When you look under the hood of your car, it can be forgivable if you haven’t got a clue what most of the things inside actually do. The same cannot be said about your sales and marketing efforts.
Knowing exactly what your customers are going to buy, not just now but also in the future, is a dream most businesses aspire to, and it is already a reality for many. The likes of Souq, Netflix, eBay and BasharaCare are all giving personalized recommendations to consumers based on their browsing and purchase behavior. Being able to offer your customers a personalized service, however, is not reserved for only large companies. Businesses of all shapes and sizes can offer personalized recommendations to their customers. The key to unlocking these recommendations for your customers is data.
Your customers create data every time they interact with your business. Everything from their online browsing behavior, to their social media and previous purchase data can be analyzed to work out what they are likely to buy next. Once you know this information, suggesting the products your customers are likely to want to buy becomes common business sense. Before you get those recommendations for your customers however, the data you collect must first be run through a series of algorithms that are collectively known as a personalization engine. Personalization engines are nifty piece of data science that, at their most detailed, allow you to make the kind of recommendations customers usually receive from friends and family.
Personalization engines are the data equivalent of a teacher recommending a book to a student. The important point to note here is that friends, family, teachers and personalization engines cannot make any suggestions without context. They all need information in order to make informed decisions about what an individual may or may not like. Personalization engines are particularly powerful, as they not only allow businesses to recommend items to customers that they are most likely to purchase. With the right data, a personalization engine may suggest products and services to a customer that they may otherwise have not considered, but that are suitable for them based on their interests, needs and desires. A good example of this is how Netflix suggests TV series and movies based off of previous viewing behavior. Going a little left field with some recommendations could pay dividends, diversifying the type of products and services an individual looks to purchase from a business. Likewise, offering a customer something a little different to what they are accustomed to, which they then enjoy, will help strengthen your relationship with that customer and the sense of loyalty they feel towards your business.
Before you look to bring a personalization engine into your business, it is worth having some background knowledge of how the technology works. There are actually three different types of personalization engine available, each working in slightly different ways and suitable for different business needs. The first of these is known as collaborative filtering. A large amount of data is collected on customers’ interactions with a business, including their previous purchases, when they have purchased, where they have purchased (online or offline) and any other engagements they may have had with the company. All of this data is then used to predict what an individual customer may be likely to buy, based on their similarity to the business’ other consumers.
The second method is called content-based filtering and it works through keywords that are used to describe individual products or services. Profiles are then built to indicate the type of product or service a specific customer may like, according to the keywords used. Individual customers are therefore recommended items that bear similarities with previous purchases or ones they have browsed online and are possibly considering. Recommendations through this method can also be made based on ratings and reviews a customer has given on other items.
The third type of engine is actually a hybrid, mixing the other two methods and in some cases, this approach can be the most effective. Netflix’s recommendations are a good example of this approach in action. Another point to note is on the potential issues these engines could encounter. The most significant of these is known as the cold-start problem. Much like how a car engine needs fuel, the personalization engine needs data, and if it doesn’t receive enough, it cannot accurately recommend items. Therefore, there is a minimum amount of data that the engine needs to get going, and for a new business with little to no purchase history, this could pose a problem.
The cold-start problem can partly be resolved through content-based filtering, if items are classified well. Potential customers will then have to list their preferences from the beginning so the engine can match those preferences with the items’ classifications. A constant feedback loop, where customers then tell the system whether or not a recommendation was useful to them, would then train the engine to get better with each personalization. A further issue could arise if a personalization engine is set up incorrectly. If the engine is fueled by data that isn’t suitable for the method used or needs of the business and its customers, then the recommendations given may simply state the obvious. A customer who buys a puppy is obviously going to need puppy food, so that recommendation will not be useful to them. Suggesting that they also get a radio because other puppy purchasers have discovered it helps keeps the dog calm at night is the kind of recommendation that will surprise and delight a customer.
Personalization engines are still a growing and niche tool for many businesses, and until now, have largely been reserved for larger businesses. With the right method and the right data however, businesses of any size can offer the kind of recommendations to their customers that will keep them coming back to you time and time again. More and more businesses will soon realize the potential in predicting what consumers want. When this happens, personalization engines will become as commonplace as their mechanical counterparts. Just as there is a car on every street, there will soon be a personalization engine powering every store.