One size doesn't fit all for online sales, research shows
New analytics research is helping online retailers optimise their product listings based on customer segmentation.
While online shopping has been gaining popularity for years, COVID-19 has pushed more consumers and businesses online for all types of transactions, creating a demand for convenience, personalisation and an abundance of options.
To meet this influx of demand, e-commerce retailers have had to shift their approach, with many choosing to adopt the marketplace model – in which third-party vendors offer their products and services through a central platform.
While a few years ago the only marketplaces that came readily to mind were Amazon and eBay, consumers are now spoilt for choice. For merchants, this means increased competition and, consequently, new strategies for selling their products and services.
The importance of listings
Many of those strategies focus on product listings – the art of determining which products to display first, and in which order.
Take, for instance, shopping on Amazon: 70 per cent of customers only purchase items they see on the first page of their product search, which means that the way Amazon decides which products to show on the first page can have a huge impact on someone's business.
Where and when to place products and services has been the subject of countless studies – however, until recently, most have been based on the assumption that consumers are a homogenous group with similar purchasing behaviours.
This theory of consumer homogeneity has led to the assumption that the most popular products, determined by how often they have been purchased in the past, should be given more visibility by being placed first on a page.
The problem with this approach is that consumers are not actually a single group but a diverse population, led by segment-specific purchasing behaviours within different markets.
"In previous research, when consumers were assumed to be homogenous, the problem of how to best place products was simple to solve," says Gerardo Berbeglia, Associate Professor of Operations at Melbourne Business School.
"But when consumers are considered heterogeneous, what we found was that even calculating the optimal ranking order of each product becomes a very difficult problem."
Understanding the customer
Working with Assistant Professor Franco Berbeglia of Carnegie Mellon University and Professor of Industrial and Systems Engineering Pascal Van Hentenryck from Georgia Tech, Associate Professor Berbeglia found it is far more effective for merchants to first segment their customers by characteristics such as age income, location or gender, and then show each consumer a ranking of different products based on their segment to maximise the chances of a purchase.
"What we have done is come up with a simple, rule-of-thumb strategy that, while not perfect, works well in practice," says Associate Professor Berbeglia.
In their December 2021 paper, Market Segmentation in Online Platforms, published in the European Journal of Operational Research, the co-authors present a simple segmentation strategy by working out the "average quality ranking" based on showing customers the products deemed the highest quality by their particular segment, taking into account factors including the number of past purchases, number of favourable reviews, the product's return rate, likes on social media and the product appeal or branding quality.
For merchants, the key takeaway is that it's not necessarily the best strategy to place their overall most popular products first.
"To become successful online, merchants need to really understand their customer base, which includes appealing to different consumer segments," Associate Professor Berbeglia says.
"Not everyone will want to purchase the most popular product – in fact, in some cases, the display of past purchases may reduce the number of sales by confusing consumers about which products to try."
While a largely homogenous customer base will respond positively to social signals such as customer reviews, a customer base with many different types of people (such as the fashion market, for example) should be analysed more carefully since social signals may be detrimental when they aren't used appropriately.
"Most people reveal a lot online, and if merchants invest in finding the types of consumers they have, they can more successfully adapt our algorithm," he says.
Associate Professor Berbeglia says the group's research will be most useful for emerging retailers, rather than the industry leaders.
"It would be impossible to develop a model that perfectly encapsulates what Amazon, Google Play or Netflix do," he says.
However for most retailers, the simple strategy is a good starting point that, with a little extra research and some tweaking, can be adapted to almost any context.
To read the full research paper, visit Market Segmentation in Online Platforms.
For more analytics information and research, visit our Centre for Business Analytics page.