Revenues, user growth, and cash flow are all vital business metrics. When a business reaches scale, they are no longer kept in Excel. Businesses track all that data in databases or database-powered products. Then, they analyze this data using BI tools (like Tableau and Looker). You enter a query (eg count data points with certain conditions) that performs joining of data silos, does summation, time series prediction, etc. This process is fairly manual; the data analyst has to intuit dependencies and correlations in the data and build a mathematical model that maps observable data to desired predictions. This process takes time and attention — the analyst cannot see all the patterns involved and isn’t perfect in the execution.
2 years ago, Salesforce bought Tableau. Last year, Google bought Looker. Similar purchases (Looker and Tableau were direct competitors). What’s the rush? At first glance, it looks like a customer acquisition — buyer gets a lot of new customers because it has a more powerful sales force. But, there’s more to it.
Both Google and Salesforce are powerful computational platforms capable of analyzing vast amounts of data. They also both have big machine learning (ML) teams. Both launched so-called AutoML products last year. What these do is suggest computational models that are best for a given task — making a particular type of input data into particular type of output data (eg predict sales based on marketing spend on certain channels). Typically, a data scientist uses his experience to pick the right model and train it on the training data (imagine a weighted multiplication filter — training picks the right weights). This process is laborious, error prone and time-consuming but the result is powerful because it allows you to make use of hidden causalities in the data. AutoML solutions pick these models automatically so no data scientist is involved — just load the data and have AutoML do its magic in finding the right mapping model (think interpolating a function). Now imagine what can be done when you analyze core business data using automated machinery rather than a few data scientists? Using the textile analogy, you could be sowing a shirt by hand, or you could be sowing them by the thousands in a factory… AutoML+BI = dangerously powerful view into any business’s health or direction. With AutoML for computation and customer data from those BI acquisitions, Google and Salesforce can analyze not just a single’s business health, but know exactly where entire industries are headed). To appreciate how dangerous, just think of Amazon’s favorite practice of offering an ecomm platform to small businesses and then undercutting them when it sees their product selling well. Here, Google could predict and manipulate entire industries. And you thought censorship was bad.