Have you ever ever considered doing analytics in digital, augmented, or combined actuality? What advantages might it carry to finish customers? We explored potential use circumstances for this expertise within the knowledge analytics house and applied a working demo in the course of the newest GoodData hackathon. On this article, I’ll briefly cowl what we did, in addition to hyperlink you to the demo software and its supply code.

First, Let Me Ask You for Suggestions

Just lately I offered a brief model of this text on the annual DATA mesh meetup and requested for the next suggestions:

The LinkedIn ballot — will probably be open for the following two weeks, please, vote!

I used to be fairly stunned how many individuals had a non-gaming expertise with not solely digital but additionally augmented and combined actuality! So many individuals approached me in the course of the after-party to debate potential use circumstances.

The Outcomes First

To be trustworthy, we strongly doubted that we’d be capable to ship something useful, however we did! The video demonstrates what we achieved. Strive it in your VR headset (or a minimum of in any internet browser).

Trace: go searching to search out the info visualization.

The subsequent chapters describe what use circumstances we found, how we applied the end result, and what execs and cons we discovered within the space of VR growth for knowledge analytics.

The Story Behind This Endeavour

As an analytics supplier emphasizing the Headless BI idea, we at all times search new knowledge consumption use circumstances.

To date we kind of touched on the next:

Like yearly, the time got here for one more hackathon in GoodData. This time round, I needed to discover one thing fully new, one thing we hadn’t touched in any respect but.

I fell in love with Oculus Quest (each one and two, ready for 3). In addition to video games, I additionally performed with it to attempt to implement one thing significant. So this yr’s matter for the hackathon was clear. Nonetheless, I’m an knowledgeable within the space of databases and the backend, so I recruited Dan Homola, our well-known front-end architect, to assist me. Fortunate me!

Discovery First!

This was the funniest a part of the hackathon. We had been speaking hours about potential use circumstances. Neither I nor Dan is an knowledgeable UX designer — we must always observe up on that later with correspondingly expert individuals! Anyway, we got here up with the next concepts.

Leverage 3D House

The already identified sorts of 3D knowledge visualizations are apparent candidates. Nonetheless, we really feel there may be extra potential within the VR/MR/AR house, the place you possibly can transfer and work together in additional methods than simply interacting in an internet browser.

Interactivity

There are such a lot of use circumstances! Numerous knowledge filters, constructing 3D visualizations with a drag-and-drop expertise.

We had been discussing one specific very fascinating use case — the interplay between semantic fashions and visualizations. In browsers, the semantic mannequin is often represented as a flat (typically hierarchical) listing of entities (attributes, information, metrics, …). Think about what you might obtain in a 3D house! Visualize a bodily/logical knowledge mannequin in a 3D house, and drag and drop entities from it, whereas at all times with the ability to see the encompassing context.

Furthermore, GoodData supplies a novel characteristic: filtering the listing of legitimate entities based mostly on the listing of entities already used within the visualization. Think about how this filtering could possibly be visualized in 3D, the place the still-valid a part of the 3D mannequin can be highlighted based mostly on relationships amongst entities.

Superior Visualizations

Once more, so many choices. We explored sources like Plotly 3D charts. An apparent candidate is any type of spatial analytics, like Google Earth on steroids.

Speech-To-Textual content

The apparent use case for the VR (or AR) headset.

Relating to analytics, we just lately printed an article describing an integration of analytics with Slack that leverages pure language question (NLQ). What if headset customers might ask for a report with their voice, utilizing pure language? Underneath the hood, the request can be reworked into an inside report definition utilizing correctly pre-trained Giant Language Fashions (LLM). Then the interior report definition can be despatched to the backend, and the corresponding visualization can be rendered in digital actuality. There are different use circumstances, like exploring semantic fashions.

Augmented/Combined Actuality

Personally, I contemplate this selection probably the most promising one.

We’re near AR headsets, that are light-weight sufficient and with batteries that final for the entire day. Or a minimum of it appears to be like like we aren’t that removed from it. As soon as the value of such glasses is just like the present worth of Oculus Quest, I count on their huge growth: AR headsets might substitute cellphones.

Being able to embed analytics into numerous AR merchandise provides you with a aggressive benefit. Personally, we had been occupied with implementing one thing like Tinder++, or a minimum of detecting individuals in our workplace and displaying their private analytics based mostly on knowledge we now have in our inside programs. I do know, too bold, however positively thought-provoking (and scary, I like to recommend studying the guide Zero by Marc Elsberg).

Prototyping

Anyway, we’re prepared. With only a few traces of code (hackathon high quality, please, be forgiving), you possibly can acquire report outcomes with our GoodData UI.SDK and embed it into VR.

Relating to developer velocity with units like Oculus:

• We misplaced plenty of time attempting to make use of the Codesandbox service. Refreshing the (publicly out there) software wasn’t deterministic, usually we noticed the outdated model time and again.
• We lastly linked Oculus to our inside community, and on to my laptop computer, the place I began the applying regionally (and frequently git-rebased it to undertake Dan’s modifications).

Relating to CI/CD: I made a decision to encapsulate the VR software to an current CI/CD knowledge pipeline (extra on that e.g. in Extending CI/CD knowledge pipeline with Meltano).

Conclusion: with a single merge request you possibly can prolong the info pipeline and the (VR) software and you may persistently ship it to varied environments.

Frontend

We discovered just one related (open supply) framework, which we might use to ship a minimum of one thing in beneath 24 hours: A-FRAME. A-FRAME supplies a WebVR (browser) expertise.

They supply fairly good documentation, although it’s typically barely deceptive.

Sadly, their SDK aframe-react appears to be not beneath energetic growth, and it isn’t working correctly. So we needed to do all the pieces with pure HTML and Javascript.

Principally, builders can generate A-FRAME HTML-like entities representing numerous sorts of objects in 3D house. It was pretty simple to generate knowledge factors as spheres positioned based mostly on underlying knowledge, with their measurement and coloration additionally being derived from the info. Fortunately, I discovered a supply code implementing thumbstick motion as an alternative choice to bodily motion (which nonetheless is smart with Oculus Quest all-in-one units).

The appliance is part of the gooddata-data-pipeline repository and is persistently delivered to 2 separate environments with the free-forever Render service (very easy to arrange!):

• vr_analytics_stg when a change is merged to the principle department. Working on prime of the GoodData staging atmosphere (workspace).
• vr_analytics when a change is merged to the prod department. Working on prime of the GoodData manufacturing atmosphere (workspace).

You will discover the whole supply code right here.

Backend

This half was too simple. I reused our public CI/CD knowledge pipeline demo.

I made a decision to make use of FAA (Federal Aviation Administration) knowledge about flights in the USA. The information is obtainable in our AWS S3 public bucket within the type of CSV recordsdata. Including AWS S3 CSV as a brand new supply of Meltano was nearly including a number of traces within the meltano.yml file. Then I wrote a number of dbt SQL fashions to remodel the info right into a mannequin that was prepared for analytics.

The GoodData logical knowledge mannequin is generated from dbt fashions (dbt-gooddata plugin), so no work was required there. Lastly, I needed to create a report appropriate to be visualized in 3D house. I made a decision to slice numerous metrics by two date granularities — quarter/yr and day of week visualized as X/Y axis. The thought was to investigate the general historical past of flights AND per day within the week. The Z axis, radius, and coloration of every knowledge level rely on metric values.

Conclusion

The Good

It’s wonderful what it’s doable to ship in beneath 24 hours!

The Dangerous

A-FRAME remains to be not mature sufficient. Nonetheless, it’s actively developed, so we are able to count on a greater expertise sooner or later. Anyway, for a severe production-ready venture, I might seek for a extra mature framework, possibly one thing like Unity.

And the Ugly

What actually hampered us initially, was the poor developer velocity to start with attributable to our attempting to make use of CodeSandbox.

We had been lastly capable of enhance it by connecting the Oculus headset to my native laptop computer by means of our inside community as a result of the CodeSandbox service didn’t work for us.

… and by the way in which, we gained! 🙌

Need To Strive It for Your self?

If you wish to put any of those rules to work, why not register for the free GoodData trial and check out making use of them your self?