Data mining is a form of business intelligence and data analysis. It is the process of analysing data to draw useful conclusions or predictions from it. It’s a technique frequently adopted by large-scale ecommerce businesses to aid with marketing and product development.
Because of the nature of the internet, ecommerce businesses will obtain a lot of data about their customers, or their prospective customers. Data is obtained whenever a purchase is made, an account is created, or a page view is made. This raw data can come from the database associated with an ecommerce website which holds all account and payment information, as well as from web analytics tools like Google Analytics.
If Google Analytics is correctly set up, whenever a visit to a site is made, information about that visitor is collected. This information could be about the visitor’s journey through the site, how they got there, what they searched to find the site, where they went after the site, how long they spent on each page etc.
Data could also be about the visitors themselves. If the visitor is logged into a Google account then the Analytics cookie will be able to read information about the account. This allows the company to start using data mining techniques to build visitor profiles to analyse the demographic of visitors to their site.
For example, the Google Analytics of a clothing website might show that 82 out of 100 visitors are female, 92 out of 100 live in the UK, and 70 out of 100 are aged 18-34, and 10% of them are shown as ‘interested in sport’ (defined by Google based on previous search history).
Using data mining techniques it’s possible to start building profiles for the type of person that visits the site. This allows the business to market and create products based on those personas. This combination of acquiring data and analysis of it through data mining techniques tells the business owners that, based on their audience, it would be beneficial for them to begin to market and sell female sports clothing.
Data mining is how Amazon have perfected upselling by promoting other items with banners like ‘customers who bought these items also bought’, ‘recommended for you’ and ‘frequently bought together’ bundle packages.
Amazon acquires a huge amount of data through various methods that they then use to customise and personalise offers and promotions to the customer. This cross-selling has been proven to increase the average order size of a basket.
But, what happens if you don’t get as far as the basket?
Data is taken at every stage of the user’s journey through the site. Data is stored whether the user only browsed a product, or added it to their basket but went away without purchasing, or followed the purchase path but cancelled the order when they saw the price of delivery. This data can be processed to tell the company that perhaps they should consider offering free delivery, or maybe they should send an email to that customer in a few days with a special offer to prompt them to return and finish their purchase.
Sales forecasting with data mining is where a company can use data to predict things useful for stock control, pricing and marketing. If you do your weekly shop online then the supermarket learns about your buying habits, and also your consumption habits. If you buy a bag of 100 teabags on the first week of January, and then again on the first week of February then the supermarket learns that it takes you about a month to use 100 teabags.
So from there, towards the end of February, they can target you with offers on teabags because they know that you’ve nearly run out. This makes you think ‘oh yeah, I do need teabags’ so you log in to your online shopping to get some teabags, and whilst you’re there you get some milk, and actually, you need some bread as well, and ‘ooooh there’s a sale on dishwasher tablets’ and that’s how they get you. This also lets them predict how much stock they need to have in store at various points of the year, and data mining powers that.
Data mining techniques can also be used as a method of fraud prevention. Data mining techniques detect abnormal patterns in a data sequence. For example, a bank might automatically temporarily suspend a credit card if their fraud prevention systems notice that it was used in McDonald’s in London and Burger King in New York within the space of an hour.
Or, a card may be suspended if it’s only usually used to pay for a cup of coffee here and there, but suddenly it’s been used to buy a 54-inch television and a quad bike. That’s when data mining techniques notice an abnormality, and so the bank gives the account holder a ring to make sure everything’s alright.
Data mining can be incredibly beneficial to a company’s growth. Basic data mining techniques can be created with the use of Google Analytics and Microsoft Excel, but for a comprehensive and effective data mining solution, it’s best to consider a bespoke solution.
The data mining wizard on Microsoft SQL Server 2016 makes it easy to model, analyse and predict trends based on a range of data from different data sources. Microsoft SQL Server 2016 is now available on CloudNX from Fasthosts.