SINGAPORE: Brands can learn a lot about how to utilise data modelling, smart AI, weighted attribution and shopping experiences from Asia’s fast growing e-commerce platforms such as Lazada, Carousell and Zalora.
Weighted attribution, for example, can be helpful to plan ahead for acquiring new customers, according to Eugene Yan Ziyou, Lazada’s data science lead.
“If I show an ad on Instagram, on Facebook, on Twitter, and then you count which ad actually leads you to come to the website, is it the Twitter ad, the last touch, or Instagram, the first touch? Or is it some combination of Instagram, Facebook and Twitter?” he asked an audience at General Assembly’s recent ‘Talk Data to Me: the Future of E-commerce’ event.
“We’ve built a Markov decision process that tries to identify how much weight to give to each of these channels, so that way, we know which channel is more suited for which kind of segment,” he explained. (For more, read WARC’s report: What brands can learn from Asia’s online marketplaces.)
“Maybe the Facebook channel has higher quality users who tend to spend more, or it could be that Instagram has younger users.”
Zalora uses A/B testing to get the most from its marketing strategies. “Every feature release, every product enhancement that we make, we always think beforehand, ‘How do we test it so that we know that we are doing better?’” said Nikhil Sahai, the regional head of marketing analytics at the platform.
“A/B testing is really the way that we see how we can prove to the management team that we are making the right decisions with our products,” he added.
Brands should also manage risk by running regular ‘health checks’ on their platforms or e-commerce stores – especially after launching a new feature: Yong Chang Yi, the business intelligence manager of C2C marketplace Carousell, looks at the impact on core marketplace metrics like transactions or listing views.
“Are there any abnormal behaviours, or if there is a down trend that might coincide with a product release,” he said.
There’s predictive chat, too. “We focus on optimising the chat experience, because that is the medium through which people discuss and arrive at a way of transacting,” said Yong.
“We are exploring how we can have smarter chat replies (to reduce) the manual effort that the seller needs to keep repeating the same messages over and over again.”
Weighted attribution, for example, can be helpful to plan ahead for acquiring new customers, according to Eugene Yan Ziyou, Lazada’s data science lead.
“If I show an ad on Instagram, on Facebook, on Twitter, and then you count which ad actually leads you to come to the website, is it the Twitter ad, the last touch, or Instagram, the first touch? Or is it some combination of Instagram, Facebook and Twitter?” he asked an audience at General Assembly’s recent ‘Talk Data to Me: the Future of E-commerce’ event.
“We’ve built a Markov decision process that tries to identify how much weight to give to each of these channels, so that way, we know which channel is more suited for which kind of segment,” he explained. (For more, read WARC’s report: What brands can learn from Asia’s online marketplaces.)
“Maybe the Facebook channel has higher quality users who tend to spend more, or it could be that Instagram has younger users.”
Zalora uses A/B testing to get the most from its marketing strategies. “Every feature release, every product enhancement that we make, we always think beforehand, ‘How do we test it so that we know that we are doing better?’” said Nikhil Sahai, the regional head of marketing analytics at the platform.
“A/B testing is really the way that we see how we can prove to the management team that we are making the right decisions with our products,” he added.
Brands should also manage risk by running regular ‘health checks’ on their platforms or e-commerce stores – especially after launching a new feature: Yong Chang Yi, the business intelligence manager of C2C marketplace Carousell, looks at the impact on core marketplace metrics like transactions or listing views.
“Are there any abnormal behaviours, or if there is a down trend that might coincide with a product release,” he said.
There’s predictive chat, too. “We focus on optimising the chat experience, because that is the medium through which people discuss and arrive at a way of transacting,” said Yong.
“We are exploring how we can have smarter chat replies (to reduce) the manual effort that the seller needs to keep repeating the same messages over and over again.”
Sourced from WARC
