- Large Chinese vending machine company with over 5,000 Internet-enabled vending machines in China
- Collected 3 years of IoT data but struggles to turn these data into strong business opportunities
- Marketing analysts bring in some new revenue but the costs are too high
- Building own AI solution would cost at least 6 months and hundreds thousands of dollars
- Using Decanter™ AI as a cost-effective, self-learning, predictive solution
- Automated price changes, no human interference required
- MoBagel’s Solution was implemented in just 2 weeks
- $1M increase in revenue over first 3 months
A large Chinese vending machine company has dispatched over 5,000 Internet-enabled vending machines across China. These machines sell different hot and cold beverages and the beverage prices can be adjusted remotely based on expected demand.
While the company has collected 3 years of IoT data, it is unable to convert these data into good business strategies. In an attempt to utilize these data, the CEO hired a team of marketing analysts to manually adjust the prices of its beverages. For example, when the forecasted demand for a certain beverage is high, the analysts will increase the price on the machine, and vice versa.
Although this plan was able to lead to slight positive increases in revenue, the CEO quickly realized that depending on marketing analysts is too costly and impracticable. He needed a cost-effective solution that could reduce the amount of human efforts required while being able to more accurately predict the demands of its beverages.
The ideal solution was clear: to implement a cost-effective, self-learning, predictive solution that could translate the machine data into real-time forecasted demand, and then automatically adjust the prices on the machines to increase the company revenue.
However, to build their own prediction solution from scratch, the company would have to hire multiple engineers and data scientists and spend at least 6 months building and testing the solution. This was not a feasible plan due to the extremely high costs associated and the long time that it would take to complete. Thankfully, MoBagel can implement the same solution in under one month and at a fraction of the cost. As an end-to-end solution that can adapt to any IoT industry, MoBagel readily helps companies collect, clean, store, analyze, predict and actionize IoT data, so companies do not have to worry about maintaining its own infrastructure.
To implement MoBagel’s solution, the company only had to insert a few lines of code using MoBagel’s SDK, which then establishes data connection between the vending machines and MoBagel’s cloud. The company then fed its three years of past raw data to Decanter™ AI engine. Decanter™ intelligently filters out noisy and meaningless information and applies deep learning algorithms to build demand prediction models. To improve the accuracy of the model, MoBagel also introduced open source data such as location, weather, and traffic that affects demand. Lastly, when Decanter™ detects any potential changes in demand, it would automatically update the price on the vending machines in real-time via webhook.
MoBagel implemented its solution in two weeks: one week for set up, and another week to build the prediction model. In the first 3 months, the company saw an average $350,000 or 23% increase in monthly revenue. This revenue growth is expected to increase as Decanter™ becomes more intelligent through its self-learning mechanism. In the process, the company also saved hundreds thousands dollars that they would have spent on developing their own solution and from replacing its team of marketing analysts.
These results were made possible because of three powerful features: 1) open source data, 2) real- time price changes, and 3) self-learning automation.
1. Open source data: the introduction of weather and location data greatly improved the demand prediction model. This could not be done manually due to the number of vending machines and different conditions. For example, we found that 1 degree Celsius increase in temperature led to an average 4% increase in demand for cold beverages.
2. Real-time price changes: Decanter™ immediately responds to potential changes in demand, thus capturing revenue that would have otherwise taken marketing analysts hours to identify.
3. Self-learning automation: Decanter™ does not require human interference, which often could lead to human delays, biases, and errors. In addition, since Decanter™ is self-learning, it is able to quickly adapt to unexpected changes in demand, and only become smarter as more data is collected.