Once a machine learning pipeline is built, predictions can be made off of that pipeline. Typically there are two broad modes of leveraging the pipeline –
- Batch – The machine learning pipeline processes all users at once and stores the model output in a DMP, CDP, or Data Warehouse. These user scores can then be accessed and used for marketing, product, advertising, and more.
- Real-Time – The machine learning processes individual users in Real-Time. Often real-time requires pipelines to run on the “edge” – in other words close to the user (in their browser for instance or on services which are close to the browser).
Batch is a more standard method for leveraging machine learning. The benefits of batch are that, because there are no real-time speed constraints, more sophisticated pipelines, feature engineering, and models can be built. All things being equal, a more sophisticated machine learning pipeline will perform better than a real-time pipeline.
However, the world is changing and batch machine learning isn’t the right approach for a broad swatch of business problems. In particular, businesses will increasingly rely on machine learning to make decisions during a session with the depreciation of 3rd party cookies (the cookie apocalypse!). However, Real-Time is often more difficult to implement and often implicitly requires less sophisticated machine learning pipelines. (At Vidora we are taking the approach of combining both batch and Real-Time in order to obtain both amazing performance and Real-Time performance).
When is Real-Time the way to go? Here are a couple of examples of when Real-Time is the right choice for a business –
- Dynamic Decisioning – Most decisions that impact a user journey in-product are good candidates for real-time. Typically product decisions will need to leverage the latest user behaviors in deciding, for instance, whether or not to show a user a paywall.
- Contextual Targeting – Targeting a user based on the context of their page is also a good candidate for Real-Time. For instance, if a user is browsing a page on stocks it might make sense to promote a business subscription to the user.
Real-Time Machine Learning is a really exciting area for businesses to leverage in order to increase conversions and maximize the value of their data. If you are interested in learning more, reach out to us at email@example.com and we can talk about some of the initiatives Vidora has in the area of Real-Time Machine Learning!