Dr Jim Goodnight talks the business of analytics

CIO: What was the impetus for starting SAS?

Goodnight: SAS itself started at NC [North Carolina] State University. Tony Barr and I began work back in 1966, 1967. At the time I was putting myself through my masters in statistics, but I had been programming for about four years and the Department of Statistics was paying me to program software to help analyse all the data on campus.

NC State is what’s called a land-grant university; one university in every state receives money from the federal government to establish a college of agriculture and engineering. The mission of these universities is to improve crops, improve livestock, and that’s where almost all the major breakthroughs in crop science come from.

Every student they graduate, every PhD, has to come up with something new. The group that I worked for was in charge of helping design the experiments and I helped analyse all of the data.

So we released our very first version of SAS in 1968. After that it was just continuous improvement, continuous refinement up until 1976 when we had grown and there was no more room for us in the building. We felt we had enough users and enough customers that we could actually live off campus. As part of the deal with the university, we had to be across the street. So we just moved straight across the street and found an office over there.

You have seen many changes throughout your tenure at SAS. How have you maintained the level of innovation?

We generally try to plan out about two years. That’s the window of time that we can get something fully developed, fully tested and ready to work with customers. It allows us to be more proactive for things like the iPad. Finally, we have a mobile device that’s big enough to do something useful. I always felt like most of our bar charts and graphs and all of our analytical stuff just doesn’t look very good on a phone; there’s not enough real estate. But all of a sudden, we were excited about making sure we had some apps out there that allow a user to download data from a server and display it on their own iPad.

Organisations sometime struggle with business analytics. What do they do well and what can they improve on?

To really effectively use these packages, you need somebody in your organisation that has thorough knowledge of what they should be doing with them. And right now there’s a little bit of a shortage in the world of analytics experts.

That’s one of the reasons we helped fund and develop a masters program at NC State University, as a Masters of Advanced Analytics. It’s a 12-month course that covers some time series, some modeling, some optimisation, and all the combinations of things that we’re using these days.

It’s not one particular analysis — it’s really combinations of many of the analytical tools. You really don’t need to be a PhD in statistics to use these tools properly; you just need some instruction on how to use them and what you’re using them for, and we’re hoping to see this kind of program spread to other universities.

How do you bring information from social networks, which is largely unstructured, back into a business?

One of our first customers was the Marriott Hotel. They were interested in what people thought of one of their brands. We go to every blog that has mentioned that particular brand and we scan each week for that brand. And we pulled all this data together and sent them an analysis on the text. You build a taxonomy of all the different adjectives that people are using to describe a particular brand and you map those to a numerical value so that ‘zero’ is that group of words which mean really bad, you have all the words that mean really great, and then all the ones in between. You have to build these taxonomies more by industry than anything else because industries tend to have much of the same taxonomies.

Then we scrape the blogs and tweets, analyse them and come up with a rating of what people think of your brand. Now, this could be very useful before and after a marketing campaign to see whether it has had any affect on people. We’re having a hard time meeting demand on that.

I suspect you have been asked this question many times over the years, but how have you maintained such strong revenue growth where other companies have faltered?

Well, one thing: We don’t sell our software outright. We license it on a yearly basis. That really keeps us on our toes because we have to make sure we’re satisfying all of our customers and we’re expanding in ways they want us to. I think that has probably been a real key in our growth and profitability. It has been very good because cutbacks are worse right now, and our licensing is so much easier for most companies.

Nowadays, we don’t ever develop new software unless we are working with a customer that has a need for it.

That’s exactly how the whole markdown optimisation and retail pricing [analytics] started. We were working with about three different retailers who were trying to lay out a roadmap of what they wanted to do, where they needed us to move to and the ideas came from our customers.

What industries do you think best understand their customers?

I think banks do, really. They’ve been in the modelling business for a long time — 25 years or so — building predictive models.

And what industries don’t do it so well?

Well, let me think about that. Take the point of sale in the grocery store, for example. Most stores are not using that data very effectively to really discern trends and shifts in buying habits and that sort of thing. A few stores are doing so but the vast majority of point of sale data just goes to do inventory.

All of that data could be used to calculate the run rate of particular items in the store. [Things like] when will it sell out, and if it’s past the end of the season, then we’d better intervene and try to reduce price on it to move it out of the door. So we are working for a lot of major retailers. And we’re using our high performance computing to do that.

And pharmaceuticals — they’re very good with analytics in the lab, analysing new drug data. But they are quite weak when it comes to using analytics to manage their companies. We are so well known as the tool to use in the laboratory that we have a hard time convincing them to use us for management purposes.

Do you think it’s a challenge for chief information officers to create a meaningful ‘story’ around the analytics, in terms of communication?

I think the biggest challenge CIOs have is the fact that some of these silos of data have built up. Take a bank for example: A credit card database keeps track of all the credit card transactions, another one does demand deposits, another for home loans, another for car loans. And none of these talk to each other.

One of the things we’re trying to do is to have a central messaging hub where every time a transaction occurs, the customer’s state record manager updates the information.

And I can give you a very good example of why this needs to be done more — and CIOs really needed to think about this: The fact is that the data that you need for fraud detection is almost identical to the data you need for marketing; it’s looking at the customer, the types of transactions, the preconceived transactions, the kind of credit card customer the person is. All this data is practically the same, yet they’re having to extract it from different databases.

And you have the compliance people that have to worry about anti-money laundering. They have to worry about providing risk information. It’s almost all the same data, but it’s not their job to worry about marketing.

Somehow, the CIO needs to reach across these silo databases and understand how they’re being used so they can see if they can achieve better efficiencies. It is, in some ways, the greatest opportunity and the greatest challenge.

For a CIO, I think it is. And then I think finding ways to extract a silo of information is an absolute — the kind of job every CIO should be doing.

There’s a lot of pressure on CIOs in most organisations just to keep all the hardware and software running, and putting out fires that occur when servers break down and people are out of service. It’s a job that can consume their entire time. But you know, things like trying to break the silos down, that’s the secondary level.

Georgina Swan is editor of CIO Australia. Follow CIO Australia on @CIO_Australia on Twitter.