Herman Widjaja, CTO, Sr. Vice President, at Tokopedia, explains how the online marketplace ecosystem is using the latest tools in artificial intelligence to innovate and serve its customers with purpose.
Online marketplace Tokopedia utilizes AI to serve over 100 million customers each month. Read below to learn how Tokopedia delivers new innovations to its customers.
getty
I run engineering at Tokopedia, one of the world’s largest online marketplaces. We’re a “Super Ecosystem,” connecting more than 10 million merchants with 100 million customers every month for product selection, payments, and delivery. We are technology and innovation trailblazers, and our mission is to democratize commerce through technology in Indonesia while transforming into an AI-first company.
Indonesia has the world’s fourth-largest population, spread over more than 17,500 islands and spanning 3,200 miles—and Tokopedia serves 99% of the districts nationwide. As you can imagine, organizing effective, timely communications and logistics for such a diverse group of buyers and sellers across a geographically dispersed country is a huge challenge.
It has meant operating as a true AI-first company, using technologies like machine learning (ML) to weave intelligence and automation into our products and processes.
Along the road to AI transformation, we have learned a couple of key principles that have helped us evolve and continuously deliver new innovations for our customers. I am convinced that focusing on these important basics has the power to drive AI growth not only at Tokopedia—but at any company looking to make the shift to AI-first. Here’s what to know.
1. AI-first demands data-first
Table of Contents
Over a decade ago, Tokopedia started as a C2C marketplace. Today, the situation is very different. We are no longer just a marketplace; we have become a technology company that delivers a host of products and services across Indonesia, including shopping, entertainment, fulfillment, new retail, and investment opportunities for Indonesians.
At Tokopedia, we’ve continuously leveraged new technologies such as Google Cloud’s AI Platform to deliver products and services based on the needs of our users and market. And that means using data—and a lot of it.
In the beginning, we used basic analytics to help us identify our efficiencies in areas like search and product discovery. But over the years, we have built up a large amount of data around the buyer and merchant experience about buying habits, payment times, goods storage, and shipping logistics.
Tokopedia’s virtual marketplace hosts various types of businesses, including retailers, restaurants, and even personal care providers. We currently have 30 different payment providers and 13 logistics companies using our platform. Additionally, 86% of our businesses are first-time entrepreneurs, mostly small operations without access to credit.
The more insights you generate, the more customers you attract. And the more customers you have, the more data you have to deliver better AI-powered predictions.
As we have grown in scale and information (across a multitude of data formats), the ability to operationalize data with AI has become critical to helping us establish a consistent cycle of improvement. The more insights you generate, the more customers you attract. And the more customers you have, the more data you have to deliver better AI-powered predictions.
To start, we centralized all of our precious petabytes of information in a data warehouse, in our case, Google BigQuery. We use it for analyzing traffic and transactional data, such as logistics and billing, and creating reports on customer insights. For example, our product managers might use it to create sales forecasts to help them plan their daily, weekly, or monthly promotional budgets.
A multi-cloud data warehouse like BigQuery also helps us deal effectively with scale, which is a real issue when it comes to data and deploying software. Online commerce has to move quickly and flexibly, and a cloud data warehouse needs to scale up to meet opportunities and back down for optimal economics. We not only benefit from serverless features and cost-effective scaling, but choosing a fully managed service also frees up more time to spend focused on things that actually deliver value for our customers.
Related: Retailers find flexible demand forecasting models in BigQuery ML
2. AI should be at the center of your work
Another important aspect—and challenge—of AI-first is finding ways to build AI into everything you do. We use ML to help us predict product categories or provide recommendations based on customers’ purchase history. But we also see the opportunity in AI and ML being applied to more innovative purposes.
For example, AI can be used for alternative credit scoring (ACS) for small businesses to help determine financial reliability, increasing financial inclusion. ACS leverages alternative data, such as social media or electronic transactions, to conduct feasibility studies, which helps unbanked or underbanked groups qualify as partners and unlocks new potential markets.
We are also experimenting with AI-based logistics and warehousing. Analyzing our transactional data for demand prediction for more effective logistics delivery time and cost could be the key to achieving our vision of same-day delivery, even if suppliers and customers are on different islands.
3. AI benefits everyone—not just data scientists
We see too many companies today leaving AI to advanced data science teams and failing to secure engineering involvement. At Tokopedia, we rely on Google Cloud engineers to not only help us architect our systems, but to also train and educate our more than 1,000-person-strong engineering team.
We want data analytics and AI to be core to the development process, empowering our engineers to be responsible for uptime, data quality, security, reliability, and cost.
AI can’t be part of an efficient, ongoing process if it’s sent off to another team to analyze. Our AI projects have helped us to find ways to develop new processes that will support our future growth, across all our teams. We want data analytics and AI to be core to the development process, empowering our engineers to be responsible for uptime, data quality, security, reliability, and cost.
Building AI into our software creation and delivery has already helped us do more for our users. From delivering better search results to optimizing customer experience with better personalization to analyzing reviews with natural language processing, we are learning something new every day. Seeing how people react to products and their search choices has helped us develop better chatbots. In addition, AI is deployed across Tokopedia for fraud detection and behavior-based credit scoring, which is improving the lives of our merchants and, again, increasing financial inclusion.
Ultimately, I believe that companies like Tokopedia that have partners like Google Cloud are leading the way for a much richer technology experience for all types of organizations. We are only successful to the extent that we help the overall market. Looking at how far mobile technologies and the cloud have come in the last decade, there is no doubt that AI tools and practices will become equally useful and widespread. If you respect the basics, there’s no limit to the sophistication that you can achieve.
Learn the skills necessary to keep up with the pace of innovation. Join us for the Applied ML Summit to learn how to apply machine learning technology in your projects and business.
