Do not use “offline eyes” in an online world!
This is especially true when it comes to knowing which data is most important in order to understand consumers. It is less about WHO they are, WHAT they did, and much more about HOW they behaved. The way consumers behave on an online platform will tell you more about their propensity to spend than other factors. Brands need to capture this data, and many do not.
Chinese consumer spending is to grow at an annual average rate of 7.7 percent per year in real terms over the next decade, becoming a key engine of global consumer demand and world growth, according to IHS forecast. Chinese consumers are moving more of their shoppings online in recent years, making China one of the most advanced e-commerce markets in the world. Since China’s emerging middle class is demanding high quality products and services, many online merchants are taking action to better understand and therefore better target this very battled group of consumers. Obviously, it comes down to data, but what data is actually more valuable?
Through a recent case-study, Artefact can bring some light to this question. When shopping online, (for services or products), brands and platforms will capture data on these consumers. By training a Machine-Learning based model to predict whether a customer will be a high spender, we combined and organized all types of data on consumers. It turns out that the type of data that has the strongest influence on their propensity to buy is not WHAT they did, not WHO they are, but HOW they behaved on the online platform. The question now becomes: how much data do you capture on HOW your consumers behave when buying your product?
Case study: an online travel service company leveraging a machine learning model
An overseas travel service platform connecting local drivers and guides to Chinese travelers provides all kinds of travel services such as picking up at an airport, renting a car, ordering a local driver guide etc. The problem they faced was that not many consumers were willing to pay for their premium service. They, however, really want to establish a reputation with high-end customers.This company hoped to use their accumulated data to recognize who their potential premium customers are and then try to retarget them more proactively.
In order to predict the possibility of a customer paying for the premium service, we first extracted numerous features. These features can be organized in 3 clusters:
- WHO are the customers? This includes a customer’s age, gender, province, city etc.
- WHAT have they done? This includes information on what they’ve purchased, whether they cancelled an order, the number of total orders etc.
- HOW did the consumers behave? This cluster includes information on the total time spent on the platform, the speed at which they clicked through the app, the number of different clicks that were made, whether a comment was left or not etc.
Most features available were on WHAT customers did (which is the starting point of any data capture exercise). But when looking at the importance of the variables, HOW the consumers behave had the strongest influence on them eventually buying premium services: 62% of the predictability was driven by these features, whereas they only represented 35%.
In addition, with consumer reviews from previous data, we had used some tools to do a sentimental analysis to extract review features. Based on these reviews we can quickly pinpoint the place that needs to be improved. It can also be used for personalized targeting, especially for those who have had a bad experience but still have the potential to be customers.
To better understand how customers make their purchase decisions, we have observed in many instances that the most important factor is the variance in customer behaviors, rather than the actions and demographics. With rich behavior data, derived features and advanced analytics, merchants can easily identify and target the customers who are likely to become premium subscribers. The good news is that in e-commerce, brands can see much more data on consumer behavior compared to offline retail. But you better anticipate it.
In summary, do not use “offline eyes” in an online world. This is also true when it comes to data.
Would you like to know more about e-commerce in China or using data in general to better understand and identify your customers? Reach out to our experts via the button below!