NCR deviates into warehousing to be more than hole in wall gang

By Stephen Ballantyne

Friday 15th September 2000

'There's a growing amount of fraud involving stolen phones but our systems can detect that sort of thing while it's happening. All these applications require real-time analysis of data, which in turn represents a move from traditional data warehousing, which is analytical after the fact, to active data warehousing'

- Peter Hand

If you've always associated NCR with cash registers and automatic teller machines (ATMs) of course you're entirely correct to do so. NCR employees may not care much that the cash register side of the business has seen serious competition from Asian manufacturers: cash registers - how quaint, how 20th century.

ATMs are a bigger and more interesting business anyway but there's another problem - NCR has been so good at selling its ATMs it dominates the market in some places, like here. Worse, the market appears to be approaching saturation - although NCR makes machines in every size from walk-in rooms that offer full banking services by remote control down to little counter-size vaults suitable for corner dairies, in practice it's the middle-sized machines that sell best and we've got about enough of those.

NCR badly needs something else to sell, preferably something related to its core competency in transaction handling and large-scale data twiddling.

Ka-ching! Large amounts of data relating to sales? That'd be data warehousing, which is the primary concern of NCR's Teradata Solutions Group. Peter Hand, who runs the Teradata in our region, was in Auckland recently.

According to Mr Hand, NCR takes data warehousing seriously enough that it has split the company into two operating divisions: Teradata for the fancy new stuff and retail financials for all the traditional NCR business.

Data warehousing, apart from being a good fit for NCR, is an interesting technology that goes well with the spirit of the time. Huge computer processing power and plummeting data storage costs can be put to use to solve the problem of providing better service to customers, at a time service is perceived by many as a primary differentiator for customer-facing businesses.

In practice, data warehousing for retail comes down to analysing the masses of data a successful business can generate, in search of hidden regularities that can be clues to how people will spend their money and on what. Getting a fix on those regularities is like weather forecasting for farmers: done right it can make the difference between profit and loss.

Here in New Zealand we have a particularly successful example of the benefits of data warehousing in The Warehouse. For some time The Warehouse has been tracking and cross-collating data it gathers about its customers and what they buy and using that information to predict what they'll want to buy next.

The system has been running successfully for about three years, and yet its data gathering is unobtrusive and primarily concentrated on solving the traditional shopkeeper's conundrum of wanting plenty of stuff that people want to buy on the shelves and not much of what they don't.

According to Julian Beavis, NCR's local Teradata person, the effect over the past three years has been to nearly double total sales without anything like a corresponding increase in stock on hand. Visit The Warehouse and see for yourself - there's usually enough of what the customers want on hand but not much more than that.

"We're building on what they have already to give them an even better forecasting and replenishment planning system," Mr Beavis said. "It goes beyond the traditional method, which is to ask yourself how much of whatever did you sell last year, to take into account predictive modelling analytics considering factors such as seasonality, demographic and product change, and many more. The goal is to reduce the out-of-stocks as well as the over-stocks."

"You don't want to give the impression that you're looking over the customer's shoulder though," Mr Hand said. "If you do that, it can completely erode the benefits of the knowledge you can gain if customers don't feel they can trust the retailer.

"We're releasing more products that leverage further on customer information gathered, however. For example, we have customer relationship management systems for banks that can sift through transactions to pull out exceptional key events, such as customers placing large deposits. This could trigger contact from a personal banker, who can offer investment advice, thus providing value for the customer and the bank.

"When you think about the number of transactions a bank might have every day, this is no mean feat. And routing the information to a call centre or a personal banker the next business day is also a significant task that we can handle.

"Another application we've released handles fraud detection for telcos. There's a growing amount of fraud involving stolen phones but our systems can detect that sort of thing while it's happening. For example, if a phone that hasn't been used for many overseas calls suddenly places three calls to Sri Lanka, we can raise an alert on that. All these applications require real-time analysis of data, which in turn represents a move from traditional data warehousing, which is analytical after the fact, to active data warehousing, which extends the insights gained and makes them available to people in the field making minute-to-minute operational decisions."

It isn't essential to be a huge company to use Teradata solutions, either. According to Mr Hand, NCR has tried to make its recent developments more scalable and to have versions of them that run on NT-based systems.

"It comes back to business value. Our experience shows a six-to-one return on investment in nine months is very achievable. If we can't build a business case that will do that, then maybe data warehousing isn't relevant to you."

Mr Beavis admitted in the past data warehousing tended to be the province of large corporations, simply because they generated huge masses of data in which tendencies could more readily be detected.

Nevertheless, smaller users have emerged - one Australian Teradata user has only seven customers, and even though that doesn't sound like much data to warehouse, there is enough business volume to justify the system.

"There's more to it than data - there's also the complexity of the data model, which can drive the processing load as much as the sheer volume of data sitting on a disk somewhere." Some Teradata users operate with databases as small as 10Gb - a hard disk big enough to hold that much information sells for a couple of hundred dollars these days.

NCR hasn't quite got down to shrink-wrapped, off-the-shelf data warehouses yet, but Mr Hand claimed it has become pretty slick at setting up a system.

"Our methodology has been used in building over a thousand data warehouses and has been fine-tuned, with lots of reusable components. The biggest part is still the extraction of data from operational systems, because no one ever knows quite what that will be until we look at it.

"However, I don't think there are systems that we haven't extracted from - we can devise a system that will be of value to any business with a customer-facing, nearly anyone, in fact."

I'll say no more for now about NCR's data warehousing systems - later this month the company plans to fly me to its data warehousing partners conference in the tropical hell that is Disneyworld, where I expect to be fully indoctrinated with NCR's current thinking on the matter, assuming the 'gators don't get me. Stay tuned.

Comments from our readers

No comments yet