What is OAS/OAC's functionality to support agile or citizen-BI use cases?

Doug Newton

Hello,
A quetion about what Oracle's strategy and available solution is in this space:

Users of Tableau, Power BI, Sisense, Qlik are used to scenarios where they don’t have an industrial-strength semantic layer… which is bad from some perspectives, but generally makes it easier for the BI tool user to just do what they need to do, without having to coordinate with a centralized semantic admin team (in our case, in IT). They just extract data sets into a BI-local data structure, and work from there.

And they don't want to work in a dev environment and promote it from dev→test→prod. They just want to do their business-driven BI work in prod using whatever data sources they'd like to cobble together.

For these more-agile scenarios, what would Oracle offer? Is it DV and datasets? 

And conversely, when these scenarios do occur, what functionality is available to system admins, to understand what is happening in the environment? So far, it seems like when power users do their own thing with datasets, we don't have good visibility into those data sources. We need some way to merge these agile use cases with enterprise oversight, so that some data sets can be migrated into the RPD over time.

Thank you for your insights.

Gianni Ceresa

Hi Doug,

Welcome to the Oracle Analytics community and forum.

Yes, as you say DV is the Oracle answer to power users who want to do their own cooking without being limited to a centrally managed semantic layer (that's the semantic model in Oracle Analytics).

In DV you have everything for a user to do a lot, sometimes too much, without needing any IT support. Of course you have permissions allowing you to "limit" a little bit what users can do, but they can connect to various sources and create datasets, or load them from files. They can use those datasets in workbooks to build their own analysis and reporting tasks. They can connect the datasets together to build more complex workbooks or, when they really need some ETL-like features, using dataflows to generate new datasets.

The challenge with DV and those power users and the freedom they can have is exactly your second question: how can a sys admin have an idea of what's going on? In my opinion this is where Oracle Analytics DV has a weak answer out of the box… You can retrieve some metadata allowing you to keep an eye on what's going on. Oracle did publish a plugin that could help you a bit in this: the Governance plugin, https://blogs.oracle.com/analytics/post/oracle-analytics-governance-plugin .

It will help you do the work of identifying what kind of source is often used by your users, it could give you a hint on what kind of content you should consider making available as part of your RPD (the semantic model). But it will be challenging, because if your users can connect to sources and build their own dataset, nothing prevent 10 users from having 10 connections, with 10 different names, with 10 different datasets with different column names, even if they are all doing the same exact job. Also, the challenge of power users is that while they should be doing the same thing, they all do it slightly differently because of their understanding of the meaning of data etc.

In OAS you will have a lot more access to analyse all the metadata of each connection, dataset, dataflow, workbook and identify more easily the topics that could be worth adding to your RPD. In OAC you have a bit less visibility (you can export a snapshot and work on that file to have access to all the details, but it's a more annoying job, involving a lot more manual work, while in OAS you just query your database to get all those answers).

This is my opinion on the topic. I observed over the year that this is a quite subjective topic: some believe DV already provide all the metadata, while in reality when you start digging you discover that some tools are missing and it is a lot of DIY (even the governance plugin has its limits…).

SteveF-Oracle

Hi Doug,

Welcome to the Oracle Analytics Community!, We are glad you joined us here.

You are correct that using Workbooks and Datasets in (/dv) allows you to perform self-service discovery and analysis without a semantic model using your uploaded datasets. The nice thing about Oracle Analytics is that you can have either, or, both options at the same time.

You would use Oracle Analytics Cloud or Oracle Analytics Server for production use, and you can use Oracle Analytics Desktop as local non-production companion for self discovery.

For Oracle Analytics Cloud there is a public extension Governance plugin that you can use in addition to Usage Tracking.

I hope that helps get the conversation started, other comments are encouraged and welcomed.

Doug Newton

Thank you for the nuanced and thorough answer. It does seem there's more work to do to enable BI development for the masses, while still providing some amount of visibility for proper data governance.

Gianni Ceresa

The "issue" is that governed analytics and a DIY analytics (or self-service analytics) coexist on the same environment, overlap a lot, while not sharing much.

This make it very challenging to define a process to identify content to move from the users' made analytics to the governed one. Mostly because it depends a lot if your users do analyses just for them, or if they can share them.

If your users can share them, you could just trust usage tracking and take objects/topics based on consumption: the more users look at a given workbook, the more that is a good candidate to be standardised and integrated in your enterprise semantic model, to make sure everybody see trustworthy data.

If your users build content mostly for their own usage, then you need to pay attention to the content of objects instead of just counting number of viewers. And the challenge here is on what is centrally managed (database connections, or some datasets) and what is free. The more power, permission, freedom, your users have, the more challenging it is to find candidates for content to be moved to the enterprise semantic model.

But this isn't really a tool weakness, it's the limit of the methods: if you let your users free, you will struggle to move content easily under the control of central IT.

You could also keep an eye out on the importance of content: what workbook is used to make important decisions in your company? That is also a good candidate to become governed, avoiding as much as possible users errors or mistakes.

All in all, it's always subjective: every company has a different way to handle this, there isn't a unique valid rule you can apply. It depends too much on your own processes, your users, what they are allowed to do, their skills etc.