Last Christmas, I Gave You My [Credit] Card…

Posted on: 17 December 2015

Online shopping has been on the rise for many years and this Christmas will most likely mark another new record for web sales. The reason is clear: it has become more convenient to shop online — why battle with hordes of stressed-out shoppers and bad weather when you could be shopping from the warmth and comfort of your couch?

But there is another reason for the surge in online shopping, according to Research Fellow in Computer Science and Statistics at Trinity College, Dr Kevin Koidl, who is also based in the ADAPT Centre for Digital Content Technology. Dr Koidl, whose key research areas lie in web personalisation and artificial intelligence, says that online retailers have become very smart in knowing when you want something, and how much you are considering paying for it.

Most of us know by now how this works. It arises via online tracking, customer segmentation and scoring. Together, the techniques are known as machine learning and artificial intelligence. Their sole purpose is to gather and assess as much data about you as possible, so as to trigger certain actions that will ensure you successfully ‘convert’ and commit to clicking the ‘buy’ button.

It’s easy to argue that we are going to be tracked anyway, so we might as well get on with it – what’s the worst that could happen?

Well, the challenge lies not in the fact that the data is collected, but in the fact that we are not able to understand how it is being processed. This is especially critical in online retail, where this lack of knowledge can mean you might be losing money by getting a bad deal that has a lofty price tag designed only for you.

This ‘dynamic pricing’ has also been a part of the offline world for a long time. Online, however, things get a little bit more complicated.

As an offline example, you can imagine you are shopping in a local grocery store. The owner goes by the name of Mary. You buy most of your groceries there and do so on a regular basis. Mary values your business and is always friendly to you. One day you are in the shop to pick up some butter and you are waiting behind someone else, who is also buying butter. Mary asks for €2.50. The person pays and walks off. You have the usual chat after which you ask how much you owe for the butter, and Mary replies €2. You are pleasantly surprised and ask why she charged 2.50 for the customer before. Mary simply replies: “He looked like he could afford it.”

This interaction may be morally questionable, but from a pure economics point of view, it’s a perfectly valid transaction. The customer could have haggled or chosen not to buy. Furthermore, there is no monopoly on butter, which means the customer could have bought the butter somewhere else. It’s business as usual. Price is determined by good old-fashioned supply and demand. So why is this different online?

A smart seller (or a smart bargain hunter) usually has one very good skill: reading the person on the other side of the trade, to assess what price can be asked for (or offered). It’s what Mary successfully did in the example above and it forms the base of every good or bad deal. In the online world, the other side is not a person but an algorithm. And, because you can’t read an algorithm, this is the precise core of the problem.

By EU law you have the right to request all data from an EU-registered entity that has stored personal data about you. You also have the right to have it deleted. This won’t help you, however, because each algorithm is proprietary and you won’t be able to request how it works. Also, simply deleting your data can be a bad idea, as you might prefer to do business with someone you know a little rather than someone you know nothing about.

To explore the problem in real-time, try a little experiment: Keep requesting the price for an airline ticket on the same website over a 30-minute spell. To add some more data points, state that you are travelling with 6 people and pick a typical Stag party location. After doing this make the same request on a friend’s phone. You will see a differently laid out website and, in some cases, a different price.

But you will have very little chance of finding out why this is. You can’t ask the seller as in the example with Mary. You can only guess. Was it the location that indicated a social trip, or was it the increased clicking that indicated an increased interest to make the purchase? No one knows except the company that designed the algorithm. And even if you manage to reverse engineer the decision matrix, it may change the day after.

So, how are you supposed to get a fair deal if the other side knows a lot about you and you know nothing about them? After all, you are not collecting data about the behaviour of the algorithm even though your behaviour is heavily tracked, which seems very unfair. Well, the good news is there is something you can do. It’s by embracing the most important thing you have left as a consumer: choice.

This effectively means you need to buy products from different online stores and, when possible, offline as well. Pay with cash, every now and then, and don’t always use your loyalty card. You can also advocate a more extended version of this by using different browsers and different email addresses. Also, leaving products in the shopping cart without buying them, or doing silly things just to see how the algorithm reacts can trigger some interesting interactions. The key is to avoid leaving an easy to predict pattern by spreading your data wide and thin.

Keep in mind your opponent in this trade is a probabilistic model that simply keeps calculating the likelihood of a successful price change. The less predictable your behaviour is, the harder it is to predict if a change in price will work.

So, when it comes to your Christmas shopping, don’t just lie back and click. Always remember there is something or someone on the other side that might give you a better deal if you try a bit harder. You almost certainly try that little bit harder offline, so why not do the same online? A little bit of effort could go a long way and help those precious Christmas dollars stretch a little further.