Traditional business analytics are not enough to predict the best course of action
Most organizations know that there’s huge value in using business analytics to parse their data. But many are just starting to see how much more they can get by applying machine learning (ML) models to that same data.
Traditional business analytics tools create helpful dashboards and reports that explain what has happened in the past and can be good for some forward-looking what-if analyses. But to get a much fuller projection into the future, properly applied ML, in conjunction with other analytics, brings a whole new level of predictive intelligence to bear.
Scenarios differ depending on the nature of the data and who is available to work with it. For the many organizations already running data warehouses, the advantages of using ML as well as other embedded analytics tools on their data can be significant. That is especially true if companies want to enable not-overly-technical business people to drive research. To do so successfully requires moving beyond just self-service analytics to fully self-service data warehouse operations.
Business analysts accustomed to self-service capabilities can easily use tools built into Oracle Autonomous Data Warehouse to load and transform the required data and build a consistent business model that all analysts can use. They can then deploy an insights tool to examine their data and get visualizations of corner cases for further examination. Under the covers, that insight tool is using machine learning, but they may want to go beyond those built-in capabilities to build custom models to address their specific business problems.
For example, marketers might want to look at purchasing patterns to see which steady customers are most likely to buy a particular up-sell or cross-sell offer. Or, on the flip side, they might want to identify which not-so-happy customers are most likely to leave for a competitor, and what new product or service might keep them from doing so.
These marketing pros, who fall into the category of what some call “citizen data scientists,” know what’s in their data mart but have little to no experience building a predictive model.
With machine learning built into Autonomous Data Warehouse, they can deploy Oracle’s AutoML UI feature to automatically create and test models to further analyze that data. AutoML can generate a report detailing which model will likely work best for the kind of results required.
While ML tools are, in and of themselves, useful, using them in conjunction with other toolsets such as graph and spatial analytics adds even more value. A spatial engine, for example, will show where a given event or transaction happened; a graph engine will show customer relationships. Adding these results as inputs to ML models can improve the quality of predictions.
These on-board tools combined with AutoML help to automate the tricky parts of a full data analysis process and at least put the marketing team on the road to a successful campaign. The team can then take their findings to data scientists for further guidance or validation.
Note that in the above example, the relevant data stay in the Autonomous Data Warehouse as different tools work their magic on it. In the alternative scenario the marketing team downloads subsets of data to an array of outside ML and other tools for analysis.
There are several key issues that arise from moving data into remote specialty analytics databases or engines. It takes time to move data around and it requires significant expertise which negates the idea of easy self-service. It also requires lots of additional services to integrate, manage and operate. A serious consequence of moving data is that the marketing team is performing analysis typically on days old data which can lead to predictions that are not very useful. For example, the up-sell or cross-sell offer is no longer relevant, or the not-so-happy customer may have already moved on.
Those factors add complexity and cost to the process. Even if the user moves results back to a single location, it’s still hard to do integrated (as opposed to separate, stand-alone) graph, spatial, and ML analytics.
And, the coup de grace is all of that data movement opens up significant security risk.
It is far easier to secure data when it stays put and the relevant on-board tools are used to massage it, than it is to securely ship it around to various remote analytics tools
For most businesses, the safer and more productive option is applying different analytics engines that are built into the data warehouse to the information stored in that data warehouse.
To sum up, companies that use Oracle Autonomous Data Warehouse’s embedded tools can mitigate the pitfalls of data fragmentation and security vulnerabilities that moving data around can cause. Using Oracle’s embedded data tools to enable analyses can be safer and more efficient than shipping data out to an array of remote specialty analysis engines. And, combining results from a raft of different-but-embedded analytics tools will make ML-enabled predictions more timely and accurate thus boosting prospects for success.
