Arrived in Denver, and Thoughts on "Super Crunchers" and Oracle RTD

I finally arrived around 7pm last night, the plan then was to try and meet up with various people but as I'd been traveling for getting on 18 hours, I thought it best to get some sleep and I crashed around 9pm. As usual on the first day over the in USA I was up early - 4.30am this time - but I actually kind of quite like that, I'm really a "morning person" and it gives me a chance to get some work emails answered and get myself prepared for today. Denver and the conference venue, as is always the case with Collaborate events, looks very impressive - I've been to the States a few times now and I take it all more in my stride, things I always like about coming to America are the drive into the city (coming down a motorway and seeing lots of skyscrapers, that's not what you get when traveling in to London along the A40), the super-smart and very new hotels and buildings, how polite everyone is, the fact that the hotel coffee shop is open at 5am. Things I still can't get used to is how tricky it is to get a good cup of tea (I've got used to asking for "hot tea", or "tazo tea" in Starbucks), how people go nuts if you inadvertently jump the queue or cross the road when the "don't walk" sign is illuminated, and having fairy cakes for breakfast. Other than that, it's great to be over here, especially as I almost didn't make it this year what with the flight cancellation.

My talk is actually today and it's on using Oracle Real-Time Decisions to make generalized, small business decisions. The thinking here is that whilst we've got the big decisions well covered off with tools such as OLAP, reporting and data warehousing, smaller decisions such as whether to grant a refund, whether to offer a discount or what supplier to use for an order are often just left to intuition, a few business rules or just whatever the person making the decision felt like doing on that day. Added up though, these small decisions can have a big impact on the success of your business, and tools such as Oracle Real Time Decisions, originally used for real-time offer generation and based on data mining technology, can be used in this situation to add "intelligence" to your small decision-making process.

This all fits in quite well with another book I'm reading at the moment, "Super Crunchers" by Ian Ayres, which looks at how two data mining/statistical techniques, "Regression" and "Randomized Tests", are behind the success of organizations like Capital One, the credit card supplier that operates in the UK and the States. Regression is a statistical procedure that takes raw historical data and estimates how various causal factors influence a single variable of interest, and is one of the main techniques Real-Time Decisions uses when working out what customer attributes have the most influence over whether an offer is accepted, or as in my Collaborate conference paper, what factors of the customer order have the most influence over whether a supplier will ship on time. Randomization is something that Capital One are famous for and involves creating randomized groups of customers and then offering them, say, a different interest rate or a picture of a smiling woman on the offer envelope to see which offer is taken up most. The randomization but is added because if you normally took two group of customers and offered them different deals, you'd need to ensure that each group had equal ages, incomes and so on to make it a fair comparison, but of course this is tricky to set up and so if you get a large enough group and ensure members are selected at random, you effectively negate differences in income, status and so on (by averaging them up) and ensure that the only factor that could truly influence, in the aggregate, whether an offer was accepted was the terms of the offer itself. This is something Real Time Decisions also does when it creates statistically similar groups of customers and applies randomized offers to them.

I've got a few hours to go before my talk (it's 6am now, the talk is at 2.45 I think) so I'm going to try and incorporate a few anecdotes from Super Crunchers in to my talk. Other than that, I'm meeting Peter Scott around 7am for breakfast (assuming he's jetlagged as well) and then I've got various talks I'm going to today including one by Edward Roske on "How Essbase Thinks", one by Mike Donahoe on the latest with BI Publisher (looking through the slides in advance, the 11g release of BI Publisher seems to come with an MDX query builder as well, this'll be interesting as this is something even MS Reporting Services doesn't come with), one by Vincent Chaznoor on Oracle 11g data warehousing and a talk by Dan Vlamis on comparing Essbase and Oracle OLAP.