I sometimes visit e-commerce conferences to see and hear experiencees from the business in becoming more data driven, and to get inspired about how we can improve our solutions. Lately, i have visited two different events and i like to share my experiences with you.
One of my conslusions is that most e-commerce organizations are currently focusing on A/B testing when they talk about implementing data driven business solutions. Some classical examples are:
By means of a statistical test the e-commerce manager tries to find out what works best to maximize conversion rates
A. The webpage with the current lay-out
B. Change one variable in the webpage lay-out (e.g. other color button, change content slightly)
A small fraction of the visitors gets to see B instead of A during their visit. If B generates a higher conversion, the webpage lay-out is changed to B. I have seen impressive results where conversion rate grows up to 40% (not percentage point) after changing just one variable. However, this method goes rather slow because it relies heavily on qualitative input from marketing managers and economical psychologists. They understand how customers think and thoroughly analyze the customer journey to improve it step by step.
Some time ago, I went to a presentation titled “The Perfect Data Driven Customer Journey”. I must confess; I am a believer of combining customer journey analysis with data science. Nowadays, most retailers have no personal relationship with their customers anymore because of digitalization. Using consumer website data is a great opportunity to better understand and serve your customers according to their individual needs, without knowing them personally.
So, you could say that I had high expectations of a presentation with such an appealing but ambitious title. The presentation was interactive and quite amusing. The presenter was handing out candy bars to the audience to reward them for giving correct answers. But unfortunately, the presentation did not give me many new insights on a perfect data driven customer journey. The personas in this use case were young singles, families and seniors. After registration on the website, visitors were classified into one of these personas based on date of birth. Subsequently, marketing managers created special content for these three target groups based on their typical interests and needs (e.g. a senior couple was presented a quiet holiday, and families to hotels with a playground for their kids). However, this was done qualitatively, by hand, without using any algorithm.
Of course, qualitatively differentiating customer journeys is a good first step to make content more personally relevant and convenient for (potential) customers. However, it is not near to perfect and relies heavily on gut feeling and generic segmentation. I found myself wondering why these modern organizations were not using a modern data science approach. Aren’t they aware of today’s possibilities? Is it due to the shortage of data scientists?
I see many organizations struggle with handling unknown visitors who have not been registered yet. Actually, you don’t need registration data in order to understand your visitor’s commercial needs. In fact, there are multiple ways to track e-commerce behavior on your website. And no, I am not talking about Google Analytics (which is inaccurate and aggregated data, hence not suitable for individual level purposes).
* I will not go into detail about specific methods and software about web tracking in this blog but my data engineering colleagues might pick up the pen later if people are interested
The first time you enter a website you truly are anonymous. However, not for long. After every click the website is able to understand your needs better. For example, clicking on a school agenda of your son’s favorite football club tells a lot about your customer profile. This approach focuses on your taste and commercial needs rather than on your name and address. For example, recommendations can be given based on what others with the same behavioral patterns bought (instead of categorical recommendations), website context can be slightly be changed if your focus is more on reviews than on price, etc.
So it is possible for websites to extract data during your first website visit. Although this might sound a bit spooky to most people, I do not agree on this one. When I visit, for example, the H&M website I certainly do not want to browse through all 10.000+ available items. I want to spend my precious time effectively and see only those items that I am likely to buy (I admit I am not a great fan of shopping). I do not mind if H&M uses data science to help me out here and personalizes their website according to my predicted needs. After all, we both have the same goal: match customers (me) with their products. I can perfectly handle a bit of collateral damage in the form of conversion optimization.
Moreover, in many cases an algorithm does not care about my personal demographics. My behavioral pattern is far more interesting:
I hear you thinking, certainly most organizations do want to have my name and e-mail address. It gives them the opportunity to send you personalized offers. This is true, you need someone’s contact info to reach out to him or her. But if the content of such offers is not perceived as valuable and relevant to you it is likely that your customer will be annoyed and perceive it as spam. Subsequently, your love for this brand or company decreases significantly and the effort of the personalized offer backfires. I would like to clarify that accurate recommendation algorithms are not only accessible for the Amazons and Spotifys of this world. Data and the knowledge to work it is all you need!
We are moving away from traditional transactional consumerism towards a more relationship-based business model. Most organizations want to build a strong brand and therefore try to serve you the best way possible when you visit their website. The increasing availability of consumer data is a great opportunity for organizations to better understand her consumers. But the usage of such data must feel safe and add value for the customer.
We have seen some impressive results with A/B testing. I wanted to show that data science can be a great addition to the e-commerce manager toolbox, and that it already adds value in the early stages of the customer journey without knowing any demographics yet. I see that both customers and e-commerce companies are ready for the possibilities that lie beyond A/B testing. Possibilities to increase not only direct sales, but more importantly customer satisfaction and brand loyalty.
I am very curious how the e-commerce standard will change in 2017. I think data science will play a great part in it. Next month, I will be attending another e-commerce conference. There I hope to see more data science and less candy bars. And remember, you don’t need to be Netflix or Google to profit from modern technological advances.