Data mining can provide fresh insights into your business, but how do you ensure you’re identifying the right information and sharing it with the right people? The experience of games retailer EB Games offers some insights.
Kevin Clarke, manager for software development at EB Games Australia, gave a presentation at TechEd Australia 2012 this week on the company’s data mining strategy, which is based around its EB World loyalty program. That has 1.1 million members in Australia (and 20 million in the US), so there’s lot of information to be analysed. Here are some of the tips Clarke shared.
Make sure you have clearly identified goals “The more you can give back to the business in terms of data mining the better,” Clarke said. “The one thing I’ve found out is you don’t want to go to the business with data insights that don’t provide any actual [usable] insights because it doesn’t make you very popular. So if you can be vetting what you want to do with the business users, getting a proof of concept validated and then bringing it back to your developers to create enterprise-scale visualisation tools, that will work better.”
Those goals won’t necessarily be complicated or dramatic. “Market basket analysis is the bread and butter of a loyalty system. You can do it without it but you don’t get much value because you lack the customer-to-transactional mapping. Understanding past purchasing patterns enables direct marketing. It’s hardly nuclear physics or anything but it’s pretty widely used.”
Engage end users through tools they are familiar with. Data mining systems are often built within IT, but the users who want that information will often perform similar analyses themselves. “It’s really important in terms of trying to get data mining into the business to enable the power Excel users who would be using a pivot table in Excel and get them attached to a data source that you have better control over,” Clarke said. “From my personal experience at EB Games there are a lot of people who do their own thing in Excel and you don’t know where they’ve got the data from. you don’t know what they’re doing with it and you don’t know where they’re saving it.”
Data mining is especially useful in mature businesses. In a rapidly growing company, past performance won’t offer much insight into future behaviour, but once a market stabilises, data mining is particularly useful. “For a while now we’ve stabilised in terms of our market saturation in Australia, so year on year sales analysis will allow us to predict more realistic comps for our stores and that drives the KPIs for each store,” Clarke said. “That’s a really important part that data mining can play.”
Data mining offers guidance, not certainty. “Forecasting is something that’s really hard to do well,” Clarke emphasised. Good analysis can identify patterns, but correlation is not causation.
Plan for how you’ll expand your analysis. EB Games runs extensive data mining on past sales, but hasn’t yet started on real-time clickstream analysis for online sales. However, Clarke is considering the possibilities.
“Clickstream analysis is not something we’re using at the moment,” he said. “A place we feel we could use this is our ecommerce environment, which is hosted out of Macquarie Telecom in Sydney, and our main data centre in Brisbane at Eagle Farm. We have a private WAN link between the two centres. Real time sales clickstream analysis would allow us to do that without running reporting queries across the production transactional database which obviously will improve performance. That’s something we might look into implementing.”
Visit Gizmodo’s TechEd 2012 Newsroom for all the news from the show.
Disclosure: Angus Kidman is attending TechEd 2012 as a guest of Microsoft.