Machine Learning — a kind of artificial intelligence — has been applied to all kinds of business processes over the past few years. By providing insights and accurate predictions, ML-based applications can improve customer service, increase product quality and cut costs. But for most companies, these kinds of projects have been out of reach. They haven’t had access to scarce ML expertise and couldn’t afford the time and computing power needed to develop effective models using the tools available.
All that is changing with the launch of Google AutoML. AutoML Tables lets business teams who aren’t ML experts — but who understand their data —automatically build and deploy high-quality ML models in days rather than weeks. More than that, AutoML Tables works with the structured data that underpins core business processes — often a firm’s largest cost centres or revenue generators. These are areas where even small improvements can have a significant business impact.
This approach to machine learning is already paying dividends for early adopters in a range of industries:
- Insurance: A large healthcare insurer used past claims data to train a model that could identify possibly fraudulent claims for further investigation. The company detected 20% more fraudulent claims using the model than with its previous fraud detection systems. With the company paying out the equivalent of more than £500 million a year, the result was a significant decrease in the cost of claims
- Media and Entertainment: An Australian pay-to-view sports broadcaster had acquired the rights to show cricket, which had previously been on free-to-air channels. To attract subscribers, the company used a year’s worth of cricket statistics, with 83 variables, to train a model that could predict the chance of a wicket falling in the next five minutes. As well as pushing these predictions to its own mobile app, the company incorporated them in online ads and electronic billboards, and integrated them with Google Assistant for a personalised experience for fans. The result was a 150% increase in subscribers per $ spent on marketing and an 18% increase in sales, while brand recall doubled.
- Retail: an online marketplace wanted to advise sellers on the best price for each item being listed, so they wouldn’t lose out on revenues by pricing too low, but wouldn’t miss out on sales altogether by pricing too high. With AutoML Tables, the development team was able to create a high quality model, that could automatically suggest suitable prices to sellers, with just one hour of configuration and training.
If you’d like to find out more about how AutoML brings the benefits of machine learning within reach of your business, come and talk to the data analytics experts in our GCP team.