforms such as Hadoop, NoSQL and related analytics systems, you have to look to the companies that have all their money riding on these new environments.